3. If you want to deliver faster,
cheaper and better, you need to
reduce your batch size.
4. Change Development Testing Staging Deploy
Cycle time (CT)
Lead time (LT)
Reaction Time
Story started
Story created
User Request
Check-In
Deploy to ProdDone
Done Done
Cycle Time
Time to complete the production of one unit from start to finish
5. Pay $3 to select all three digits at
once
Pay $1 for the first digit, find out if it
is correct, then choose if you wish to
pay $1 for the second digit, and then
choose if you wish to pay $1 for the
third digit
Pick the numbers in two ways
Front-Loaded Lottery
A Lottery ticket pays $3000 to the winning three digit number
6. Value of Feedback
Probability of occurrence
Accelerated Feedback reduce
required investment by 67%
100 %
Spend $1
10% 1%
Saving
= $0.90
Saving
= $0.99
$3$2$1
• Embedded shutdown options
• Values added by those options
• Options comes from buying
information in small batches
7. “In product development batch size is the
quantity of information, money, or other
‘stuff’ that is transferred from one state to
another at a single moment of time.”
Donald G. Reinertsen
9. Large Batch Small Batch
Unfinished Requirements
10 weeks 1 weeks
20 requirements
200 requirements
From “The Principles of Product Development Flow” by Donald G. Reinertsen
Large Batch vs Small Batch
Increased Quality and decreased frequency of bad releases
Risk = Batch Size * WIP
Risk = 20 = 1 * 20Risk = 2000 = 10 * 200
100 times as much Risk
10. 10 weeks
20
100
2 week
From “The Principles of Product Development Flow” by Donald G. Reinertsen
Batch Size and Arrival Rate
Capacity optimization and Economical Choices
11. Specify, build,
test and deliver a
SINGLE
line of code
Cost
Batch Size
Total Cost
Holding Cost
Transaction Cost
Specify, then build,
test, deliver
ALL
lines of code
U-Curve optimization
problem
From “The Principles of Product Development Flow” by Donald G. Reinertsen
12. 1 2 3 n
Batch Size and Risk Reduction
Releasing software more frequently reduce defect probability and defect cost
25%
50%
n = number of changesets
probability = (1 / 2n) * 100 [percentage]]
n=1: (1 / 21) * 100 [percentage] = 50%
n=2: (1 / 22) * 100 [percentage] = 25%
n=3: (1 / 23) * 100 [percentage] = 12,5%
Given: defect probability of a changeset ~ size of changeset
13. Batch Size and Risk Reduction
Reduced Variability
Average Fixing Time: 0,5 days
Average Number of Defects: 12 (WIP)
=
Average Waiting Time: 6 days
Sunk cost
of Bug-X
Opportunity cost
of Bug-X
1 2 3 4 5 6 7 8
10k
5k
15k
20k
14.
15. How to reduce Batch Size?
• Loose Coupling to uncouple batches
• Independently working with small sub-systems
• Parallelism and Distribution
• Reduce Hand-Overs
• Good infrastructure
• Automation
• Set WIP Limits
• DevOps practices