OVERALL THEMES Learnings from Session 1 (additional, not already SESSION 3 – Waiting-Line Service Processes SESSION 4 – Managing Quality

Ways of thinking: included) Utilisation and waiting times – general patterns: Control charts – capture voice of process, how stable over time
-Ops process approach To determine capacity of process, determine scarcest 1. Increasing avg util, increases avg waits -X-bar chart plots sample means, R-chart plots sample ranges
resource (BN) 2. Increase in util results in longer waits the higher the -By central limit theorem, sample means follow a normal distribution
-Better ops can increase quality and decrease cost Ops Process Approach:
-Global, not local optimisation starting util level is (utilization degrades performance at an with same mean as underlying process, and smaller standard
1. Understand process 2. Identify Problems
-Manage variability increasing rate) deviation (sigma/sqrt(n))
3. Identify alternatives for action 4. Evaluate alternative
-Manage probabilities (don’t take as given) 5. Implement an alternative 6. Go back to Step 1 3. Waits are worse the more variability in system -Identify and investigate samples falling outside control limits
-Consider both actual and perceived waiting 4. Waits are shorter the more capacity flexibility Process capability – meeting customer specifications (voice of cust. )
SESSION 2 – Managing Process Flows Higher the capacity, higher the ability to handle variability Calculating chances that process meets
Truths: Throughput rate: Rate at which units flow through process. specifications (centred, Cp):
-Bottlenecks determine capacity, and can move Throughput time: Amount of time flow unit spends in the
process (depends on how boundaries are defined) Calculating chances that process meets specs
-With variability, set capacity = demand will result in Inventory: # of units within boundaries of a process. (non-centred, Cpk):
huge inventories or waiting times Note: 6 sigma = 6*std deviation, 3 sigma = 3*std dev
-Pacing system to bottleneck reduces inventories and System in Equilibrium if avg # of flow units in system stays
the same over time. For either Cp or Cp : the greater, the better
wait times < 1 excessive defect rate ; = 1 normal process variation ; +/- 3 SD
-Little’s Law (I = R*T) barely meeting specf ; > 1 process variation is smaller than the range
Little’s Law: I = R x T
--Waiting time grows exponentially w/ higher I = Avg # of inventory in system of specf
utilisation R= Avg throughput rate in system (the arrival rate!) Special cause of Variation (making process out of control):
-Capacity flexibility = smaller queues and waiting times T= Avg. throughput time for flow units
-Can be traced back to specific action/problem
-Control charts: process stability over time. Capability If several independent resources work together to -Can be addressed without changing entire process
measures: how well process meets specifications accomplish a task, the capacity of doing the task is the Facility = a total servicing unit; can be composed of several
Common cause of Variation (making process not capable):
MINIMUM capacity over all the resources channels or just one
Trade-offs: -Individually small/hard to distinguish, e.g.
Channel = servicing point within a facility
-Efficiency vs breadth of service Goldratt’s Theory of Constraints environmental parameters, characteristics of process
Estimating wait times
-Batch sizes – larger batches cause longer throughput 1. Identify bottleneck Precision vs accuracy
A = average arrival rate, S = average service rate
time + more inventory, but higher capacity 2. Exploit bottleneck (maximise utilisation) Accurate: av. of observations is close to target
M = number of servers in parallel
-# of servers vs waiting times + quality of service 3. Subordinate non-bottlenecks (pace to bottleneck) Precision: observations are close to each other
U = Utilization = A/(S*M)
-Tighter parameters for quality results in more false 4. Elevate bottleneck (increase capacity) Precision is often harder to achieve than accuracy (requires reducing
L = av. # of customers in system; those being served +
alarms but faster problem identification 5. Repeat step 1 process variability (removing special + common causes of variation))
those in line
Lq = av. # of customers waiting in line (can approximate) Littlefield: Key learnings
The Goal – key learnings
Match capacity strategy with market positioning
-Running bigger batches results in less setups but longer
Variability causes lost revenues even with utilisation well below 100%
throughput times
W = total av. time in system Forecast future util. in both mean +some measure of mean’s variation
SESSION 1 – Process Analysis -Increase capacity at bottlenecks and run smaller batches at
Wq = total av. waiting time (syst. in equilibrium) Revenues are a lagging indicator of profitability; operational metrics
Process: Procedure that converts inputs into outputs non bottlenecks
CVIAT = coefficient of variation of inter-arrival times are a leading measure of profitability
Set-up Time: Amount of time to prepare before -Often, most of the throughput time in a process is wait
processing time CVST = coefficient of variation of service times SESSION 5 –Quality and Operational Excellence
Run Time: Time to process an item (excl. set up time) -Balance flow with demand, rather than balancing capacity “Coefficient of Variation” = ratio of the standard deviation Main components of TPS
Capacity: Max # of items that can be processed in a with demand of a variable to its mean 1. Standardisation 2. Visibility 3. Improvement
given length of time (gives us an upper bound of the -Focus on global optimization, not local optimization If inter-arrival and service times are exponentially Toyota: Key learnings
amount of R that we can generate in an hour/day Learnings from Session 2 (additional, not already included) distributed, coefficient of variation = 1, thus: Think about what any particular aspect of TPS requires in terms of
Bottleneck: Resource that limits production -> Waiting lines build up even at non-BN due to variability in Lq for M servers in parallel = capabilities
Utilization: Ratio of amount of resource used arrivals and variability in processing times at each stage. –Heijunka (level scheduling) is hard to achieve as it requires a
Cycle Time: Inverse of the capacity Learnings from Session 3 (additional, not already capability of high flexibility
Managing variability is key
Buffer: Storage area for WIP included) –Low inventories requires high quality
-> Running bigger batches results in higher capacity but
Throughput Time: Time for item to go through entire -> To modify characteristics, change # of channels or Toyota combines a push philosophy for long term planning with a pull
longer throughput times and more inventory. Run bigger
process. average service rate philosophy for shop floor control
batches at BN and smaller batches at non-BN
Starvation: When process stage comes to a halt due -> Both, processes and perceptions can be mngd Any production system needs to be managed dynamically
-> If there is variability, setting cap = demand -> huge
to slower pace of preceding stages -> For complex queuing systems, simulate to find best If the environment changes, even a very well thought out production
inventories or waiting times
solution system will fail