MACHINE LEARNING IN ECOMMERCE APPLICATIONS

MANAGERIAL ECONOMICS PROJECT – PGPEM 2018 – Group 13

ANSHU SINGH [1816057]

ATUL JHA [1816018]
CHARU GUPTA [1816017]
DEVENDER SINGH [1816058]
GAURAV BAIDYASEN [1816006]
MANASI SAWANT [1816053]
RAHUL PANDEY [1816043]
SHRINIVAS SHARMA [1816054]
 Machine Learning in E-commerce Applications

CONTENTS
ABSTRACT ...................................................................................... Error! Bookmark not defined.
INTRODUCTION................................................................................................................................ 2
APPLICATIONS OF MACHINE LEARNING IN E-COMMERCE INDUSTRY........................... 3
a. CUSTOMER BEHAVIOR AND SEGMENTATION ...................................................................................................... 3
b. ASSORTMENT PLANNING AND MERCHANDISING ............................................................................................. 4
c. PROMOTION RECOMMENDATION ENGINE ............................................................................................................ 5
d. MOBILE ANALYTICS ............................................................................................................................................................ 8
e. PRODUCT DEVELOPMENT AND DESIGN .............................................................................................................. 12
f. MARKETING ANALYTICS ............................................................................................................................................... 13
g. SUPPLY CHAIN MANAGEMENT .................................................................................................................................. 13
CASE STUDIES IN E-COMMERCE INDUSTRY .......................................................................... 15
h. MYNTRA- ASSORTMENT PLANNING ...................................................................................................................... 15
i. AMAZON- PRODUCT RECOMMENDER ................................................................................................................... 18
j. ALIBABA SMART SUPPLY CHAIN (ASSC) .............................................................................................................. 21
k. HDFC - CONVERSATIONAL BANKING WITH CHATBOT ................................................................................. 22
l. WALMART- GLOBAL CUSTOMER ACQUISTIONS ............................................................................................... 23
PARADIGM SHIFT IN E-COMMERCE ......................................................................................... 26
m. CUSTOMER TRAFFIC/CLICK RATE .......................................................................................................................... 27
n. CUSTOMER CHURN ANALYSIS .................................................................................................................................... 29
IMPLICATIONS OF ECONOMIC CONDITIONS AND COMPETITION ................................. 31
CONCLUSION.................................................................................................................................... 32
APPENDIX: ....................................................................................................................................... 33
o. MOBILE ANALYTICS ........................................................................................................................................................ 34
p. SUPPLY CHAIN MANAGEMENT .................................................................................................................................. 34
q. INCOME .................................................................................................................................................................................. 35
r. GAME THEORY.................................................................................................................................................................... 36

1|Page
 Machine Learning in E-commerce Applications

ABSTRACT
The e-commerce industry has evolved over the past decade in multiple phases. Newer
technologies, advanced and cheaper computational power, data science, big data, machine
learning, big data and digital marketing have substantially contributed towards the growth of e-
commerce. Data growth has played an important role to understand customers in an accurate
and an efficient way. This has directed to a paradigm shift in the space of e-commerce industry.
Data is no more a by-product of the business model but it can be considered as one of the key
drivers to understand customer’s behaviour, returns on promotion, customer acquisitions and
customer retentions. Machine Learning can help e-commerce retailers offer intelligence-
powered shopping experiences to customers, increase conversions & curtail cart abandonment.
The average customer of today owns and uses at least four devices – mobile, tablet, laptop, and
a desktop/Laptops daily. Most of the customers nowadays use their smartphones to make online
purchases. This project would be focussed on different areas of application ranging from
assortment planning, marketing analytics, pricing analytics, SMART supply chain management,
promotion planning, product development and design and so on. The project covers extensive
application of machine learning in Amazon, Myntra, Walmart and HDFC. The project would also
focus on the multiple experiments which are deployed in different e-commerce businesses and
functions. Lastly, the project would discuss about the impact of economic conditions and
competitor analysis in e-commerce domain.

INTRODUCTION
Technological enhancements have helped Data Science field and many industries to grow faster
and make decisions more rapid. The integration of technology, business and data science has
not only provided a path to solve complex business problems but also allowed companies to take
data-driven decisions. Machine Learning has proved to be a vital tool to work on non-traditional
problems such as handling/managing and generating insights from unstructured data, customer
analysis, marketing analytics, computer vision, image processing, text mining and many more.
Ecommerce industry has leveraged data science and Machine Learning application across various

2|Page
 Machine Learning in E-commerce Applications

functions and channels. There are multiple fronts where ML and DS can be deployed in
ecommerce industry.

Few areas of applications are –

 Customer Behaviour – Build Demand Planning metrics, Identify the price elasticity for a
commodity, Impact of cross- elasticity across different categories and segments
 Customer Segmentation - Attributes of a customer by revenues-
 Revenues -High valued customer, medium-valued and risk customers
 Product Category – Up-sell, cross-sell and Market basket Analysis(MBA*),
recommendations
 Supply Chain Management – Inventory Planning, Logistics optimization, and Production
cost optimization.
 Promotional Planning – Build promotion recommendation engines, target
marketing(market mix modelling), Selection of promotions across categories,
Cannibalization
 Cost Optimization – Optimization of different costs as labour cost, inventory cost, logistics
cost and many more.
 Customer Acquisitions and Customer retention strategies- Marketing strategies,
customers traffic, no of visits, no of clicks, customer trajectory

APPLICATIONS OF MACHINE LEARNING IN E-COMMERCE INDUSTRY

 CUSTOMER BEHAVIOR AND SEGMENTATION
Companies today recognize that each customer comes with a varied background, values and
expectations. However, they need to make sense of these differences, figure out behavioral
patterns, cluster users in relevant buckets, and then use this information to attain marketing
goals. These are some ways in which consumer behavior data can be used effectively.

3|Page
 Machine Learning in E-commerce Applications

1. CONTENT AND EXPERIENCE CUSTOMIZATION:

We can analyze parameters like location, demographics, consumption history of the user and
offer a very personalized experience across devices to the user. Personalizing content helps
customers to engage better with the brand. Brands dynamically serve content that a
customer may like thus reducing the time spent in discovering the product and drive faster
decisions. This boosts user satisfaction, reduces churn, thereby increasing the LTV (Lifetime
Value) of a customer.

2. CROSS-SELL, RE-TARGET, BUILD LOYALTY:

Recommendations like ‘people who bought this, also viewed this’- help to cross-sell products
to customers. This information is also used to extend Offer Coupons, drive Loyalty Schemes,
and re-target old customers who haven’t engaged with the brand lately.

3. PREDICT FUTURE TRENDS:

Machine learning is used in e-commerce to predict future trends- like demand of an item,
Price Elasticity of an item or the Cross-Price elasticity of two items, especially during the Sale
Season to determine optimum price and quantity that needs to be stocked in warehouses.

Thus, analysing data based on customer behaviour helps companies increase their ROI by Faster
customer acquisition, Customer retention and engagement, and reducing customer Churn.

 ASSORTMENT PLANNING AND MERCHANDISING

With unlimited product options available within each category and with constraint of limited
inventory a retailer can maintain and display on website, Assortment Planning & Merchandising
becomes very crucial. Assortment planning is about picking the right product and ordering in right
quantity to match the market demand. Attracting the right consumers and getting them to
complete their purchases. This is where merchandising comes in to save the day.

4|Page
 Machine Learning in E-commerce Applications

Customers generally do not like cluttered display of products. Also different types of products
demands different planning. It’s easy to spot the dogs in your assortment using sales data, but
it’s far from obvious what slow sellers should be replaced with, as a slow seller you delete might
be an important product to some of your best customers. Below graph shows how Fashion
assortment planning is different from basic assortment planning.

Fashion v/s Basic Product Planning [ref]

To address this deficiency, Machine Learning techniques can make assortment planning &
merchandising vastly more scientific. It is observed that most of the time customers don’t buy
products; they buy a bundle of attributes. For instance when you say you want to buy TV, you
think about screen size, resolution, price, LCD versus plasma, and brand, not just the product TV.
Machine Learning approach sales data and various attributes to forecast the demand for
potential new products. Armed with these data, retailers can test their hunches more
scientifically. [ref]

 PROMOTION RECOMMENDATION ENGINE

Banking, Retail and e-commerce are some of the mature markets which have adopted machine
learning data science to grow their business. Ecommerce industry relies heavily on machine
learning to understand consumer behaviour, study competition, articulate promotional
strategies, deliver smart supply chain models.

5|Page
 Machine Learning in E-commerce Applications

This section is particularly focussed on interpreting promotion strategies in ecommerce space.

Some of the key areas of applications are

6|Page
 Machine Learning in E-commerce Applications

SALES PROMOTIONS are the most popular means to attract customers. There are many methods
to executive it – price promotions such as EDLP, BOGO, discounts and non-price promotions such
as vehicle promotions (aisle, sidekick and so on). Some of the key algorithms as regressions
techniques, clustering techniques and few advanced ML techniques such as Neural Networks,
Random Forest, and Support Vector Machines are widely used for forecasting customer sales
trends and segmenting them into high, medium and risk buyers. It, in turn, increases the top line
growth of a firm.

CONSUMER BUYING PATTERNS AND SEGMENTATION Micro targeting customers is a useful tool
to understand consumer buying patterns and consumer segments. It is related to segmenting
your audience lets your business offer the right promotions to the right individuals.
Customization speaks directly to consumers—signalling that you know exactly what they need.

7|Page
 Machine Learning in E-commerce Applications

Classification algorithms such as K-Means Clustering, Decision Trees, and Ensemble methods and
so on are widely used to solve segmentation problems.

CROSS-SELL AND UPSELL are another opportunities to increase the top-line growth and a great
value addition to the customer. Buyers adore convenience, and they want to save time shopping
at one place. And it eliminates the hassle of sifting through multiple ecommerce sites.

 MOBILE ANALYTICS
While Smartphone sales are getting stagnant, but their penetration and their Share in
Ecommerce is increasing rapidly.

Global Smartphone
Shipments
2
Units Shipped
(in Billions)

1.5
1
0.5
0
2012 2013 2016 2017 2021
Year

8|Page
 Machine Learning in E-commerce Applications

Analysing customer behaviour on Mobile Platform helps e-commerce firms in enhancing their
Customer Experience, Sales etc.
Using ML Models firms can analyse and react to some of the events such as Context Based Search,
Login Independent - Multiple Personas etc., which will enhance its conversion ratio and customer
experience significantly.

Interaction with
eCommerce App

Search for a Product

GEO Fencing / Awareness Multiple Personas

Same Login ID

Identify the User

Is the Person in Based on Usage
Shopping Mall? Pattern <Not only
Login ID>

Display More
Display various Optimize the App Display Product
Research Related
Buying options Design based on Information Per
Information
<Ex: Various Price Preferences (Say – Preferences of
<Ex: Various Models
Options> Theme etc) Identified Persona
of Product>

9|Page
 Machine Learning in E-commerce Applications

OTHER BENEFITS OF ML AND SMARTPHONES

VISUAL SEARCH
Using Image Recognition Technologies and AI, e-commerce Vendors can provide Visual Search,
to the customers.

PRODUCT DESCRIPTION SUMMARIZATION

Considering the small form factor of smart phone, Machine Learning has started playing a
significant role in displaying information per user preferences.

10 | P a g e
 Machine Learning in E-commerce Applications

EMOTIONS AI
Technology that recognize and decipher a person’s mood will help e-commerce since it can satisfy
their customer needs considering emotions as well in its decision making.

AUGMENTED REALITY
With advent of powerful mobile processors and technology to support Augmented Reality (AR
Mobile Headsets), in the future e-commerce website will start using AR to help customer choose
the product after trying it out virtually.

To summarize, Machine Learning is the heart of promotional recommendation engines.

Ecommerce industry has witnessed a significant shift in demand through the application of

11 | P a g e
 Machine Learning in E-commerce Applications

Machine Learning Customer Acquisitions and Customer Retention Strategies are primarily
achieved through smart promotional strategies.

 PRODUCT DEVELOPMENT AND DESIGN

The use of machine learning is not only restricted to collecting data and utilizing it for servicing
the customer. As the AI technology has improved multi-fold in recent time it has influenced the
way to transform the design process. Machine learning is being used by varied industries such as
architecture, web design, car design, and even in composing music.

For instance, an AI-based platform, Shimon, developed at the Georgia Institute of Technology in
the US, is already composing new music.

Several companies and start-ups have tried to automate the process of design in fashion industry
too. Some fashion brands are experimenting with IBM’s supercomputer Watson and Google’s
technology tools to create designs. StitchFix, an American online retailer, has a tech platform
called Hybrid Designs; Myntra has platform-Rapid that automates fashion design using AI. Dozens
of AI start-ups in fashion, including Stylumia, are attempting to do what Myntra is doing and
more.1

12 | P a g e
 Machine Learning in E-commerce Applications

 MARKETING ANALYTICS
Today Machine learning algorithms are employed extensively by companies for marketing
analytics. Ad-networks, Programmatic Ad space buying, advanced targeting on social media
platforms have evolved because of this.

Brands often use DMP (Data Management Platforms) to store, analyse and collect 1st and 3rd
party information of user’s journey online. This information is then used to serve ads to targeted
users based on certain parameters (audience tracking) or re-target high intent users with
products that they have recently seen.

For example If you look at a certain item on Bluestone and then leave the website without
purchasing anything, the Google ad platform will pick your browsing history from Bluestone
website and show you these items on Google Ad spaces.

Google has created SEM algorithms that are able to manage your ad campaigns and place bids
automatically on your behalf. Once you define a set of goals, keywords, and type of audience to
target, the SEM algorithm will learn and optimize on its own, even pick best performing creatives
from a set of creatives.

These algorithms have changed the way brands market today and made it possible to have a
targeted and measured marketing communication.

 SUPPLY CHAIN MANAGEMENT

Supply Chain management is the management of movement of goods/raw material from
suppliers/producers to consumers in form of finished goods involving several important steps
such as sourcing and procurement, demand planning and inventory management, and logistics.
In e-commerce, it is essential that movement of goods happen in time bound and orderly fashion
in both ways (B2C and C2B), hence the need of an efficient Supply Chain Management.

Key activities of Supply Chain in e-commerce consist of following functions:

13 | P a g e
 Machine Learning in E-commerce Applications

1. SUPPLY CHAIN REPLENISHMENT: Integrated Production ,distribution and procurement

processes to increase replenishment and reduce inventories by synchronising Demand and
Supply
2. DEMAND PLANNING AND INVENTORY MANAGEMENT: Integrated software to predict
demand and manage the inventories in network of warehouses
3. LOGISTIC AND TRANSPORTATION: Optimal distribution system for transportation of goods

Key challenges in supply chain of e-commerce businesses include:

 Finding the right product to sell
 Predicting the demand for future
 Cost optimisation on inventory
 logistics to meet time bound delivery

SMART SUPPLY CHAIN in e-commerce uses AI to help forecast product demand; prepare,
replenish and allocate inventory for optimal turnover; determine the right products to offer; and
smart logistics using AI to find the cost-effective delivery routes.

14 | P a g e
 Machine Learning in E-commerce Applications

CASE STUDIES IN E-COMMERCE INDUSTRY

 MYNTRA- ASSORTMENT PLANNING
Myntra has taken advance steps as compared to other in applying technology in fashion design:
it seems to have figured out a solid process. Broadly, there are three reasons why Rapid is ahead
of other similar efforts.

1. Myntra has lots and lots of sales and browsing data. The company has access to customer
data on Flipkart and Jabong as-well. These three platforms control roughly 70% of all
online fashion sales in India, according to Myntra estimates.
2. Myntra has the cash to invest in AI technologies, such as GPUs, and attract top engineers.
3. It has built expertise in the supply chain side of fashion and has close relationships with
suppliers and manufacturers because of its large private label business. Myntra is able to
convince suppliers to shrink the process from 180 days to less than 45 for its fast- fashion
products.

Many companies could achieve one or two aspects but few can combine all the three. “What
we’re doing is not just a technology solve or a supply chain solve,” said Rajesh Narkar, vice-
president of Myntra’s private labels business. “It’s about disrupting the way brands are built.
Today, we’ve reduced the (manufacturing) process to less than 45 days. The idea is to do it even
faster. We’re asking ourselves: if there’s customer intent to buy something, can we serve it

then?” 30

15 | P a g e
 Machine Learning in E-commerce Applications

How Data Gathering Works For Myntra Design Creation

When the Rapid team planned to develop the ideas of generating completely Machine Learning
driven designs, the team came across GANs (Generative Adversarial Networks). Ian Goodfellow,
a researcher at google, Invented GANs.

16 | P a g e
Machine Learning in E-commerce Applications

Neural networks, which comprise connected

transistors that replicate the structure of
neurons in the human brain, are a fixture in
AI. GANs go further. They comprise two
neural networks. In Myntra’s case, for
instance, one network comes up with the
designs after being fed data. The other
network critically vets the designs and gives
feedback to the first network on where it’s
going wrong. Through this interplay, the two
neural networks are trained to think like
humans.

Source: http://fortune.com/ai-artificial-
intelligence-deep-machine-learning/

17 | P a g e
 Machine Learning in E-commerce Applications

The Rapid team at Myntra trained and tweaked the GANs for more than a year, and kept
consulting with Goodgellow to make the best use of nascent technology. Finally in September
2017 the team was ready with new designs to go live on the platform.

The designs were included in 2 private brands of Myntra, Moda Rapido and Here&Now. These
two brands together do a sales of 12-13 crore INR every month and are Myntra’s most profitable
brands, generating net margins that are around 10 percentage points higher than the others. 2

 AMAZON- PRODUCT RECOMMENDER

When it comes to the e-commerce sales of U.S. retailers, Amazon Inc. dominates the list of the
top 10 companies — and is followed by Apple Inc., Wal-Mart Stores Inc. and Macy’s Inc.

18 | P a g e
 Machine Learning in E-commerce Applications

POSSIBLE REASONS OF AMAZON’S SUCCESS

Today shoppers don’t want to do all the legwork in the path-to-purchase. According to a survey,
64% of respondents wanted retailers to provide smart recommendations or suggestions based
on prior purchases to aid in the buying decision.

[ref]

Apart from other reasons such as fastest on door delivery, largest catalogue, and amazing
customer service, secret key to Amazon’s growth is their hyper charged recommendation engine
(on-site recommendations, email recommendations, and recommendations with Facebook

19 | P a g e
 Machine Learning in E-commerce Applications

Dynamic Product Ads) and their constant implementation of machine learning in every part of
their business.

Amazon Onsite Recommendations features such as “Recommended for you”, “Frequently

Bought Together”, “Recently Viewed” has made browsing the site easy for customer by putting
products in front of you that you’re likely to click, learn more about and buy.

Amazon is not only targeting customers who are online and are on Amazon right now, but only
is targeting them when offline by recommending products offline via email. Highly relevant
emails are critical for improving your click-through rate, conversion and revenue per email
metrics.

If Amazon started sending disconnected emails, these emails would likely be marked as spam or
they’d just unsubscribe from future mailings.

The Recommendation email containing items frequently purchased together, with the aim to get
you to buy a product you were interested in a while back and cross-sell its accessories in order
to increase average order value and the amount of revenue they generate from each customer.

Holistically, recommendation engine is an effective way of creating a personalized shopping

experience for each customer. It has helped Amazon increase average order value and the
amount of revenue generated from each customer.

20 | P a g e
 Machine Learning in E-commerce Applications

 ALIBABA SMART SUPPLY CHAIN (ASSC)

 Smart Supply Chain:

 Smart Logistics:

21 | P a g e
 Machine Learning in E-commerce Applications

 HDFC - CONVERSATIONAL BANKING WITH CHATBOT

A big new application of machine learning is in chatbots. Of late, e-commerce websites have been
applying chatbots, or software conversing with the user through auditory or textual methods, to
aid the customer find the product they seek, track order status, and to answer other general
queries. Customers get instant and accurate answers, without having to put up with the difficulty
of calling and waiting for a live agent attend their call at the customer service.

HDFC Bank has joined hands with Niki.ai, the artificial intelligence firm to bring in ‘conversational
banking’. In May this year, Ratan Tata and Roni Screwvala’s Unilazer Ventures picked up stakes
in Niki.ai.

This service allows bank customers to conveniently access their bank accounts through chatbots
that facilitate commerce and banking transactions without getting out of the chat window. With
over 71% of our transaction taking place on digital platforms, internet and mobile are now core
banking channel. [ref1][ref2][ref3][ref4]

22 | P a g e
 Machine Learning in E-commerce Applications

Hence, Bot implementation for retailers is seeking to make the entire customer experience
seamless and relatable. Equipped with qualities ranging from intelligence to empathy, bots are
here to stay in the retail industry. What is left to be seen is how it matches the same experience
of an in-store purchase. Retailers will want to use bots to guide customers, such as those
searching for gift ideas. Brands can also gain from bots by offering storytelling, product
exploration, and other such marketing functions.

 WALMART- GLOBAL CUSTOMER ACQUISTIONS

Walmart is on its way to transform into an e-commerce giant and this started back in September
2016 when they bought jet.com to get into the competition with Amazon.com in United States.

Since then Walmart has been on an acquisition spree and targeting e-commerce focused retailers
such as Moosejaw.com, Bonobos.com, Parcel.com and Flipkart.com

23 | P a g e
 Machine Learning in E-commerce Applications

Jet.com

•Acquired by Walmart in September 2016 on subsiduary based model for 3.3 $ billions.
•March 2016- Bought Hayneedle.com ,online specialty store that emphasizes on indoor/outdoor home furnishings,
decor and other furnishings for commercial use
•December 2016- Bought shoes.com, online footwear retailer from Boston
•March 2017- Modcloth.com, American online retailer in indie and vintage-inspired women’s clothing

Moosejaw.com

•Acquired by Walmart on February 15th, 2017

•Its an online and B&M retail store focusing in outdoor recreation apparel and gear for camping, hiking, rock climbing
and snowboarding.
•brought along its partnerships with around 400 brands.
•“Activist organization” Patagonia Incorporation ,An American clothing firm that sells sustainable outdoor clothing
using down feathers and wool material.
•The North Face Incorporation (Another American outdoor product organization producing outwear and equipment for
campers, climbers, hikers, mountaineers, skiers, snowboarders and athletes and included various level of customers
such as college students , enthusiasts and professionals);
•Marmot (An outdoor clothing and sporting goods firm founded in 1974 as Marmot Mountain Works), and Arc'teryx
(Canadian outdoor clothing and sporting goods firm , founded in North Vancouver, British Columbia in 1989).

Bonobos.com

•Acquired by Walmart on June16th, 2017

•Its an e-commerce firm located in NYC which designs and sells men’s clothing .
Parcel.com

•Acquired by Walmart in September 2017.

•Parcel is one of a kind New York based delivery service (24/7) that uses routing algorithms to sort, load and deliver
package to customers.
•Parcel exposes the ‘Parcel API’ and its easy integration with RESTful API is quick and orders can be created in JSON and
send over using https protocol and the order window response by Parcel will be returned in form of JSON which can be
easily used and applied by the websites and application with minimal changes at their end.
•Parcel have their own employees based workforce and fleet of vehicles (mostly leased). The strategy is to use this
service to better the experience and to deliver and the regular goods and grocery (from both Walmart and jet.com) to
the customer on the same day or as per their convenience.
•Parcel has very unique design which allows them to offer their services generically to any website or application intend
to offer the delivery of their products in NYC.

Flipkart.com

•Acquired by Walmart in May 10th 2018, Flipkart is an Indian electronic commerce company based in Bengaluru, India
(with its subsidiaries Myntra.com and Jabong.com) for 16 billion dollars, making it the world’s biggest ecommerce deal.
•This is Walmart’s step towards beating Amazon in Asian markets by targeting Indian customers.
•Flipkart has already been working with Data Scientists on creating Machine Learning based models to optimize the
address classification and eliminating the fraudulent addresses, optimizing the logistics and delivery time of the
products.
•Flipkart uses eLoc (based on digital mapping and deep mapping technology) for solving last-mile logistics. With this
solution, they are able to resolve the inconsistencies and identify the correct address with 98% accuracy rate.

Timeline for E-Commerce Based Acquisitions by Walmart

24 | P a g e
 Machine Learning in E-commerce Applications

Shoe.com

Modcloth.com
Hayneedle.com
Jet.com

Parcel.com Moosejaw.com

Walmart.com

Myntra.com

Flipkart.com

Bonobos.com
Jabong.com

Walmart’s Net of E-Commerce Acquisition

With all these e-commerce acquisitions, how is Walmart’s going to make sense of it all? The
answer is WalmartLabs. WalmartLabs is premier retail research lab where they are uncovering
various innovations on retail to revolutionize Walmart’s approach towards digital commerce.
1. Platform developed around Performance Ad, displaying contextual and targeted product
advertisement to the customers, their teams are working with Data Scientists and
Researchers to develop prototypes, proof of concepts and implement data models for
improving monetization for Walmart and enhancing the online shopping experience.

25 | P a g e
 Machine Learning in E-commerce Applications

2. Walmart has huge inventories, retail warehouses and therefore they are rich in customer
data. Using ML algorithms they have developed model for logistics and warehouse
stocking which optimizes the process of stocking the goods and maintain the inventories.
They’ll be able to predict customer’s demand triggered by events such as natural disasters
(hurricane/ blizzards), elections or Christmas. Using these algorithms they can stock the
products expected to be in demand, maintain their inventories and can meeting the
expectations of the customers.
3. Product Matching in e-commerce applications using deep learning.
Deep neural networks and image recognition is used for identifying products from their
image and using the same logic, similar products can be identified and suggested to
customer pertaining to massive technological advancement. These algorithms are also
used in identifying the facial expressions and classifying the mood/emotion.

These series of e-commerce acquisition has helped Walmart to grow in retail market. With
majority customer base as millennials, Walmart’s target is to compete with existing online giants
by creating an online ecosystem offering competitive prices for various range of products and
fast delivery /pickup services for customers. With their current financial framework of strategic
capital investment, operating discipline and efficient growth, their goal is to spend money wisely,
to invest in e-commerce, technology and increase revenue, which will lead them to offer and sell
the products at lowest prices to their customers resulting in a win-win.

PARADIGM SHIFT IN E-COMMERCE

E-commerce industry has brought paradigm shift in the business community and the coming
times looks ever promising. The idea is to make the experience time-efficient, secure and
optimized for both producers and customers without compromising the quality. Creating models,
identifying trends using Machine Learning brings significant promise to the future of e-
commerce.

26 | P a g e
 Machine Learning in E-commerce Applications

 CUSTOMER TRAFFIC/CLICK RATE

We all get the marketing e-
mails/calls with a link suggesting
a product, which could be new or
created around our
purchase/browsing history.
Based upon our requirement, we
take the action (click/not click) and when the link is clicked, the response gets recorded. This
scenario aids the e-commerce firm to gather their customer click rate numbers, also abbreviated
as CTR (Click through ratio).

CTR
Click through ratio is the ratio of ‘total number of
customers who clicked the link’ to ‘total number of
impressions (people who saw the link, includes both
clicked and non-clicked)’and is a useful parameters in
identifying the products for customer considering
personalization and is also used to measure the success of
an advertisement campaign.
High CTR indicates relevancy and trend of the
product/subject under the lens although it’s not the sole
parameter to determine the success of a campaign.
CTR is a recursive process and proper keywords and
responses increases the accuracy of product
recommendation.

Click Through Ratio (CTR)

27 | P a g e
 Machine Learning in E-commerce Applications

Abnormal percentage change in CTR explains a new phenomenon (addition of a new variable in
the demographics of the campaign or vice versa) and hence the change in strategy as per the
new output and meeting the requirements.

Customer Traffic Rate Customer Click Rate Conversion Ratio

• Total number of • Customer Click Rate is • Conversion ratio is the

customers who visited the ratio of ‘number percentage of total
the e-commerce of clicks” to ‘number number of actions
website at a particular of impressions” for a completed out of the
time. particular online total number of
• Customer traffic rate campaign. visitors on an e-
helps in analyzing the commerce website or
customer behavior, application.
effect of a promotion
or change in pattern
of inflow of customers
due to external
factors, affecting the
overall sales.

Comparison of Customer Traffic Rate, Customer Click Rate and Conversion Ratio

Customer traffic, Conversion Ratio and CTR are correlated and helps in collecting intelligent
information which can be used to create products as per the demand. Collected data is used to
develop training data for machine learning models and new patterns can be identified. Better the
training data, better classifier can be developed and hence the consistent patterns leading to
better predictions. These patterns could be next coming of age phenomena which will help in
creating/offering better products and ending the ones which are not required anymore, the
redundant ones.

28 | P a g e
 Machine Learning in E-commerce Applications

 CUSTOMER CHURN ANALYSIS

Customer Churn is defined as the number of customers who have left their current e-commerce
platform (unsubscribed from the e-mail list, deleted the mobile app) and moved on looking for
better/new platforms. In this age where offering are in form of SaaS, it’s very crucial for an e-
commerce platform to retain their customers as getting new customers can be multiple times
costly, compared to retaining the existing ones.

•Customer dissatisfaction, it could be related to a bad customer service ,

defective product, or unpleasant experience while doing online shopping.

•Transactions not going fine properly, multiple failures, all mode of payments
were not offered.

•Security of customer data was not taken care of. Spams and privacy issues

• Bad delivery/return Service, Customer's time/preferance not respected.

Basic Reasons of Customer Churn

The idea is to identify the need of the customer on a regular basis, meeting the expectations and
taking constant/ recursive feedback until the issue is resolved. This can be done by better

29 | P a g e
 Machine Learning in E-commerce Applications

customer service, marketing campaign, or in some cases dispensations as well. For an e-

commerce platform, Customer Churn Analysis can be complex, although a subscription based
model eases the complexity for data scientist.

The number of inactive users can be identified and can be offered incentives to create
interest and hence the activity.

The users are active but not purchasing anything for a long time. Gift-
Vouchers/’crediting mobile wallets with amount’ can be offered in such scenarios
along with suggestion of personalized products.

Attempt to add new customers can be done using reference model, where the
mobile-wallet gets credited with the gift amount (for both referrer and new
customer) as soon as referred customer registers/subscribe.

24/7 Customer Service for the better experience in future, to implement the changes and
meeting the expectations. Using crowdsourcing for new ideas and identifying the trends.

Churn Analysis Application Aspects and Benefits.

In today’s fast-paced word, customers are impatient, impulsive and needs to be gratified. If it’s
too late and the customer has already made up their mind to move on, may result in an increase
in customer attrition rate. To be in good books, ‘customer vulnerability’ should be taken care of
by developing the platform adhering to local as well as global standards and policies. Using ML,
solutions can be automated as a part of business flow using ‘Churn Analysis’ output.

Template Explaining Basic Building Block of Churn Analysis.

30 | P a g e
 Machine Learning in E-commerce Applications

Machine learning identifies the customer behavior and suggests about red-flags/good-things
about the product. Solutions can be created/modified after studying the attrition specific
patterns, and further development can be suggested based upon the richness of data.

IMPLICATIONS OF ECONOMIC CONDITIONS AND COMPETITION

Use of Machine learning has affected all the major activities of the ecommerce like pricing,
product development, marketing, supply chain, consumer behaviour and the most basic
transaction in economy which is to buy and sell a goods/services i.e. the speed with which they
will be able to conduct sales transactions which will have an impact on the way companies do
their business below are some of the key area where it has helped industry to do their business
in an effective way

STREAMLINING BUSINESS OPERATIONS TO INCREASE THE VALUE OF BUSINESS.

Machine learning gives advantage to producers as this will help in using the existing capital and
labour effectively which will bring business a new source of growth with enhanced level of
productivity and efficiency, ML incorporate data from all aspects of the business providing
Managers to take the informed decision with historical and streaming data to enhance
operations starting from procurement of raw material to finished product to quality assurance
to compliance for regulation. In the above customer segmentation and behaviour we have tried
to address the aspect of production which is to produce the product as per the need of the
customer by analysing parameters like location, demographics, consumption history of the user
and offer a very personalized experience.
Assortment planning is another area where we have seen a significant impact where the time
line of the producing a product came down from 180 days to less than 45 days. Idea is take that
to next level where producers can serve the customer if they intent to buy they should be able
to produce.

31 | P a g e
 Machine Learning in E-commerce Applications

STREAMLINIG SCM FOR OPTIMAL DISTRIBUTION SYSTEM

Machine Learning provides an opportunity to improve SMC by providing insight into the supply
and demand aspects which earlier technology was not able to address, this leads to minimizing
the risk at the supplier side, reducing the shipping cost, improve in supplier delivery performance,
lower inventories, planning production along with factory scheduling and this helps managers to
get end to end visibility

COMPETITION
Machine leaning is used extensively to price the commodities /goods in market we have seen
intense pricing wars among the enterprises to lower the price to reasonable level this will
definitely help both customers as well as the enterprise itself to bring down the cost of running
the organization. Organizations/Enterprises have used all the aspects of game theory like
prisoner’s dilemma to price commodity/Goods to come to Nash equilibrium to price (Low Price,
Low price) which ultimately lead to the profit maximization for enterprises and consumers make
utility maximization until enterprises does not fall into the price cut trap by reducing the price at
the cost of profit which will lead to destruction of the competition or if they form cartel and
increase in price this will lead to loss of customer confidence. Enterprises should be rational while
dealing with the prices.

CONCLUSION

Machine Learning in e-commerce has an impact at the microeconomic level of retail, wholesale
and labour market transactions which will have an enormous impact on the performance of
companies, the economic welfare of consumers and the labourer working in the industry.
Actionable Information lead growth will further increase productivity, reduce cost and intensify
already existing competition between players which will ultimately lead to consumer lead growth
of the industry.

32 | P a g e
 Machine Learning in E-commerce Applications

The gains in efficiency in the ecommerce and economic benefits at the microeconomic level will
affect the Macro was well as the social economic behaviours of the society but that will be
restricted due to the access to ecommerce system which in country like India will not reflect the
right macro conditions but for develop world it will be one of the factors for macro economy
going forward. Government need to form better policies such that the competition is not killing
business/ small time players they need to persuade, incentivise business such that they are able
to effectively build a sustained structure to help the growth of the industry by taking consumers
interests also into consideration.

APPENDIX:

MICROECONOMICS /MANAGERIAL ECONOMICS IMPLICATIONS

INFORMATION ECONOMICS
Machine Learning will help in reduction of the asymmetry information which will lead to reduce
in transaction cost especially on the information transaction cost as the cost of actionable
information currently used by consumer survey and other methods are costly and are not able
to provide the full picture of the need and demand of the society.

INDIVIDUAL /CONSUMER AND MARKET DEMAND

Consumer preferences will be marked, and information will help to come to the right preference
of the consumer, which will reduce the time to shop/buy a product which will be helpful for
consumer to utilise the time in any other activity

INDUSTRY/PRODUCER
Machine Learning will help organization/companies to act accordingly to serve the customers
even they can come up with right signalling like warranty/guarantee if people are not taking some
brand or product. PA problem which arises due to asymmetry will be reduced as organisation can
set the right correct capacities and wages and incentive can set accordingly

33 | P a g e
 Machine Learning in E-commerce Applications

GOVERNMENT
Government will be able to form right policy for taxing so that the burden on the taxes falls from
most of the buyers as they will also have the required information so that government can work
on the right tax levels (VAT, GST etc.) and subsidies on the required items for the common people
overall there can be reduction of price control from government or concern agency not to get
too involved in policing than to form right policy framework.

MOBILE ANALYTICS
BEHAVIOURAL ECONOMICS
It will help industry to understand the pattern of the demand and the need of the people at the
right time and which will help industry to be better prepared for the future demand and changes
that are needed in their product portfolio.

INDIVIDUAL/CONSUMER AND MARKET DEMAND

Consumer will also benefit as the demand for the product is delivered in time and which will
ultimately lead to reduction in cost and time to get on demand product.

UNCERTAINITY AND CONSUMER BEHAVIOUR

Consumers will not part way with the product /service if they are able to get that at the right
price else due to substitution effect demands my shift to other product which will be
counterproductive for the supplier/producer.

SUPPLY CHAIN MANAGEMENT

SUPPLY AND DEMAND
With the use of ML there can be reduction in the arbitrage in the market i.e. the assumption of
the perfectly competitive market could be closer to become a reality, Supply and demand gap
will be reduced as we will have right demand which we can provide to the suppliers / produces
so that there is no gap in the demand and supply which will help companies to run the business

34 | P a g e
 Machine Learning in E-commerce Applications

at optimal level with winning strategy for long term survival and the right price will result in
economic equilibrium

COST OF PRODUCTION: PRODUCER

Cost of production will come down if the right information is passed in the right time to the
producers which will help them to have supply /output with help of setting all their factors of
production in the right mix

INDIVIDUAL/CONSUMER AND MARKET DEMAND

As the demand will met with the predictive and other methods consumer will be happy to get
the required product in time and at the right price

WELFARE ECONOMICS
Cost–benefit analysis can be performed better by the organization and it will lead to forming a
better policy framework by government based on the trending analysis with the right set of
information market failure will not occur which will reduce the burden on the society, even to an
extent externality like pollution can be reduce by right supply chain and for reducing the market
inefficiencies

INCOME
PROFIT MAXIMIZATION FOR AN ORGANIZATION/PRODUCER
Since the organisation will have the demand which is as per the consumer need they can
maximise their return by running as efficiently as per the demand

LABOUR WAGE
Since the plant might be running efficiently there will not be too much discrimination in the
wages of the management and labour as the wages determined will be as per the market demand
Skills and Substitution effect on the employment will be reduced .

35 | P a g e
 Machine Learning in E-commerce Applications

GAME THEORY
“Prisoner’s dilemma” game can indicate the reasons for the price war, as in this classic case where
purely rational self-interested prisoners will betray each other, and the only possible outcome
for two purely rational prisoners is to not cooperate even if it’s in their best interest.
On similar lines we consider two e-commerce firms (oligopolistic companies) in market which sell
homogeneous product and each wants higher market share to earn higher profit.
Below are the conditions that they meet
1) Information of both their prices are in public
2) Repeated Game both companies are changing and their prices based on others price

In this case firms find themselves in prisoner’s dilemma where if one cut the prices then other
will also cut the prices to capture the market, they know that if they collude they can gain but
they are more worried if other party default so this dilemma continues to have a price war among
them.

NASH EQUILIBRIUM IN PRICE WAR AND CARTELIZATION

In the above scenario we know that (cooperate, cooperate) is the Nash equilibrium then for
above case (low-price, low-price) is the best suited Nash equilibrium for the ecommerce
companies but this will be a disaster for their existence that will result to zero profit for them but
every company tries to form a normal low prices strategy to form a monopoly price and then get
monopoly price (if they succeed then they earn bigger profits) which is usually higher price.
Another case can be pricing the goods/services normal and both make profit. In fact, the perfect
competitive equilibrium is the “Nash equilibrium” or “non-cooperative game equilibrium”. In this
equilibrium, each enterprise makes profit maximization, consumer makes utility maximization, it
will result in a zero profit, and namely, price equals marginal cost. If two online retailer’s decision-
makers are rational, then there is no price war between them. Because the best choice depends
on the choice of the other in this game, when one chooses high-priced that the other do so best,
but when the other choice is low-priced, best choice is high-priced, so each retailer hopes to

36 | P a g e
 Machine Learning in E-commerce Applications

achieve not only their own (high-priced, high-priced), it also expects the other one will choose a
high-priced, so (high-priced, high- priced) should be reasonable result of the game.

Price competition cannot take place repeatedly in reality. So, if there is a possibility that
enterprises can avoid price wars, reaching win-win results under certain price competitions.
There is a probability that those survived firm could coordinate the price. Oligopolies can cease
the price war and enjoy the high profits together, if the requirement of cooperation is met. For
this very reason there is good chance of price collusion happening or cartelization.

37 | P a g e
 Machine Learning in E-commerce Applications

References:
1. https://www.thestreet.com/story/14398943/1/walmart-not-one-dealmaking-says-ceo.html
2. https://economictimes.indiatimes.com/industry/services/retail/softbank-ceo-confirms-walmart-
flipkart-deal/articleshow/64093437.cms
3. http://www.therobinreport.com/walmart-jet-com-can-crush-amazon/
4. https://www.adweek.com/digital/walmarts-ecommerce-ceo-explains-why-its-many-
acquisitions-will-help-it-reach-millennials/
5. https://stories.flipkart.com/ai-ml-flipkart-indian-address/
6. https://www.fromparcel.com/developer/api/docs
7. https://medium.com/walmartlabs/product-matching-in-ecommerce-4f19b6aebaca
8. http://ilfstechnologies.com/wp-content/uploads/2018/01/customerchurn.png
9. https://gallery.azure.ai/Collection/Retail-Customer-Churn-Prediction-Template-1
10. https://towardsdatascience.com/churn-prediction-with-machine-learning-e6612cd5538f
11. https://insideretail.asia/2017/06/07/how-alibaba-uses-artificial-intelligence-to-change-the-way-
we-shop/
12. https://www.azoyagroup.com/blog/view/retail-tech-innovation-in-china-new-retail-smart-
supply-chain-ai/
13. http://wps.prenhall.com/bp_turban_ec_2012/207/53241/13629779.cw/content/index.html
14. https://www.digitalcommerce360.com/article/us-ecommerce-sales/
15. http://altitudelabs.com/blog/amazon-product-recommendation-engine/
16. https://www.martechadvisor.com/articles/customer-experience-2/recommendation-engines-
how-amazon-and-netflix-are-winning-the-personalization-battle/
17. https://blog.wiser.com/modern-guide-assortment-optimization/
18. https://www.nchannel.com/blog/ecommerce-merchandising-strategy/
19. https://www.cgsinc.com/blog/what-assortment-planning
20. https://www.digitalcommerce360.com/2017/08/09/e-commerce-grow-17-us-retail-sales-2022/
21. https://www.abetterlemonadestand.com/ecommerce-chatbots/
22. https://suyati.com/blog/chatbots-retail-marketing-future/
23. https://suyati.com/blog/hdfc-bank-introduces-chatbot-banking-niki-ai/
24. https://suyati.com/blog/machine-learning-helps-ecommerce-taxonomies/
25. https://www.abetterlemonadestand.com/ecommerce-chatbots/
26. http://www.parkeravery.com/pov_Fashion_vs_Basic_Assortment_Planning.html

38 | P a g e
 Machine Learning in E-commerce Applications

27. https://www.smartrmail.com/blog/3-ways-to-achieve-amazon-like-sales-growth-with-product-
recommendations/
28. http://inspirationfeed.com/dynamic-content-the-secret-ingredient-for-email-marketing/impact-
of-recommendation-engine-on-amazon-and-netflix-3/
29. https://www.livemint.com/Companies/tchtq74FOkMM43szMIHh0M/At-Myntra-machines-tell-
designers-how-to-make-clothes.html
30. https://www.livemint.com/Companies/tchtq74FOkMM43szMIHh0M/At-Myntra-machines-tell-
designers-how-to-make-clothes.html
31. https://www.livemint.com/Companies/tchtq74FOkMM43szMIHh0M/At-Myntra-machines-tell-
designers-how-to-make-clothes.html
Books and Reports:
1. Gartner Studies: Market Guide for Digital Commerce Search, Published: 12 January 2017 ID:
G00311302
2. 2017_Internet_Retailer_Top_500_Report

39 | P a g e