CTL.

SC4x – Technology and Systems

Key Concepts Document

This document contains the Key Concepts for the SC4x course.

These are meant to complement, not replace, the lesson videos and slides. They are intended to
be references for you to use going forward and are based on the assumption that you have
learned the concepts and completed the practice problems.

This draft was created by Dr. Alexis Bateman in the spring of 2017. It was updated Fall 2017 by
Mr. Ahmed Bilal.

This is a draft of the material, so please post any suggestions, corrections, or recommendations
to the Discussion Forum under the topic thread “Key Concept Documents Improvements.”

Thanks,

Chris Caplice, Eva Ponce and the SC4x Teaching Community
Fall 2017 v2

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
1
 Table of Contents
Introduction to Data Management ....................................................................................................... 4
Data Management ................................................................................................................................... 4
Querying the Data .................................................................................................................................... 5
Data Modeling ..................................................................................................................................... 7
Relational Models .................................................................................................................................... 7
Designing Data Models ............................................................................................................................ 8
Relationships and Cardinality ................................................................................................................... 8
Keys .......................................................................................................................................................... 9
Database Normalization ..................................................................................................................... 11
Objectives of Normalization ................................................................................................................... 11
Summary of five normal forms ............................................................................................................... 12
Client Server Architecture ...................................................................................................................... 13
Database Queries ............................................................................................................................... 14
Structured Query Language ................................................................................................................... 14
Creating Databases and Tables ............................................................................................................. 16
Database Conditional, Grouping, and Joins ......................................................................................... 18
Database Conditional Clauses ................................................................................................................ 18
Sorting and Sampling Data .................................................................................................................... 20
Joining Multiple Tables .......................................................................................................................... 21
Topics in Databases ............................................................................................................................ 23
Introduction to Machine Learning ...................................................................................................... 29
Overview of Machine Learning Algorithms ............................................................................................ 29
Model Quality ........................................................................................................................................ 30
Machine Learning Algorithms ............................................................................................................. 32
Dimensionality reduction ....................................................................................................................... 32
Principal component analysis (PCA) ....................................................................................................... 32
Clustering ............................................................................................................................................... 33
Classifications ......................................................................................................................................... 33
Comparing predictor accuracy ............................................................................................................... 35
Sensitivity and specificity ....................................................................................................................... 35
Supply Chain Systems - ERP ................................................................................................................ 38
Supply Chain IT Systems ......................................................................................................................... 38
Enterprise Resource Planning ................................................................................................................. 38
ERP Communication ............................................................................................................................... 40
The Value of ERP for SCM ...................................................................................................................... 41
Supply Chain Systems - Supply Chain Modules ................................................................................... 43
Advanced Planning Systems ................................................................................................................... 44
Transportation Management Systems (TMS) ........................................................................................ 45
Manufacturing Execution Systems ......................................................................................................... 48

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
2
 Supply Chain Systems - Supply Chain Visibility .................................................................................... 49
Tracking and Tracing .............................................................................................................................. 50
Technologies for Supply Chain Visibility ................................................................................................. 51
Supply Chain Systems - Software Selection & Implementation ........................................................... 54
Architecture ........................................................................................................................................... 54
Cloud Computing .................................................................................................................................... 54
Software Vendor Selection ..................................................................................................................... 57
Implementation ..................................................................................................................................... 58

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
3
 Introduction to Data Management
Summary
Supply chains are moving at ever-faster rates with technology and systems supporting this
movement. Data management plays a critical role in enabling supply chains to move at the
speed and precision they do today, and the need for advanced data management will only
continue.

In recent years there has been an explosion of information and this is especially true in supply
chains. A few examples introduced include Amazon’s massive supply chains selling 480 million
unique items to 244 million customers while UPS is delivering 20 million packages to 8.4 million
delivery points. This information is coming from multiple sources, in addition to sensors, the
“internet of things”, and regulations requiring increasing amounts of information.

All of this information is commonly referred to as the “Big Data” challenge. Data is driving our
modern world, but how can we be sure of it and use it most effectively? As we will review –
data is messy, it requires cleaning and programming. Data is frequently trapped in siloes
coming from different sources, which makes working with it more challenging. In addition, data
is big and getting even bigger daily. The tools we have all become comfortable with
(spreadsheets) can no longer handle that amount of data, so we must use different tools to
enable greater analysis.

To better understand the role of data and how to manage it, the following summaries cover an
introduction to data management, data modeling, and data normalization – to get us started on
a solid ground with handling large data sets – an absolute essential in supply chain
management.

Data Management
In data management supply chain managers will be faced with immense complexity. This
complexity is influenced by the volume (how much), velocity (pace), variety (spread), and
veracity (accuracy). Each of these components will influence how data is treated and used in
the supply chain.

There are several reoccurring issues that supply chain managers must be aware of as they are
working with data:
• Is the data clean?
• Is the data complete?
• What assumptions are you making about the data?
• Are the results making sense? How can I check?

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management
MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
4
 Cleaning data is one of the most important; yet time consuming processes in data analysis. It
can greatly influence the outcome of analysis if not completed properly. Therefore – SC
professionals should always plan enough time for basic data checks (meaning if you get garbage
in, you will get garbage out).
There are several typical checks you should always look for:
• Invalid values - negative, text, too small, too big, missing
• Mismatches between related data sets - # of rows, # of cols
• Duplication – unique identifiers
• Human error – wrong dates, invalid assumptions
• Always explore the outliers – they are the most interesting!

When cleaning data, you should be organized. This means you must make sure to version the
documents you are working with and keep track of data changes.

Querying the Data

Once you have a clean and organized set of data, querying the data can make data extremely
powerful. Querying data refers to the action of retrieving data from your database. Because a
database can be so large – we only want to query for data that fits certain criteria.

There are several basic options that can help you get some quick answers in big data sets, such
as using Pivot Tables:
• There are data summarization tools found in LibreOffice, Google Sheets, and Excel
• They automatically sort, count, total or average the data stored in one table or
spreadsheet, displaying the results in a second table showing the summarized data.
• Very useful in tabulating and cross-tabulating data

No more spreadsheets!
Unfortunately, as we dive deeper into the big data challenge, we find that spreadsheets can no
longer service all of our needs. We have the choice of working with structured or unstructured
data. A database is a structured way of storing data. You can impose rules, constraints and
relationships on it. Furthermore, it allows for:
• Abstraction: Separates data use from how and where the data is stored. This allows
systems to grow and makes them easier to develop and maintain through modularity.
• Performance: Database may be tuned for high performance for the task that needs to
be done (many reads, many writes, concurrency)

Spreadsheets are unstructured data. You have a data dump into on spreadsheet and you need
to be able to do lots of different things. Spreadsheets will always be great for a limited set of
analysis such as informal, causal, and one-off analysis and prototyping. Unfortunately, they are
no longer suited for repeatable, auditable, or high-performance production. Unstructured data
commonly has problems with: redundancy, clarity, consistency, security, and scalability.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
5


Learning Objectives
• Understand the importance of data in supply chain management.
• Review the importance of high quality and clean databases.
• Recognize the power of querying data.
• Differentiate between unstructured and structured data and the need for tools beyond
spreadsheets.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
6
 Data Modeling
Summary
Now that we have been introduced to data management and the issue of big data, we now
deep drive into data modeling where we learn how to work with databases. Data modeling is
the first step in database design and programming to create a model for how data relates to
each other within a database. Data modeling is the process of transitioning a logical model into
a physical schema.

To understand the process of data modeling, we review several components including
relational databases, data organization, data models for designing databases, and what
constitutes a good data model. A data model consists of several parts including: entities and
attributes, primary keys, foreign keys, and relationships and cardinality.

Relational Models
The relational model is an approach to managing data that uses structure and language where
all data is grouped into relations. A relational model provides a method for specifying data and
queries. It is based on first-order predicate logic, which was described by Edgar F. Codd in 1969.
This logic defines that all data is represented in terms of tuples, grouped into relations. There
are several definitions to be familiar with as we reviewed previously with relational models:
• Entity: object, concept or event
• Attribute (column): a characteristic of an entity
• Record or tuple (row): the specific characteristics or attribute values for one example of
an entity
• Entry: the value of an attribute for a specific record
• Table: a collection of records
• Database: a collection of tables

Tables and Attributes

Data in relational tables are organized into tables, which represent entities. Single tables within
a database can be seen as similar to a spreadsheet. However, we use different words to refer to
“rows” and “columns”. Attributes are the characteristics of an entity.
Tables
• Tables represent entities, which are usually plural nouns
• Tables are often named as what they represent (typically plural nouns, without spaces):
e.g. Companies, Customers, Vehicles, Orders, etc.

Attributes
• Characteristics of an entity (table), typically nouns
• Examples in the form of: Table (Attr1, Attr2, ... AttrN), Vehicles (VIN, Color, Make,
Model, Mileage)

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
7

Entity Types and Entity occurrence: an entity is any object in the system we want to model and
store. An entity occurrence is a uniquely identifiable object belonging to an entity type.

Designing Data Models

There are several steps to designing a database to store and analyze data.
1. Develop a data model that describes the data in the database and how to access it
2. Data model defines tables and attributes in the database (each important concept/noun
in the data is defined as a table in the database)

Data models help specify each entity in a table in a standardized way. They allow the user to
impose rules, constraints and relationships on the data that is stored. It also allows users to
understand business rules and process and analyze data.

Rules for a Relational Data Model

There are several rules for relational data model:
• Acts as a schematic for building the database
• Each attribute (column) has a unique name within a table
• All entries or values in the attribute are examples of that attribute
• Each record (row) is unique in a good database
• Ordering of records and attributes is unimportant

What makes a good relational data model? A good relational model should be complete with all
the necessary data represented. There should be no redundancy. Business rules should be
effectively enforced. Models should also be reusable for different applications. And finally, it
should be flexible and be able to cope with changes to business rules or data requirements.

Relationships and Cardinality

When we begin to work with the data – we have to understand how data relates to each other
and data uniqueness of the attributes. Some of this can be managed through entity types and
attributes. Relationships + cardinality = business rules.

Entity and Attributes

An entity is a person, place, thing, or concept that can be described by different data. Each
entity is made of a number of attributes. Entity types should be described as part of the data
modeling process, this will help with the documentation and determination of business rules.

How to draw and entity-relationship diagram:
An ERD is a graphical representation of an information system that visualizes the relationship
between the entities within that system.

• ERD or entity-relationship diagram is a schematic of the database
• Entities are drawn as boxes
V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management
MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
8
 • Relationships between entities are indicated by lines between these entities
• Cardinality describes the expected number of related occurrences between the two
entities in a relationship and is shown using crow's foot notation (see figures below)

Cardinality – crow’s foot notation
General Meanings Mandatory vs. Optional

Domain Validation Entities: Also known as pick lists or validation lists. Domain validation
entities are used to standardize data in a database, they restrict entries to a set of specified
values. They are tables with a single attribute that enforces values of attribute in related
table(s).

Keys
Primary keys are attributes used to uniquely identify a record while foreign keys are attributes
stored in a dependent entity, which show how records in the dependent entity are related to
an independent entity.

Primary key: one or more attributes that uniquely identify a record. The attribute has be
uniquely suited.

Foreign Key: Primary key of the independent or parent entity type is maintained as a non-key
attribute in the related, dependent or child entity type, this is known as the foreign key

Composite key: is a primary key that consists of more than one attribute, ex: charter airline,
every flight has a different number.

Many to Many Relationships: A many to many relationship refers to a relationship between
tables in a database when a parent entity contains several child entity types in the second
table. ex- Vehicle can be driven by many drivers, drivers can drive many vehicles. In this case an
associative table (entity), aka junction table is appropriate where the primary key of parent is
used in primary key of child.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
9
 Referential integrity
Referential integrity maintains the validity of foreign keys when the primary key in the parent
table changes. Every foreign key either matches a primary key (or is null).

Cascade rules: choose among delete options
• Cascade restrict: Rows in the primary key table can’t be deleted unless all corresponding
rows in the foreign key tables have been deleted
• Cascade delete: When rows in the primary key table are deleted, associated rows in
foreign key tables are also deleted

Learning Objectives
• The data model describes the data that is stored in the database and how to access it.
• Data models enable users to understand business rules and effectively process and
analyze data.
• Understand that business rules are imposed on the database through relationships and
cardinality.
• Recognize that data models may vary for a given dataset as business logic evolves.
• Remember that the data modeling process may reveal inconsistencies or errors in the
data, which will have to be corrected before importing into a database.
• Selection of entities and associated attributes from a flat file is not always obvious.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
10
 Database Normalization
Summary

Database normalization, or normalization, is an important step in database management.
Normalization is intrinsic to relational databases and is the process of organizing attributes into
relations (or tables). This process is vital in reducing data redundancy and improving data
integrity. In addition, normalization helps organize information around specific topics that can
be used to digest the massive amount of information in databases into something digestible.

When SC professionals are presented with large amounts of raw data, that raw data may be
stored in a single table, containing redundant information or information about several
different concepts. The data can be separated into tables and normalized to allow for better
data handling and comprehension. To get to this place, updating a data model can be done
collaboratively during meetings and discussions to define the business rules. During updates,
normalization prevents mistakes and data inconsistencies. Normalization helps prevent
redundancy, confusion, improper keys, wasted storage, and incorrect or outdated data.

Objectives of Normalization
1. To free the collection of [tables] from undesirable insertion, update and deletion
dependencies.
2. To reduce the need for restructuring the collection of [tables], as new types of data are
introduced, and thus increase the life span of application programs.
3. To make the relational model more informative to users.
4. To make the collection of [tables] neutral to the query statistics, where these statistics
are liable to change as time goes by.

**Remember our relational model definitions
• Entity: object, concept or event
• Attribute (column): a characteristic of an entity
• Record or tuple (row): the specific characteristics or attribute values for one example of
an entity
• Entry: the value of an attribute for a specific record
• Table: a collection of records
• Database: a collection of tables

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
11
 Summary of five normal forms
1. All rows in a table must contain the same number of attributes; no sub-lists, no
repeated attributes.
2. All non-key fields must be a function of the key.
3. All non-key fields must not be a function of other non-key fields.
4. A row must not contain two or more independent multi-valued facts about an entity.
5. A record cannot be reconstructed from several smaller record types.

Normal Forms
First Normal Form – the basic objective of the first normal form (defined by Codd) is to permit
data to be queried and manipulated, grounded in first order logic. All rows in a table must
contain the same number of attributes; no sub-lists, no repeated attributes, identify each set of
related data with a primary key. First normal form can make databases robust to change and
easier to use in large organizations.

Second Normal Forms – must first be in first normal form, all non-key fields must be a function
of the primary key; only store facts directly related to the primary key in each row.

Third Normal Form - must first be in second normal form. All the attributes in a table are
determined only by the candidate keys of the table and not by any non-prime attributes. Third
normal form was designed to improve database processing while minimizing storage costs.

Fourth Normal Form - must first be in third normal form. A row should not contain two or more
independent, multi-valued facts about an entity. Fourth normal form begins to address several
issues when there is uncertainty in how to maintain the rows. When there are two unrelated
facts about an entity, these should be stored in separate tables.

Fifth Normal Form - must first be in fourth normal form. A record cannot be reconstructed
from several smaller record types. Size of this single table increases multiplicatively, while the
normalized tables increase additively. Much easier to write the business rules from the three
tables in the fifth normal form, rules are more explicit. Supply chains tend to have fifth normal
form issues.

Normalization Implementation Details

Normalization ensures that each fact is stored in one and only one place, to ensure data
remains consistent. Normalizing the data model is a technical exercise. It does not change
business rules! However, through the process of meetings and decisions it may help the rules
be further defined through review. Care in data normalization is needed to preserve data
quality. There are times when normalization is not an option – this happens when there are
large, read only databases for report generation of data warehouses.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
12
Client-server model
Client Server Architecture
Client Server Model
•  Clients can connect to servers to access a specific service
Clients can connect to servers to access a specific service using a standardized protocol, see
figure below.
using a standardized protocol

Web
Applica9on
Client

Database User MySQL

Interface Client Database
Server

Analy9cs
Client

46

Database Servers
Databases are hosted on a server and not usually accessible through a file system or directory
structure. The main options for hosting a database is: on a single server, in a database cluster,
or as a cloud service. All of these systems are designed to abstract the implementation details.
A client has software that allows it to connect and communicate with the database server using
a standardized protocol. There are many different user interfaces for many databases.
Databases can be accessed remotely or on the Internet.


Learning Objectives
• Identify and understand database normalization.
• Review why we normalize our data models.
• Understand the step-by-step process of data normalization and forms.
• Learn and apply how we normalize a relational data model.
• Recognize the drawbacks of normalization.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
13
 Database Queries
Summary
As we continue our discussion of database management, we dive into the issue of database
queries. The ability to make effective queries in a large database enables us to harness the
power of big data sets. SQL (Structured Query Language) is a language and contains the
commands we use to create, manage, and query relational databases. As in all technology and
systems applications, there are a multitude of vendors who offer SQL variations, but in general
they have a common set of data types and commands. SQL is portable across operating systems
and in general, portable among vendors. Having covered the commonly used data types in
previous lessons, in this next section we will cover very commonly used queries.

Structured Query Language

SQL is used to query, insert, update, and modify data. Unlike Java, Visual Basic, or C++, SQL is
not a complete programming language; it is a sub-language of approximately 30 statement
types. It is generally embedded in another language or tool to enable database access. A few
definitions we need to be aware of as we explore SQL are:
• Data definition: Operations to build tables and views (virtual tables) 

• Data manipulation: INSERT, DELETE, UPDATE or retrieve (SELECT) data 

• Data integrity: Referential integrity and transactions enforce primary and foreign keys 

• Access control: Security for multiple types of users 

• Data sharing: Database accessed by concurrent users 


A few issues to make note of as you work with SQL is that it has several inconsistencies. For
example, NULLs can be problematic and we will explore that later. In addition, when working
with SQL it is important to recognize that it operations on declarative language, not procedural
language. This means that you write the command in such a way that describes what you want
to do, not HOW you want to do it. It is left up to the application to figure it out.

Variations among SQL Implementation

Because different databases use SQL, there can be variation in how SQL is implemented. The
variations include:
• Error codes 

• Data types supported (dates/times, currency, string/text variations) 

• Whether case matters (upper, lower case) 

• System tables (the structure of the database itself) 

• Programming interface (no vendor follows the standard) 

• Report and query generation tools 

• Implementer-defined variations within the standard 

• Database initialization, opening and connection 


V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
14
 As we have already learned, a data type defines what kind of value a column can contain.
However, because there is variation across databases – we will use the data types for MySQL
for the purpose of this discussion. MySQL has three main data types: numeric, strings (text),
and dates/times. See the following figures:

Core MySQL Data Types – Numeric
Core My SQL Data Types - Numeric

Numeric Type Description
INT A standard integer
BIGINT A large integer

DECIMAL A fixed-point number

FLOAT A single-precision, floating-point number

DOUBLE A double-precision, floating-point number

BIT A bit field

Core MySQL Data Types – Strings (Text)
Core MySQL Data Types – Strings (Text)

String Type Description

CHAR A fixed-length, non-binary string (character) 7

VARCHAR A variable-length, non-binary string

NCHAR Same as above + Unicode Support
NVARCHAR Same as above + Unicode Support
BINARY A fixed-length, binary string
VARBINARY A variable-length, binary string
TINYBLOB A very small BLOB (binary large object)
BLOB A small BLOB
TEXT A small, non-binary string


8

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
15
 Core MySQL Data Types – Dates/Times
Core MySQL Data Types – Dates/Times
Date / Time Type Description
DATE A date value in 'CCYY-MM-DD' format
TIME A time value in 'hh:mm:ss' format
DATETIME Date/Time in 'CCYY-MM-DD hh:mm:ss’ format
TIMESTAMP Timestamp in 'CCYY-MM-DD hh:mm:ss’ format
YEAR A year value in CCYY or YY format

Creating Databases and Tables
To get started, we will need to know how to create databases and tables. While MySQL can be
an intimidating program, but once you master some of the basics, you will be able to work
effectively with large data sets.
• To create a Database, we use the CREATE DATABASE command 

9
• Once you have created the database, you will now apply the USE command to tell the
system which database to use 


Once you have created a database, you will want to create tables within the larger database:
• New tables are declared using the CREATE TABLE command
• We can also set the name and data type of each attribute 

• When creating new tables, we can specify primary keys and foreign key relationships
• We can also decide whether or not NULL or empty values are allowed 


Inserting Data into a new Database

Once you have created a new database, you are ready to insert data. The data model will act a
guide to load data into a new database. If the database builds well, it may mean that you have
found the real business rules. Or, if you have some errors, you may have the real business rules,
but the data may be messy. Finally, if it builds with many errors – this may be the case that the
business rules are not accurate and are closer to what they want or think they have, not what
they use. In many cases, it is useful to get sample data and browse it during the process of
building the model.

SQL Select Queries

SQL SELECT query is used to fetch the data from a database table, which returns data in the
form of a result table. SELECT returns a set of attributes in a query. In most applications, SELECT
is the most commonly used data manipulation language command. SELECT statements are
constructed from a series clauses to get records from one or more tables or views.

Clauses must be in order; only SELECT and FROM are required:
• SELECT attributes/columns
V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management
MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
16
 • INTO new table
• FROM table or view
• WHERE specific records or a join is created
• GROUP BY grouping conditions (attributes)
• HAVING group-property (specific records)
• ORDER BY ordering criterion ASC | DESC
• DISTINCT return distinct values

Wildcards in SQL
Most database implementations offer additional regular expressions – wildcards. A wildcard
character can be used to substitute for any other character(s) in a string. Regular expressions
can be used to find records, which match complex string patterns. For instance, MySQL has:
• [list] match any single character in list, e.g. [a-f]
• [^list] match any single character not in list, e.g. [^h-m]

Editing a Table
In some cases you will be faced with the need to edit a table. In this case you will use the
following:
• INSERT is used to add a new record to a table that contains specific values for a set of
attributes in that table
• The UPDATE keyword is used to modify a specific value or set of values for a set of
records in a table
• DELETE is used to remove records from a table that meet a specific condition

Learning Objectives
• Become more familiar with SQL.
• Recognize different implementations of SQL have differences of which to be aware.
• Review the different data types.
• Learn how to create new databases and tables.
• Understand how to use a SELECT query.
• Be familiar with wildcards and when to use them.
• Review how to edit a table.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
17
 Database Conditional, Grouping, and Joins
Summary
In the next section, we examine how to deal with database conditional, grouping, and joins. As
we get further into SQL, we will need to refine our approach to make our actions more
effective. For example, we will need to narrow the set of records that get returned from a
query. We will also need to make statistical queries across different groupings or records. In
addition, we will need to sample or order our output results. Another challenge will include
integrating data from other sources within our database. The following review will cover these
challenges and others as we continue to work with SQL.

Database Conditional Clauses

A conditional clause is a part of a query that restricts rows matched by certain conditions. You
can narrow SELECT statements with conditional clauses such as WHERE IN, or the BETWEEN
keyword. WHERE IN statements are used to identify records in a table with an attribute
matching a value from a specified set of values. The BETWEEN keywords are used to identify
records that have values for a particular attribute that fall within a specified range

WHERE IN: WHERE attribute IN is used to select rows that are satisfied by a set of WHERE
conditions on the same attribute. Example:

SELECT *
FROM Offices
WHERE State IN ('CO', 'UT', 'TX');

is equivalent to:

SELECT *
FROM Offices
WHERE State = 'CO' OR State = 'UT'
OR State = 'TX'


BETWEEN Keyword: Select records where the attribute value is between two numbers using
BETWEEN, range is inclusive and also works with time and date data.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
18
 Null Values
Null values are treated differently from other values; they are used as a placeholder for
unknown or inapplicable values. If values are empty or missing, they are stored as NULL. A few
issues to be aware of for NULL values:
• NULL values evaluate to NOT TRUE in all cases
• Check for NULLS using IS and IS NOT
• When a specific attribute may contain NULL or missing values, special care must be
taken when using conditional clauses

Grouping Data and Statistical Functions

Once you are a bit more comfortable working with SQL, you can start to explore some of the
statistical functions that are included in many implementations of SQL. These functions can
operate on a group of records. Using the GROUP BY clause will return a single value for each
group of records. To further restrict the output of the GROUP BY clause to results with certain
conditions, use the HAVING key words (analogous to the WHERE clause).

Aggregate Statistical Functions in SQL
Commonly used functions include:

More advanced statistical functions can be created using the basic statistical functions built into
SQL such as calculating the weighted average or getting the z-score values by combining
different functions.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
19
 Sorting and Sampling Data
You will also be faced with the need to sort and sample the data. Several clauses will help you
will that including ORDER BY, LIMIT, and RAND.

ORDER BY: The ORDER BY clause specifies that the results from a query should be returned in
ascending or descending order

LIMIT the number of returned records: A LIMIT clause restricts the number of records that
would be returned to a subset, which can be convenient for inspection or efficiency

Randomly select and order records: The RAND() function can be used to generate random
values in the output or to randomly sample or randomly order the results of a query. For
instance:

Reorder the entire table:
SELECT *
FROM table
ORDER BY RAND();

Randomly select a single record:
SELECT *
FROM table
ORDER BY RAND()
LIMIT 1;

Generate a random number in the output results:
SELECT id, RAND()
FROM table;

Creating New Tables and Aliases

AS Keyword (Aliases): The AS keyword creates an alias for an attribute or result of a function
that is returned in a query

CREATE TABLE AS: Use CREATE TABLE with AS to create a new table in the database using a
select query. It matches columns and data types based on the results in the select statement.

Results from a query can be inserted into a new table using the CREATE TABLE with the AS
keyword. As seen in the following:

CREATE TABLE new_table
AS ( SELECT column_name(s)
FROM old_table);

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
20
 SELECT INTO: Results from a query can be inserted into an existing table using a SELECT INTO
clause if the table with the appropriate structure already exists.Take the results of a select
statement and put them in an existing table or database:

SELECT column_name(s)
INTO newtable [IN externaldb]
FROM table1;

Joining Multiple Tables

The relational database model allows us join multiple tables to build new and unanticipated
relationships. The columns in a join must be of matching types and also must represent the
same concept in two different tables. This can help us to contextualize or integrate a table in
our database with data from an external source.

We want to learn how to take data from different tables and combine it together. This may
include data from other data sources that complement our own, such as demographic
information for a zip code or price structure for shipping zones for a carrier. The process of
merging two separate tables is called “joining”. Joins may be done on any columns in two
tables, as long as the merge operation makes logical sense. See below for a visual
Joins - visually
representation of joining:

Two tables with shared

field/data

SELECT *
FROM tb1, tb2;

SELECT *
FROM tb1, tb2
WHERE tb1.bg = tb2.bg;

SELECT grey, pink

FROM tb1, tb2
WHERE tb1.bg = tb2.bg;
36

Columns in a JOIN
• They don’t need to be keys, though they usually are
• Join columns must have compatible data types
• Join column is usually key column: Either primary or foreign
• NULLs will never join
V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management
MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
21


Types of Joins and Views

Join from 3 Tables: Joining three tables together just involves one additional join between two
already joined tables and a third table.

Join Types
Different types of joins can be used to merge two tables together that always include every row
in the left table, right table, or in both tables. The following are different types of JOIN:
• INNER JOIN: returns only the records with matching keys (joins common column values)
• LEFT JOIN: returns all rows from LEFT (first) table, whether or not they match a record in
the second table
• RIGHT JOIN: returns all rows from RIGHT (second) table, whether or not they match a
record in the first table
• OUTER JOIN: Returns all rows from both tables, whether or not they match (Microsoft
SQL, not MySQL)
• In MySQL, JOIN and INNER JOIN are equivalent

Views
Views are virtual tables that do not change the underlying data but can be helpful to generate
reports and simplify complicated queries. They are virtual tables that present data in a
denormalized form to users. They do no create separate copies of the data (they reference the
data in the underlying tables). The database stores a definition of a view and the data is
updated each time the VIEW is invoked.

There are several advantages to VIEWS. User queries are simpler on views constructed for
them. They offers a layer of security that can restrict access to data in views for users. They also
provide greater independence, meaning that the user or program is not affected by small
changes in underlying tables.

Learning Objectives
• Learn how to work with SELECT for conditional clauses.
• Recognize the role and use of NULL values.
• Review how to group data with the GROUP BY clause.
• Introduce the existence of statistical functions in all SQL functions.
• Recognize how to apply aggregate statistical functions.
• Review sorting and sampling techniques such as ORDER BY, LIMIT and RAND.
• Learn how to create new tables and aliases using the AS keyword.
• Becoming familiar with joining multiple tables.
• Recognize the types of JOINs.
• Identify when and how to use VIEWs.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
22
 Topics in Databases
Summary

In this section, we will be covering some additional database topics. These topics are going to
help SC professional apply databases in the real world. To learn on how to optimize
performance in relationship database, this section will introduce you to indexing and its impact
on performance. Then, you will learn another type of database that is recommended for cases
where analytics and reporting needs to be performed to help make data driven decisions. We
will then move away from relationship database and explore NoSQL database. NoSQL database
can offer flexibility and superiors performance for some type of applications especially in cases
where data is unstructured. The next topic that we'll look at is what are the benefits of storing
data and processing data on the cloud. Finally, we'll think about how to preprocess and clean
our data sets before we put them into databases. This skill will be really essential for us to make
sure that we actually store and retrieve reasonable data every time we use our database.

Indexes and Performance
The best way to improve the performance of SELECT operations is to create indexes on one or
more of the columns that are tested in the query. The index entries act like pointers to the
table rows, allowing the query to quickly determine which rows match a condition in the
WHERE clause, and retrieve the other column values for those rows. All MySQL data types can
be indexed. Primary and Foreign Keys are automatically indexed by SQL to allow them to be
searched faster.

Although it can be tempting to create an indexes for every possible column used in a query,
unnecessary indexes waste space and waste time for MySQL to determine which indexes to
use. Indexes also add to the cost of inserts, updates, and deletes because each index must be
updated. You must find the right balance to achieve fast queries using the optimal set of
indexes.

Example of Indexing
Consider a Customer table that contains the following columns: Customer ID, Customer name,
City, state, Postal (zip) code and Address. Customer_ID, here, is the primary key. Let us assume
that we want to index State and City so that one can quickly narrow down the lists of customer
in a particular state and city. This can done by using the below syntax where IX_City_State is
just the name of the index that will be created on the table “Customer” and the attributes are
“State” and “City”

CREATE INDEX IX_City_State
ON Customer (State, City)

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management
MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
23
 Databases and Data Warehouses
So far, we have looked at one type of databases that are known as online transaction
processing, or OLTP. There is another class of database that you should be familiar with. These
are known as online analytical processing or OLAP. OLTP and OLAP, each have different use
case and purpose. The below tables list down the common use case and key differences

Use cases
between these databases:

Table: Use Cases

OLTP OLAP
Manage real-time business operations Perform analytics and reporting
Supports implementation of business Supports data driven decision making
tasks
e.g. Transactions in an online store e.g. Data mining and machine learning
e.g. Dashboard showing health of e.g. Forecasting based on historic data
business over the last few days/hours
Concurrency is paramount Concurrency may not be important

Key differences

Table: Key Differences

OLTP OLAP
Optimized for writes (updates and Optimized for reads
inserts)
Normalized using normal forms, few Denormalized, many duplicates
11
duplicates
Few indexes, many tables Many indexes, fewer tables
Optimized for simple queries Used for complex queries
Uncommon to store metrics which can Common to store derived metrics
be derived from existing data
Overall performance: Designed to be Overall performance: Designed to be
very fast for small writes and simple very fast for reads and large writes,
reads however relatively slower because data
tends to be much larger


12

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
24
 NoSQL Database
A NoSQL database provides a mechanism for storage and retrieval of data that is modeled in
means other than the tabular relations used in relational databases. Unlike SQL, the data in
NoSQL database is unstructured and is not stored in tables. NoSQL databases uses other
mechanisms to store data, some of which are discussed in the next section.

NoSQL database offers the following benefits over SQL database:
• NoSQL may be faster in some use cases, design data model around application. Simple
reads and writes are very fast with NoSQL solutions
• NoSQL databases is flexible as compared to SQL database and allows storage of data
when the relationship between many of entities is not clear.
• NoSQL databases are easy to scale because of their simplified structure.

Common types of NoSQL databases
Transactional business data is still commonly stored in relational databases due to consistency
and how reads and writes are handled
• Key-value database – "look-up table" or "dictionary"
Types of NoSQL database
n Simplest examples can be made more complex with
Key-value database • The least complex NoSQL option, which stores data in a schema-less
formats like JSON
way that consists of indexed keys and values.
n Each record may have different data fields or attributes,
• Each record may have different data fields or attributes, which are
which are stored together with a unique key
stored together with a unique key
n Data model is not predefined, empty fields are not stored
• Data model is not predefined, empty fields are not stored
Key Value
Key Value
1000 {name: "Chris", language:
1000 Chris "English", city: "Boston", state:
1001 Julie "MA"}

Common types of NoSQL databases

1002 Clark
1003 MA
1001 {name: "Julie", state: "NY"}
1002 {name: "Clark", language:
"French", language: "Spanish"}
• Document-oriented database
17
Document-Oriented • Similar to key-value concept but adds more complexity.
n Similar to key-value database
database • Each document (grouping of key value pairs, also known as
n Key value pairs can be further grouped into collections,
collection) in this type of database has its own data, and its own
typically related to entities
unique key, which is used to retrieve it.
n Note duplicated data Offices

Employees Key Value

Key Value 10000 {name: "Boston Office",

city: "Boston", state:
1000 {name: "Chris", language: "MA", employee:
"English", city: "Boston", state: "Chris", employee:
"MA"} "Clark"}
1001 {name: "Julie", state: "NY"} 10001 {name: "New York
1002 {name: "Clark", language: Office", state: "NY",
"French", language: "Spanish"} employee: "Julie"}
18

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management
MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
25
 Common types of NoSQL databases
• Column-oriented database
Column-oriented • Stores data tables as columns rather than rows. Each row is indexed
n Imagine "indexing" every column in a relational database
database and thus has “RowID”
• with the row ID
"RowID" acts as the "key" and the attribute acts as the "value"
• n "RowID" acts as the "key" and the attribute acts as the
Now information about each column is stored more efficiently for
"value"
some queries, allowing for excellent scalability and high
performance.
n Now information about each column is stored more

efficiently for some queries First Name RowID

Christopher 1, 3
Employees
Julie 2
Common types of NoSQL databases
1
RowID First Name
Christopher Massachusetts State
State
RowID
2 Julie New York Massachusetts 1
• Graph database
3 Christopher New York New York 2, 3

Graph database Connections exist, and are not created during query like a
• n Stores data that’s interconnected and best represented as a graph 19

relational database, efficient for highly connected systems

• This method is capable of lots of complexity

https://neo4j.com/developer/graph-database/ 20

Cloud Computing

Cloud computing enables users and enterprises to access shared pools of resources (such as
computer networks, servers, storage, applications and services) over the internet.

Cloud computing offers various benefits over traditional systems, including:
• Low start-up cost
• Low risk development and testing
• Managed hardware (and software)
• Global reach
• Highly available and highly durable
• Scale on demand, pay for what you use
• In some cases, can work with local infrastructure if needed

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management
MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
26
 Types of Cloud Computing

Infrastructure as a service • Outsource hardware
(IaaS) • User provides operating system, database, app, etc.
• Most flexible, higher upfront total IT cost
• Example: Rackspace
Platform as a service (PaaS) • Outsource operating environment
• Cloud platform provides OS, database, etc.
• Examples: Amazon, Microsoft, Google
Software as a service (SaaS) • Outsource software
• Configure the third party app
• Examples: Salesforce, HubSpot
• Least flexible, lower upfront total IT cost

Data Cleaning
Data must be cleaned or pre-processed before it can be inserted into a database. This is a time
consuming but a mandatory exercise, as the quality of the database depends on the integrity of
the data.

Types of data cleaning solutions

Data cleaning can be performed with free software and tools, however the learning curve for
these can be steeper than that of commercial software. These options are listed below:

Off the shelf software • Graphical user interfaces, no programming required, enables
collaboration with non-programmers
• Offer various advance features including capability to operate on
large data set, reproducible workflow and version control
• Software is not free
• Examples: Trifacta, Paxata, Alteryx, SAS

Open-source • Working in data frames and data dictionaries, requires some
programming programming skills, but languages are relatively google- friendly
languages • Offers the same benefits as “Off the shelf software” but requires
programming experience and can be sometime time consuming.
• Software is free
Unix command line • Not as user friendly as previous options, but very fast and versatile
tools • Excellent for breaking up large datasets that would crash other
software
• Software is free and often built in the OS.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
27

Learning Objectives
• Understand Indexing and its impact on performance.
• Introduce online transaction Processing(OLTP) and online analytical processing (OLAP)
• Review use cases and key differences of OLTP and OLAP
• Introduce NoSQL database
• Review various types of NoSQL databases.
• Introduce Cloud computing and review types of services offered through cloud.
• Become familiar with advantages of cloud computing.
• Recognize importance of data cleaning

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
28
 Introduction to Machine Learning
Summary
In this lesson, we explore machine learning. This includes identifying when we need machine
learning instead of other techniques such as regression. We break down the different classes of
machine learning algorithms. In addition, we identify how to use machine-learning approaches
to make inferences about new data.

Review of Regression
Linear regression uses the value of one or more variables to make a prediction about the value
of an outcome variable. Input variables are called independent variables and the output
variable is known as the dependent variable.
• Linear regression output includes coefficients for each independent variable.
o This is a measure of how much an independent variable contributes to the
prediction of the dependent variable.
o The output also includes metrics to be able to assess how the model fits the
data. The better fit of the model, the better you are able to make accurate
predictions about new data.
• Using coefficients calculated from historic data, a regression model can be used to make
predictions about the value of the outcome variable for new records.

Overview of Machine Learning Algorithms

Machine learning algorithms are primarily use to make predictions or learn about new,
unlabeled data. There are several classes of algorithms:
• Classification: assigning records to pre-defined discrete groups
• Clustering: splitting records into discrete groups based on similarity; groups are not
known a priori
• Regression: predicting value of a continuous or discrete variable
• Associate learning: observing which values appear together frequently

Supervised vs. Unsupervised Machine Learning

Supervised learning uses outcome variables, known as labels, for each record to identify
patterns in the input variables or features related to the outcome variable.
• Correct answer, label is known in the training data
• Label is applied by a person or already exists
• Labeled data are used to train an algorithm using feedback
• Apply or test the trained model on new, unseen data to predict the label

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
29
 Supervised learning workflow
Supervised learning workflow
Learning Flow

Training Making PredicBons
Training Making PredicBons
Unlabeled
Raw Data Labels
Data
Unlabeled
Raw Data Labels Data

Algorithm
(Training)
Model

Algorithm
Model
(Training)

Model PredicBon


Model PredicBon

EvaluaBon Unsupervised learning workflow
Learning Flow
EvaluaBon 16
In unsupervised learning, the outcome variable values are unknown, therefore Raw Data
relationships among the input variables are used to identify patterns of clusters of
records.
• Finds previously unknown patterns in the data without labels or guidance 16
Algorithm
• No training/testing/validating process because correct answer is unknown

Model

Model Quality
Machine learning models should be trained on an unbiased set of data that is Produc<on
representative of the variance in the overall dataset. Bias quantifies the lack of
ability of a model to capture underlying trend in the data. More complex models
decrease bias but tend to increase variance. Variance quantifies a model’s sensitivity to small
changes in the underlying dataset.
• Ideally want low bias and low variance, but there is a tradeoff between the two
quantities
• If there is a bias in the training data or if too many features are included in a model, the
model is at risk of being overfit. In overfit models, the coefficients, known as
parameters, will not be generalizable enough to make good predictions for new records.
• A large and representative sample of the labeled data should be used to train the
model, the remainder is used for testing.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
30
 Overfitting vs underfitting
• Underfitting – model is too simple, high bias and low variance
• Overfitting – model is too complex, low bias and high variance
o Overfitting is a more common pitfall
o Tempting to make models fit better by adding more features
o Results in a model that is incapable of generalizing beyond the training data

Learning Objectives
• Be introduced to machine learning
• Become familiar with different types of machine learning algorithms
• Be able to differentiate supervised and unsupervised learning and their processes
• Recognize model quality and the tradeoffs between bias and variance
• Learn how to identify when a model is over or underfit

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
31
 Machine Learning Algorithms
Summary
In this lesson, we are going to dive deeper into machine learning algorithms. Each model has
different properties and is best for different types of tasks. We review how to compare them
with performance metrics. We need to be able to group records together without labels to
inform prediction using unsupervised classification. In addition, we review capability to
confidently reduce the number of features included in an analysis without losing information.
The lesson also introduces how to compare predictor accuracy and test for sensitivity and
specificity.

Dimensionality reduction
Dimensionality reduction is a term for reducing features included in analysis. It is often needed
for analysis with many features. Trying to reduce dimensionality randomly or manually leads to
poor results
• Results need to be interpreted by humans, should be tractable
• Increasing the number of features included increases the required sample size
• Features should not be included or discarded from analysis based on instinct
o Dimensionality reduction techniques should be employed, such as principal
component analysis.
• Summary statistics are a means of dimensionality reduction

Principal component analysis (PCA)

PCA is a mathematical approach to reduce dimensionality for analysis or visualization. It
exploits correlations to transform the data such that the first few dimension or features contain
a majority of the information of variance in the dataset. PCA determines which variables are
most informative based on the distribution of data and calculates the most informative
combinations of existing variables within the dataset. PCA works well for datasets with high
dimensionality.
• No information is lost, first few components hold much of the information
• Same premise as linear regression except without a dependent variable
o Linear regression solution is the first principal component
o Disregarding the information describing the principal component, PCA calculates
the second most informative component, then the third, and so on
• Linear combinations form a set of variables that can be used to view the data – new
axes
• Components are ranked by importance, so all but the first few can be discarded, leaving
only the most important information with very few components
• The coefficients in the table give the proportion of each of the original variables that
went into each component

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
32
 • Relative signs +/- indicate that two variables are positively negatively correlated in that
particular component
• The components are difficult to interpret using only the coefficient values, plotting often
improves understanding
PC1 = (a*var1 + b*var2 + c*var3 + …)
PC2 = (d*var1 + e*var2 + f*var3 + …)
PC3 = (g*var1 + h*var2 + i*var3 + …)

Clustering
Another way of thinking about dimensionality reduction is how close each point is to other
points. The idea is to separate data points into a number of clusters that have less distance
between the points internally than to other clusters. Clustering can be helpful to identify groups
of records that have similar characteristics to one another. When data is unlabeled, clustering
can be used to group records together for deeper inspection. Upon deeper inspection of the
records in each cluster, users can understand the patterns that lead to records being grouped
together, and also identify reasons for records being grouped separately.

K-means clustering
k-means clustering starts with selecting the number of clusters, k. k cluster-centers are placed
randomly in the data space and then the following stages are performed repeatedly until
convergence. K-means does not determine the appropriate number of clusters, this is set by the
user based on intuition or previous knowledge of the data The algorithm can terminate with
multiple solutions depending on initial random positions of cluster-centers and some solutions
are better than others.
• Data points are classified by the center to which they are nearest
• The centroid of each cluster is calculated
• Centers are updated to the centroid location

Classifications
• Clustering and PCA allow users to see patterns in the data, which is the best that can be
done because there are no labels to guide the analysis
• With supervised learning, the label is included in the learning process:
o Unsupervised: what features are most important or interesting?
o Supervised: what features are most informative about the differences between
these groups?
• Classification methods: each record falls into some category or class, predict the
category of a new record based on values of other features in the record
• Regression methods: one variable depends on some or all of others, predict the value of
the dependent variable based on the values of the independent variables

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
33
 Classification Trees
Classification trees split data to find optimal values for features, used to split data by class. Tree
diagrams show the class makeup of each node, and the relative number of data points that
reach each node
• Tree pruning
o Tree pruning removes rules associated with overfitting from the tree
o The new tree misses a few points classified correctly, but contains only meaningful
rules, more generalizable to new data

Naïve Bayes classifier

• The Naïve Bayes algorithm considers the value of each feature independently, for each
record, and computes the probability that a record falls into each category
• Next, the probabilities associated with each feature are combined for each class
according to Bayes' rule to determine the most likely category for each new record
• Almost completely immune to overfitting - Individual points have minimal influence;
Very few assumptions are made about the data

Random forest
Random forest is an ensemble classifier that uses multiple different classification trees. Trees
are generated using random samples of records in the original training set. Accuracy and
information about variable importance is provided with the result.
• No pruning necessary
• Trees can be grown until each node contains very few observations
• Better prediction than classification
• No parameter tuning necessary

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
34
 Comparing predictor accuracy
Cross - validation
• Models should be good at making classifications of unlabeled data, not describing data
that is already classified.
• Randomly divide data into a training set and a test set
o Hide test set while building the tree
o Hide training set while calculating accuracy
o Computed accuracy represents accuracy on unseen data
• Techniques are available to do this multiple times, ensuring each record is in the test set
exactly once, e.g. k-folds

Comparing models
• Several standard measures of performance exist, can run multiple models and compare
metrics:
o Accuracy
o Precision
o Recall
o And more
• Application drives which performance metrics are most important for a given task

Sensitivity and specificity

Sensitivity and specificity are statistical measures of the performance of a classification test.
Sensitivity measures the proportion of positives results that are categorized correctly.
Specificity measures the proportion of negatives results that are categorized correctly. So for
example, if you needed to measure, say, the ability of a test to always detect a problem, you
would look at sensitivity. Or if you need to look at the performance of those classifiers at not
generating false positive classifications, then specificity. See example below:

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
35

The ROC Curve
• The Receiver Operating Characteristic (ROC) curve plots the true positive rate
(Sensitivity) versus the false positive rate (100 - Specificity) for different cut-off points
• Each point on the curve represents a pair of sensitivity/specificity values corresponding
to a particular decision threshold
• A test with perfect discrimination (no overlap in the two distributions) has an ROC curve
that passes through the upper left corner (100% sensitivity, 100% specificity)
• The closer the ROC curve is to the upper left corner, the higher the overall accuracy of
the test

Confusion Matrix
A confusion matrix is a table that is often used to describe the performance of a machine
learning algorithms on a set of test data for which the true values are known. Below is an
example confusion matrix:

Predicted High Predicted Low
Actual High 100 20
Actual low 10 200

The table tells us the below about the classifier:
• There are two possible predicted classes: "high" and "low". If we were predicting the
average spending by a user them a high will imply these customers are high spender, and
“low” will mean they are low spender.
• The classifier made a total of 230 predictions
• Out of those 230 cases, the classifier predicted "high" 110 times, and "low" 220 times.
• In reality, 100 patients in the sample are high spender, and 200 are low spenders.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
36
 Learning Objectives
• Be further introduced to machine learning algorithms and how to work with them
• Become familiar with dimensionality reduction and when and how to use it
• Recognize when to use clustering as an approach to dimensionality reduction
• Review different classification methods such as classification tress and random forest
• Learn how to compare predictor accuracy
• Become familiar with sensitivity and specificity as indicators of a binary classification
test

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
37
 Supply Chain Systems - ERP
Summary
In this next segment, we explore supply chain IT systems. Because supply chains are essentially
made up of three flows: information, material, and money – IT systems support the information
flow. For example, in a supermarket, they have to deal with different types of supply chain data
such as supplier inventory, facility management and payroll, sales, and expired and obsolete
inventory. There are many daily transactions in a supermarket that need to be captured and
ensured for consistency and timeliness. That information needs to be somehow translated into
usable information for business decisions, and then these objectives need to be efficiently
achieved. The amount of information for transactions per week in a single supermarket can
number in the millions. This is for a single store.

On an enterprise level, companies need systems that help them manage and organize this
information for use. While supply chains are always portrayed as neat and linear systems, the
reality is much different, as we have learned over the previous courses. Flows move up and
down the chain and through many partners until they reach their final destination. Supply
chains need IT systems because while teams may sit in different functional units they frequently
need to share information. In addition, many different firms interact in the supply chain, they
need systems to carry that information between them, this helps de-silo the supply chain.
There needs to be coordination across functions, which is the essence of supply chain
management and can be facilitated with systems like Enterprise Resource Planning (ERP).

Supply Chain IT Systems

Supply chains need IT systems because they are large, complex and involve many players. They
often become intertwined and individual actors impact others. Decision-making is based on
common data and interaction with other functions in a firm. And supply chains need IT systems
because supply chains require communication for so many interactions B2B, B2C, M2M, etc.
(B2B = business to business, B2C = Business to Consumer, M2M = machine to machine)

Enterprise Resource Planning (ERP) systems serve as a general ledger and central database for
all firm activity. The next are Supply Chain Planning Systems. These systems are primarily for
production planning and scheduling, demand planning and product lifecycle management. The
last are for Supply Chain Execution; which are transportation and warehouse management
systems and manufacturing systems. The first we will tackle are Enterprise Resource Planning
systems.

Enterprise Resource Planning

In the following section we cover why firms use ERPs; the core functions of ERPs; data needed;
communications of the systems; and strategic benefits of an ERP. Most firms have an ERP
because many functions in a firm such as sales, inventory, production, finance, distribution, and
V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management
MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
38
 human resources have siloed databases. With a centralized ERP, these databases can more
easily be exchanged and shared.

Benefits of ERP allow enterprises to organize processes and data structure, integrate
information into unified repository, make data available for many users, eliminate redundant
systems and data, reduce non-value added tasks, standardize process designs, as well as be
more flexible. There are significant drawbacks to using ERP, these include: significant
implementation time and maintenance that come at a cost, data errors ripple through systems,
competitive advantage can be dampened, firm reliance on a single vendor, shortage of
personnel with technical knowledge of system, and high impact of down time of said system.

ERP Core Functions

Most ERP Systems share the same core functions. They tie together and automate enterprise-
wide basic business processes:

Customer Management is the face to consumers and serves the following functions:
• enables order entry, order promising, open order status
• allows marketing to set pricing schemes, promotions, and discounts
• provides real-time profitability analysis, and
• permits order configuration, customer delivery schedules, customer returns, tax
management, currency conversion, etc.
Manufacturing is the face to production and serves the following functions:
• includes MRP processing, manufacturing order release, WIP management, cost
reporting, shop floor control etc.,
• provides real time linkage of demand to supply management enabling real time
Available-to-Promise (ATP) & Capable-to-Promise (CTP), and
• serves as primary interface to “bolt-on” advanced planning and scheduling optimization
modules.
Procurement is the face to suppliers and serves the following functions:
• integrates procurement with supplier management,
• facilitates purchase order processing, delivery scheduling, open order tracking,
receiving, inspection, and supplier performance reporting, and
• creates requests for quotation (RFQ)
• manages negotiation and pricing capabilities.
Logistics is the face to internal and external supply chain and serves the following functions:
• runs the internal supply chain for enterprise,
• provides connectivity to remote trading partners (3PLs, carriers, etc.),
• handles distribution channel configuration, warehouse activity management, channel
replenishment, planning, distribution order management, etc., and
• serves as primary interface to “bolt-on” warehouse and transportation management
systems (WMS and TMS).
Product Data is the face to all material and serves the following functions:
• describes products enterprise makes and/or distributes,

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
39
 • contains proprietary data on costs, sources, engineering details, dimensions, weight,
packaging, etc.,
• interfaces with inventory, manufacturing, and product lifecycle management, and
• sometimes included in partner collaborations in order to compress time to market of
new products.
Finance is the face to the CFO and serves the following functions:
• strong suit of most ERPs (but also a double edged sword!),
• provides real-time reporting of all transactions resulting from inventory movement,
accounts receivable, accounts payable, taxes, foreign currency conversions, and any
other journal entries, and
• supports detailed reporting and budgeting capabilities.
Asset Management is the component that controls key assets.
• controls enterprise’s fixed assets
• established and maintains equipment profiles, diagnostics and preventive maintenance
activities, and depreciation activities.
Human Resources is the face to employees.
• manages all aspects of human capital with enterprise
• monitors performance of transaction activities to include time, payroll, compensation,
expenses, recruitment, etc.
• supports employee profiles, skill development, career planning, performance
evaluations, and retention.

ERP Data
There are three types of ERP Data:
Organization data: represents and captures the structure of an enterprise.
Master data: represents entities (customers, vendors) with processes. It is the most commonly
used. But because specific processes use materials differently and specific data needs differ by
processes – this adds to complexity of master data needs. Material types can be in different
states and can be grouped differently based on firm needs.
Transaction data: reflects the outcome of executing process steps. It comes in organizational,
master and situational data. Transaction documents include purchase orders, invoices, etc.

ERP Communication
Business-to-Business (B2B): Commerce transactions between manufacturers, wholesalers,
retailers. Each business represents a link in the supply chain.

Business-to-Consumer (B2C): Sale transactions between firms and end-consumers. The volume
of B2B transactions is much greater than B2C.

Accelerating and validating B2B and B2C transactions: For B2B this is achieved through
Electronic Data Interchange (EDI). For B2C this is achieved through a website and email.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
40
 Electronic Data Interchange (EDI): “The computer-to-computer interchange of strictly
formatted messages that represent documents other than monetary instruments.” There is no
human intervention in the process.

ERP Communica-on



§  B2B EDI example:
ERP systems can “communicate” via EDI, sharing near real-time information. See figure below.

Purchase Order

Order Confirma-on
SAP
Data Order Cancella-on Oracle
Format Data
Business X ETA Business Y Format
Gateway Order Cancella-on Gateway
ASN
Order Receipt

§  ERP systems can “communicate” via EDI, sharing near real--me informa-on.
The Value of ERP for SCM
§  The data is usually translated and validated to be imported into an ERP system.
There are three important values of ERP for supply chain management: reduction of the
§  Any info file can be shared given appropriate ERP fields to capture and display its
bullwhip effect, enabling widespread analytics, and extending the enterprise.
content.
§  What other info would businesses want to share?
Reducing the Impact of Bullwhip Effect
One of the key values of an ERP system is reducing or preventing the Bullwhip Effect. The
MIT Center for
Transportation & Logistics Adapted from Omar Elwakil (2016)
24
Bullwhip Effect is phenomenon where information distortion leads to increasing order
fluctuations in the upstream supply chain (forecast-driven supply chains). It is driven by several
behavioral causes like overreaction to backlogs and lack of transparency. There are also many
operational errors such as forecasting errors, lead-time variability, and promotions.

ERP can reduce the impact of the Bullwhip Effect by extending visibility downstream to
customer demand and upstream to participants enabling collaboration and information
sharing. It also facilitates point of sale capturing and helps reduce batch size and demand
variability through smaller and more frequent orders.

Enabling Analytics
ERP systems play a key role in enabling analytics. They are primarily retrospective, serve as the
ledger of the firm, and provide the CFO with financial snapshots. They also enable other forms
of analytics for Business Intelligence (BI): which transforms raw data into meaningful
information to improve business decision-making. These can be descriptive, predictive, and
prescriptive.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
41
 Extending Enterprise
While ERP systems are primarily used in intra-firm process management to connect various
departments and provide access to data, they also serve an extending function for better
connection with partners. ERPs serve a value in connecting End to End Supply Chains with
better connections across SC participants, providing shared understanding, reducing
coordination and monitoring costs, and responding quickly to market feedback.

Learning Objectives
• Introduction to supply chain IT systems, their value, application, and constraints.
• Review ERP, its setup functionality, and applications.
• Recognize core functions of ERP.
• Be familiar with data house in ERP systems.
• Review how that data is used to facilitate communication.
• Understand some of the value of ERP systems for supply chains.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
42
 Supply Chain Systems - Supply Chain
Modules
Summary
In this next segment we review different supply chain modules as a subset of supply chain IT
systems. To understand where we are now with supply chain IT systems we need to review the
evolution of supply chain tools. We journey from the 1960-70’s with the Bill of Materials
Processor, mainframes based database systems, and material requirements planning (MRP) to
the 1980’s with the second coming of MRP that added finance and marketing, Just-In-Time
manufacturing methodology, expansion to other functions, and the precursor to the ERP
systems of today. In the 1990s, most MRP systems were absorbed into ERP suites; there was
the introduction of Advance Planning Systems (APS), and wider adoption of ERP systems. In the
2000s, many of these systems adopted web-based interfaces, improved communication, and
adopted shared or cloud based solutions. There was also a major consolidation of supply chain
software vendors and expansion of ERP systems to include SCM.

Now as we explore how to further extend the enterprise and its ability to adequately manage
its information on its own and together with other companies, many firms use a series of IT
modules. These systems are sometimes a part of ERP, may be standalone applications, or can
be part of a larger supply chain ecosystem. We will review two main functionalities including
Advance Planning Systems (APS) and Execution Applications. Advanced Planning Systems (APS)
are long range, optimization based decision support tools while execution applications include
Warehouse Management Systems (WMS), Transportation Management Systems (TMS), and
Manufacturing Execution Systems (MES).

Planning vs. Execution

Although planning modules seek to enable future planning and enabling efficient processes,
there is often a gap between the planning and execution tasks. The figure below illustrates this
gap. Planning vs. Execution

Consists of a continuum of tasks, but . . .

ROA
Impact
Execution
Supply Tasks
Chain
Strategy/
Network
Design Tactical
Transportation
Modeling Transportation
Procurement Shipment
Consolidation Fleet
& Carrier Routing/
Selection Scheduling

. . . there is a gap!
Planning Tasks

MIT Center for
Transportation & Logistics

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
43
 Questions, Approaches, and Technologies Change based on timeframe
Questions can be strategic such as: “What carriers should I partner with and how?” “How
should I flow my products?” or they can be tactical such as: “How can I quickly secure rates for
a new DC/plant/lane?” “What lanes are having performance problems?” or operational:
“Which carrier should I tender this load to?” “How can I collaboratively source this weeks’
loads?”

The timeframe also drives the approach. For instance in the strategic face – a company will be
establishing a plan and strategy, have event based enablement and complete non-routine
analysis. In an operational timeframe they will be executing the strategic plan, operate on
transaction based rules and processes, and have automated actions.

And technologies also align with timeframes. For instance, strategic timeline will allow for
analysis engines tools like optimization, simulation and data analysis and communication via
the web, file exchange and remote access. The tactical timeline allows for the same analysis
and communication technologies while the operational timeline allows for communication but
also workflow software such as compliance tracking, rules, and transaction processing.

Advanced Planning Systems

We now take a closer look at advanced planning systems that are primarily used as decision
support systems. They typically include functionality for network design, demand planning,
production planning, production scheduling, distribution planning, and transportation planning.
Advanced Planning Systems utilize large scale mixed integer linear programs (MILPs) and
sometimes simulation.

Planning Horizons
Advance Planning Systems help with planning horizons. The following provide a rough guideline
but each firm differs and it is unique to specific industries.:
• 3 months out – Master Production Schedule (MRP, DRP)
o <4 weeks out - Frozen MPS
o 5 to 8 weeks out – Slush MPS – some changes allowed (+- 10%)
o >8 weeks out – Water MPS – more changes are allowed (+- 30%)
• 3-18 months out – Aggregated Planning
• >18 months out – Long Range Planning – Network Design, etc.

Flow
Inputs (from ERP or other systems): Current costs, manufacturing and storage capacities,
consensus forecast, sales orders, production status, purchase orders, and inventory policy
recommendations, etc.

Decision Process: Large scale optimization (MILP) across multiple facilities and time horizons in
a single planning run; unconstrained, constrained, and optimal plans

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
44
 Outputs: Demand forecast and plan for meeting demand; a feasible production plan for future
periods to include allocation of production to plants; allocation of orders to suppliers;
identification of bottlenecks; Root Cause Analysis.

Transportation Management Systems (TMS)

TMS is software that facilitates procurement of transportation services, short-term planning
and optimization of transportation activities, assets, and resources, and execution of
transportation plans (Gonzalez 2009). It is often geographic and mode specific. The core
functions of TMS are: transportation procurement; mode and carrier selection; carrier
communication; routing guide generation and maintenance; fleet management; audit,
payment, and claims; appointment scheduling; yard management; and route planning.

Transportation Execution
The TMS serves as the interface to the carriers while connecting the Order Management
System (OMS), Payment Systems, and the ERP. Its main objective is to: move products from
initial origin to final destination cost effectively while meeting the level of service standards and
executing the plan using the procured carriers. This is broken down in phases below:

PLAN: Create Shipments from Orders
EXECUTE: Select and tender to Carriers
MONITOR: Visibility of the status of Shipments
RECONCILE: Audit invoices and pay for Transportation

There are many considerations to be made in TMS such as:
§ How do orders drop? Batched vs Continuous?
§ How much time is allowed between drop and must-ship? Weeks? Days? Hours?
Minutes?
§ What percentage of orders change after release?
§ How do they change? Quantity? Mix? Destinations? Timing?
§ What is the length of haul?
§ How many orders are “in play” at any time?
There are also key decisions like carrier selection and load building.

TMS Carrier Communication & Selection

Useful EDI Transaction Sets
• 204 – Motor Carrier Load Tender: Used by shippers to tender an offer for a shipment to
a full truckload motor carrier. It may be used for creating, updating or replacing, or
canceling a shipment.
• 990 - Response to a Load Tender: Used by motor carriers to indicate whether it will pick
up a shipment offered by the shipper
• 214 - Transportation Carrier Shipment Status Message: Used by carriers to provide
shippers and consignees with the status of their shipments.


V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management
MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
45
 Carrier Selection
Example of carrier selection
Carrier C
Continuous Move

Primary
Spot
Carrier A
Carrier A
Carrier E

Spot

Carrier B
Carrier B
Carrier B
Carrier B

Types of Capacity Dedicated

Carrier D
Primary - Contracted Carrier
Dedicated Fleet Spot
Continuous Move
Spot Carrier
MIT Center for
Transportation & Logistics 15

Linking Approaches
Approaches Must Be Linked
Tier III
Spot execution –
Tier II highly variable
Increased flexibility
in execution
Flexible Assmt Dynamic Pricing
Dynamic Carrier in Private
How do I select

Dynamic Selection Exchange

each carrier?

III IV
I II
Strategic Strategic
Lane Assmt Lane Assmt
Static w/ Tier Pricing

Tier I
Uses strategic Contract Dynamic
routing guide
How do I price each load?
MIT Center for
26

Transportation & Logistics

Warehouse Management Systems & Automation

WMS is a software system that facilitates all aspects of operations within a warehouse or
distribution center and integrates with other systems. It is not the same as inventory
management systems; WMS complements IMS. Examples of the benefits of a WMS include:
real-time stock visibility and traceability, improved labor productivity, and improved customer

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
46
 service. Some of these benefits are closely tied to automation of material handling and
paperless device interfaces.

Examples of Warehouse Automation include:
Automatic identification technologies: Bar codes and bar code scanners, radio frequency tags
(RFID) and antennae, smart cards and magnetic stripes, vision systems.
Automatic communication technologies: Radio frequency data communications, synthesized
voice, virtual displays, pick to light / voice systems
Automated material handling technologies: Carousels, conveyors/robotics, flow racks, AS/AR
Systems

WMS Software Components

Order Processing
• Order checking & batching
• Allocation
• Auto-replenishment
Receiving
• ASN planning
• In bound tracking
• Delivery appointment scheduling
• PO verification
• Returns processing
Put-Away
• Palletizing
• Zoning and slotting
• Random/directed put away
• Routing for putaway & replenishment
Picking
• Batch/Wave/Zone/Directed picking
• Carton/pallet select
• Assembly/kitting
• Pick-to-light/voice
Shipping
• Pallet sequencing & Load planning
• Pallet layering
• Trailer management
Labor Management
• Individual/team performance mgmt
• Labor scheduling
• Time standards
Equipment Support
• Interface to automated equipment
• Equipment maintenance
V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management
MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
47
 Manufacturing Execution Systems
MES is a software system that manages and monitors all work-in-process (WIP) in the
production process. It is integrated with the ERP to manage the execution of release of
production orders to finished goods delivery, trigger supply chain replenishments, and enhance
product traceability through manufacturing.

The functionality of a MES include:
• Machine scheduling
• Process management
• Document control
• Labor management
• Inventory management
• Product (WIP) tracking
• Performance analysis
• Labor management
• Quality management
• Production reporting

Learning Objectives
• Become familiar with systems that are common in supply chains that extend the
enterprise.
• Differentiate between Advanced Planning Systems and Execution Systems.
• Recognize the gaps in planning vs. execution and the timeframes embedded in both.
• Review Advanced Planning Systems, their use and application.
• Become familiar with the main execution systems in SC such as TMS, WMS, and MES.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
48
 Supply Chain Systems - Supply Chain
Visibility
Summary
In this next segment, we will introduce the concept of supply chain visibility. Supply chain
visibility provide us with a process of determining the current and past locations (and other
information) of an item within the supply chain. The two major components of a supply chain
visibility are track and trace. Tracking give us the capability to follow the route of an object as it
goes downstream throughout the supply chain from the start to finish. Tracing, on the other
hand, enable us to identify the source of an object or group of objects, within records,
upstream in the supply chain. Track and trace requires three different types of technologies:
Capture, Transmit and Access. ‘Capture’ transforms physical data such as location, time,
temperature, status, etc. into digital data. ‘Transmit’ enable movement of digital data from a
local source to global or cloud system and/or database. While ‘Access’ allow companies to
handle, manage and analyze data that gets generated throughout the supply chain.

Supply Chain Visibility

The ability to track and trace the path and status of all parts, components, and products from
original source to final use. Examples of questions that can be answered includes: Where is
order #12345 right now? How long on average does it take to get through the IB port? Where
are all of the items from lot #1299 located right now? Etc.

Supply Chain Visibility can be hard to achieve because of the following reasons:
1. Multiple parties involved in every shipment
2. Form of the product/conveyance can change
• Level of aggregation (conveyance, container, pallet, SKU, item, etc.)
• Consolidation/Deconsolidation (shipment, order, etc.)
3. Firms have different needs for visibility
• Real Time vs Near Real Time vs Post Hoc
• Tracking vs Tracing
4. Requires three different types of technology
• Data Capture & Translation
• Data Transmission & Upload
• Data Access & Actionability

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
49
 Tracking and Tracing
Tracking
The capability to follow the route of an object as it goes downstream throughout the supply
chain from the start to finish. Tracking enable supply chain professionals to answer questions
such as below:
• Where is it now?
• Where will it be?
• When will it get there?

Tracing
The capability to identify the source of an object or group of objects, within records, upstream
in the supply chain. Tracing enable supply chain professionals to answer questions such as
below:
• Where did it come from?
Supply Chain Visibility – Track vs. Trace
• Where has it been?
• Where are all the similar ones?

Trace now Track

Where did it come from? Where is it now?
Where has it been? Where will it be?
Where are all the similar ones? When will it get there?
• The capability to follow the route of an object
• The capability to identify the source of an
as it goes downstream throughout the supply
object or group of objects, within records,
chain from the start to finish.
upstream in the supply chain.
• Real time or near real time is required for
• Speed of recall is critical, but need not be
reacting or making tactical changes. But not
real time.
for post-hoc analysis.
• Requires record keeping of past activities,
• Requires current information for present to
locations, and events.
future looking analysis.

Tracking

Primary Processing Distributor Retail Consumer

producer company

Tracing

MIT Center for
Transportation & Logistics
4

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
50
 Technologies for Supply Chain Visibility

Broadly speaking, there are three types of technologies that are required to implement track
and trace within a supply chain These includes:
1. Capture & Translation
2. Transmission & Upload
3. Access & Actionability

Capture & Translation

These transform physical data such as location, time, temperature, status, etc. into digital data.
Examples includes Barcodes, Scanners, RFID, Smartphone, Internet of Things.

Let’s look at the top two commonly used technologies for capture and translation:

Barcode RFID
An optical, machine-readable, representation of data System that uses electromagnetic fields to
usually describing something about the object that automatically identify and track tags attached to
carries the barcode. objects. The tags contain electronically stored
information and can be passive (does not require
power) or active (requires power).
Benefits Benefits
• Enables accurate and efficient data collection all • Can be read without line of sight and from a
along the supply chain greater distance
• Easily track movements by SKU to identify fast/slow • Tags are highly durable and are reusable.
movers, monitor promotions, etc Multiple tags can be read simultaneously and so
• Enables the profiling of individual consumers are faster than barcodes.
• Relatively low cost and extremely accurate • Tags can have read/write capabilities - information
compared to key-entry. can be updated or added.
• Passive labeling – requires no power. Tags can potentially contain highly detailed data
Difficult to replicate or counterfeit and data can be
password protected.
Limitations • Limitations
• Require line of sight within ~15 feet (~4.6 meters) • More expensive than barcodes
in order to be read. • Does not work well with metal or liquids –
• Items must be scanned individually – labor interference for radio waves
intensive. • Requires assembling and inserting a computerized
• Barcodes can be easily counterfeited or replicated. chip.
• Typically, only the type of item is contained, not an • RFID still has two separate chips (read only and
individual unique code readable/writable), which cannot be read by the
• Read only and information cannot be updated or same machine or scanner.
added.

Transmission & Upload

These enable movement of digital data from a local source to global or cloud system and/or
database. Data can be transferred when the items reach a certain milestone, also known as
milestone reporting or can be transferred in real time using GPS. Examples are Electronic Data
Interchange, Application Programming Interface, Global Positioning Systems.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management
MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
51
 The below table summarizes the comparison between different transmission technologies:
Electronic Data Interchange (EDI) Application Programming Real time GPS
Interfaces (API)
Computer-to-computer exchange of Similar to EDIs but uses APIs to Transmits real-time location
business documents in a standard exchange information over the information from vehicles to a
electronic format between business internet. APIs are set of subroutine common platform.
partners (technology is over 30 definitions, protocols, and tools for
years old). building application software.
Benefits: Benefits: Benefits:
• Simplification of information flows • Provide “real” real-time data and • Able to track every asset in real
that require human interaction can interact with multiple time
and paper documents, systems via their universal • Enables the continuous monitoring
• Elimination of sorting, distributing, nature. of assets
organizing, and searching paper • Cloud-based technology, so it can • Many technologies are now
documents, be maintained and updated available – smartphones, ELDs,
• More efficient storage and ease of without disrupting shipping etc.
manipulation of electronic operations. • Able to implement contingency
records, • Lower implementation and plans in real-time (re-routing,
• Eradication of manual re-entry or maintenance costs (perhaps) diversion, etc.)
data – which reduces data errors, than EDI. • Able to ensure driver adherence to
• Greater trading speed and • Extremely flexible in terms of routes and other requirements.
reduced cycle times, which leads what and how to communicate. • Creates tremendous data that is
to • Enables two-way communication amenable to machine learning and
• Total cost reductions of 20% to – able to send responses to data other big- data analysis techniques
35% when replacing a paper based calls with success or error
system. messages
• Able to request additional data
on demand from API endpoints
• The standard for data
transmission in almost all
industries besides logistics
Limitation: Limitation: Limitation:
• Latency – EDI messages typically • APIs are not generally compatible • Can be expensive to operate
held and released in batches using with legacy systems • Requires power and active
timer- based processing • Not standard practice in the connectivity throughout
• Implementation – Establishing EDI Logistics industry and no current • Not available for all moves – ocean
direct connections between standards n Testing during moves are problematic
partners is cumbersome and time implementation can cause • Not fully integrated into TMS or
consuming (months) systems to crash. other systems
• Expensive - Can be very expensive, • Requires development resources • Inundation of data does not
especially when using a 3rd party to implement always lead to elucidation –
to translate the data • Potential security ramifications. almost forces adoption of some
• Not Interactive – EDI only data analytics tools
operates in one direction – there • Data received is context free
is no ability to respond with a (latitude, longitude, time)
“received” message • Data timing is difficult to merge
• Rigidity –EDI can sometimes with other milestone based data
require changing an established
business process. It’s not very
flexible: scheduled transmissions,
cannot request data on demand,
set milestones.
V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management
MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
52
 Data Access and Actionability Technology
These technologies enable companies to handle, manage and analyze data that gets generated
throughout the supply chain. Common technologies to make it happen includes:
• Integration into Existing Systems
• Supply Chain Event Management (SCEM) Systems
• Supply Chain Control Towers

The table summarizes each of these technologies:

Integration into Existing Systems Supply Chain Event Management Supply Chain Control Towers
(SCEM) Systems
• Extended Transportation • Exception based tools with “If- • Central hub or platform that
Management Systems (TMS) Then” rules captures data along supply
• Add-ons to Enterprise Resource • n Screens out the non- chain
Planning (ERP) tool necessary status updates and • Requires expert staffing and
• Challenges with merging real- only alerts if needed n allows for end-to-end visibility
time and milestone status data! • Follows the ”Sense-Alert- • Becoming common offering by
Respond-Learn” methodology 3PLs

Learning Objectives
• Become familiar with the concept of Supply chain visibility.
• Differentiate between Tracing and Tracking.
• Review the three technology requirements for supply chain visibility.
• Understand the benefits and limitation of various technologies used within supply chain
visibility.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
53
 Supply Chain Systems - Software Selection
& Implementation
Summary
As we have journeyed through various types of supply chain systems, we will now cover the
process of software selection and implementation. As firms are in the process of selecting their
supply chain systems they will need to be aware of various factors that will guide their decision
of whether to select the system or not. So we now discuss the process of software selection
and evaluation criteria. While selecting the appropriate software system can be challenging,
implementation is far more difficult. The process of implementation can be long and costly,
using additional resources. In this lesson, we cover general guidelines on what to be prepared
for when implementing software systems.

Architecture
Evolution of Architecture
To understand where supply chain systems stand now, it is helpful to understand the evolution
of the architecture starting in the 1970’s. The following are the various forms of architecture for
the last fifty years:
• Mainframe (1970s)
• Personal Computers (mid-1980s)
• Client-Server (late 80s to early 90s)
• Wide Web and Web 2.0 (mid-90s to present)
• Cloud or Post-PC (today and beyond)

Today there are a variety of software systems available to businesses. In terms of architectural
format, they can choose between “On Location” or “On Premise” – meaning that the firms host
the software in their own facilities or on their hardware and within their own firewall. However,
companies are increasingly opting for cloud computing options. This means that they have
several deployment models available to them.

Cloud Computing
As cloud computing becomes increasingly more popular, there are a variety of offerings that
can be tailored to firm needs. They are Infrastructure as a Service (IaaS), Platform as a Service
(Paas), and Software as a Service (Saas), third party management increases from IaaS to SaaS.
We discuss each format as well as benefits below:

Infrastructure as a Service (IaaS): In this format, the third party provides the firm with the
computing infrastructure, physical or virtual machines and other resources. Firm owns and
manages the software application. The benefits of this are:
V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management
MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
54
 • No need to invest in your own hardware
• Infrastructure scales on demand to support dynamic workloads
• Flexible, innovative services available on demand

Platform as a Service (PaaS): In this format, the third party provides firm computing platforms
to include operating system, database, web server etc. Firm owns and manages the software
application. The benefits of this are:
• Develop applications and get to market faster
• Deploy new web applications to the cloud in minutes
• Reduce complexity with middleware as a service

Software as a Service (SaaS): In this case, the third party provides firm with access to the
application software and handles installation, setup, maintenance, and running. Firm is
charged by use. Benefits include:
• You can sign up and rapidly start using innovative business apps
• Apps and data are accessible from any connected computer
• No data is lost if your computer breaks, as data is in the cloud
• The service is able to dynamically scale to usage needs

While there are many benefits to cloud computing, there are widespread disadvantages of
cloud computing that include but are not limited to: vendor outages, unrestricted government
access, security & privacy risks, and key data and processes require network access.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
55
 Software Selection Sources
There are different sources of software that firms need to be aware of. These include a
customized in-house system designed for a business, an ERP expanded system with additional

Pros and Cons for So3ware Sources

bells and whistles tailored for a company, best of breed solutions (of market solutions), and
best of breed platforms. These are discussed in further detail in the chart below:
Source Advantages Disadvantages
Customized •  Best fit to the firm and its processes. •  Excep7onally difficult and 7me
In-House consuming to develop
System •  Most expensive total cost of ownership
•  Difficult to maintain
•  Can result in “inward looking” solu7on
ERP •  Rela7vely fast implementa7on •  Tends to be inflexible in terms of
Expanded •  Less expensive than in-house process
Systems customiza7on •  Could require change in business
•  Efficient from IT perspec7ve processes
•  Easier to upgrade with ERP •  Not guaranteed to be best solu7on
enhancements approach
Best of Breed •  Best performing market solu7on for each •  Difficult to integrate different systems
Solu7ons func7on •  Can have slow performance
•  Requires the use of middleware
between the applica7ons
•  Upgrading individual components can
cause ripple effect problems
Best of Breed •  Very good, if not best, solu7on for each •  Requires the use of middleware
Plaborms func7on with easier integra7on between between the applica7ons
individual modules

MIT Center for
14
Outsourcing
Transportation & Logistics

There is also the option of outsourcing some of these systems to different providers. For
instance, in logistics, 3PLs or Third Party Logistics Providers, serve as an organization that can
run the software as well as perform all of the business processes. Having a 3PL run your logistics
eliminates the need for hardware and software. 3PLs can possibly replace personnel within the
firm. The use of 3PLs is most common with smaller firms.

The main reasons to outsource are to reduce capital expenditure for software and hardware. It
may also reduce costs as a result of partner’s economies of scale; they often have the ability to
do it faster and better as well as be more flexibile and agile. It may also be an opportunity to
increase levels of service at reasonable costs. Firm can focus on core business and bring in
expertise that is not affordable in-house. There are myriad other reasons to outsource, but
there are also many not to, discussed below.

At the top of the reasons not to outsource are security and privacy concerns; someone else has
access to the firm’s data. There are also worries of vendor dependency and lock-in. The firm
may lose in-house expertise to a core function. There are also high-migration costs as well as
concerns over availability, performance, and reliability. There are additional reasons not to
V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management
MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
56
 outsource. Firms need to weigh the pros and cons of outsourcing matched with business
objectives to decide which is suitable for them.

Software Vendor Selection

In the end, a firm must select its vendor. Some firms throw a dart on the wall, and that is its
choice. Others have an organized and formalized fashion to select a vendor and in general it
goes as follows:
1. Form a Project Team (Internal and/or External) & Objectives
2. Understand the Business and Needs: review current business processes, prioritize
needs/functionality, create Request for Information (RFI)
3. Create Initial Short List of Potential Solutions & Vendors
4. In-depth Review of Short Listed Vendors: have vendors conduct realistic product
demonstrations, references from current users.
5. Create and Distribute final Request for Proposal (RFP)
6. Make the Decision: negotiate contract, price, and service level agreements (SLAs) and
establish an implementation plan plan

While cost is one of the primary factors in decision-making, there are many other criteria that
need to be evaluated on top of cost. They are:
• Functionality – does the system features fit the firm’s processes and needs?
• Ease of Use – how fast is the initial learning curve and on-going use?
• Performance – what are the processing speeds?
• Scalability – how well can the system expand and grow with the firm?
• Interoperability – how well does the system integrate with other systems?
• Extendibility – how easily can the system be extended or customized?
• Stability – how reliable is the system in terms of bugs and up-time?
• Security – how well does the system restrict access, control confidential data, and
prevent cyber hacking?
• Support – how is the quality of the vendor in terms of implementation, support,
training, thought leadership etc. ?
• Vendor Viability – how is the vendor’s financial strength and willingness to supply
updates and enhancements? Will they be here in 3 years??

Because there are a variety of criteria that firms will be evaluating vendors on, a scorecard is a
popular way to capture financial and non-financial attributes. The criteria can be scored as rank,
ratings, and grades. Scorecards tend to be very detailed and can even be broken down by
specific features. The selection can be made between vendors or between alternative hosting
platforms.

Total Cost of Ownership

Software License: Direct cost of the software system itself – assuming ownership.
Maintenance: Ongoing annual costs to guarantee upgrades and bug fixes.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
57
 Platform/Hardware: Cost of needed hardware to run the new software.
Training: Cost of training initial and on going personnel
Implementation: Cost of getting the system to go live! These vary widely between systems and
firms.
Customization: Cost of modifying the system itself to fit the firm’s processes. Nothing in SCM is
used straight out of the box (vanilla).
System Integration
Cost of interfacing this system with other modules and modifying existing systems to fit

Implementation
While selecting a vendor can be difficult and time consuming, the actual process of
implementation can take an even more significant amount of time and consume a lot of
resources. There are a few different approaches to implementation. They include Direct (or Big
Implementa7on Approaches
Bang), Parallel, Pilot, and Phased (or Rolling). Each of these has its own positives and negatives,
but the approach must suit the needs of the business.
Moving from an old mul7-module system to a new mul7-module system.
# Modules # Loca7ons Comments
Converted Converted
Direct All All •  Switch from the old to new system occurs on one day
or •  Pain of switch concentrated for en7re firm
•  Fastest implementa7on 7me, but highest risk
Big Bang
•  Post-implementa7on produc7vity drop
•  High poten7al for system wide failures due to insufficient tes7ng/training

Parallel All/Some All/Some •  Old and new systems kept on for tes7ng period
•  Lowest risk of failure, but highest cost and longest implementa7on 7me
•  Employees do double entry work

Pilot All One •  Full implementa7on of all modules at one loca7on

•  Iden7fy bugs or issues that are corrected prior to larger rollout
•  Contains any poten7al failure from infec7ng all loca7ons
•  Tests individual modules and integra7on simultaneously

Phased One All •  Implementa7on of one module at a 7me across the network
or •  Longer implementa7on dura7on than direct, but with lower risk
Rolling •  Users have more 7me & learn as they go - no dip in performance a3er
•  Learn and fix as you go – beser process for later implementa7ons
•  Loss of managerial focus over 7me and a con7nuous state of change
•  Poten7al for missing data during transi7onal implementa7on period
•  Might require temporary bridges from old to new systems during transi7on


MIT Center for
Transportation & Logistics
27

There are a few best practices to keep in mind when going about implementation. They
include:
1. Secure senior executive commitment: ability to gather and use resources, empower
team.
2. Form interdisciplinary team(s)
3. Create a clear and specific scope document
4. Build extensive testing into the project plan (you can’t test too much)
5. Include extensive user training into the project plan
V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management
MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
58

Learning Objectives
• Recognize selecting a software vendor is an intertwined decision between architecture
and source.
• Understand tradeoffs between On-Premise and Cloud based systems.
• Know the differences between In-House, Best of Breed, ERP Extensions, and Outsourced
forms of software systems.
• Review the selection process, recognizing there are multiple attributes, and the total
cost of ownership is complex.
• Understand the challenge of implementation and the various approaches to
implementing systems within a firm.
• Review best practices of implementation.

V2 Fall 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management

MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected]
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
59