Title: Migrating from Waterfall to Lean
1Migrating from Waterfall to Lean
- Sterling Mortensen
- President, Business Productivity Consulting
- sterling_at_howtoproductivity.com
- Bret Dodd
- Process Strategist, Hewlett Packard
- bret.dodd_at_hp.com
2The Current State
- Me (The Change Agent)
- I feel like I am lying about status
- Projects are always out of control
- There are always late surprises
3HP Results w/Lean
4HP Results w/Lean (cont)
- Corporate Value
- 70 million program cost avoidance in 20 months
on Edgeline - 27 million in annual savings (FW lab budget
reduction) going forward - Program TTM reduced by 1 year
- 7.4x productivity improvements sustained during
staff budget reductions
5The Field Guide To Lean Development
6Why is status often wrong?
Reqd scope
75 of work remaining
EVM 50 of tasks complete
Only 25 of work is done
Spec
Imp
Des
Test
ReSpec
ReDes
ReImp
ReTest
Demo
Rework Again
Unplanned Work
7Unplanned Work Work Discovered Through Learning
- SW Development IS Learning
- Learning doesnt happen in nice, phased steps
- Risk reduction requires accelerating learning in
high risk areas early in a project - Rate of learning f(1/time to feedback)
8What Do You Do?The Basic Outline
- The goal is to enable work to move through your
system quickly - Organize people around cohesive work chunks
(features or requirements) - Prioritize work chunks based on value, risk, and
operating order - Have dedicated people to staff all areas of the
work chunk - Only start work chunks that can be fully staffed
- Drive each work chunk to DONE
- Learning is complete
- Tested Demo
- Defects fixed
- Once a work chunk is DONE
- Identify highest priority work chunk that can be
staffed - Make a choice
- Start identified work chunk, OR
- Utilize available staff to complete a work chunk
that is in process
9Iterative Development Life Cycle
24 Hrs
Daily Standup Meeting
1 month
Iteration
Demonstrate Functionality
Iteration features
Product Backlog features
Select
Get Feedback
10Test Strategy
- A feature or feature chunk is not done until it
has passed ALL tests - This means that tests have been run, defects
fixed, and tests rerun passing - Learning is complete once ALL tests pass! (see
speaker notes) - You should consider a test framework such as
FURPS when creating your test strategy - Functionality
- Use this as an opportunity to layer your testing
to include unit testing, subsystem testing,
system testing, exploratory testing, etc. - Usability
- Use this as an opportunity to get actual customer
feedback - Reliability
- Performance
- Supportability
- Use this as an opportunity to get feedback on
customer experience, service technician
experience, and TCO (Total Cost of Ownership)
11Test Strategy (cont)
- Test strategy has a very significant impact on
the economics of your development system - Timing of test execution
- time to feedback is the single most important
metric to understand - It drives human behavior
- It affects the time to fix a defect
- Frequency of testing
- automated vs. manual, scripted vs. exploratory
- Quality of testing
- Requirements-based coverage analysis vs. code
coverage analysis - We want to emphasize that lean and associated
test strategy is as much or more about
organizational culture and discipline as it is
about technical knowledge and skill - The success of nearly any test approache will be
determined by the culture of testing often and
fixing defects as they are found (see notes)
12Why This Works - Economics
13Why This Works Focus on Closure
- Small batch low WIP
- high productivity
- Predictability
- Linear vs. non-linear closure
- Manager satisfaction
- Employee satisfaction
- Get done sooner
- Continuous quality reduces rework
- Better economics
50 of customer value measurable
Reqd scope
50 of customer value measurable
Less than 50 of the waterfall calendar time has
been used
80 of the waterfall calendar time has been used
14Why It Works Risk Management
- Uncertainty Risk Areas the learning is needed
- Focusing early iterations on high risk areas
enables learning, reduces uncertainty risk
15Metrics Guiding Proper Execution
- Closure Rate of Features
- Time to Feedback for features
- WIP Arrival Rate
- Variability of Closure Rate of Features
- Variability of Cycle Time of Features
16Cumulative Flow Diagram
Eventual convergence indicates that backlog can
be worked off.
Avg Backlog arrival rate
Avg WIP Arrival Rate
Backlog
Average Closure Rate
Avg rates can be computed using run charts and,
later, SPC charts. Variance can be accounted for
to yield predictability
WIP
Amount of WIP queue size. Queue size has a
direct correlation to quality and cycle time
Done
17Managing Variance
Features
Cycle Time
I2
I3
I4
I5
F2
F3
F4
F5
I1
F1
Iteration
Feature
- Planning becomes much simpler when you understand
the average delivery rate of your system - Confidence in your planning grows tremendously
when you understand the variance present in your
system - Decomposition of large features into chunks helps
to reduce variance and enables time to feedback
to be short
18WIP gt 900
WIP gt 500
WIP
19(No Transcript)
20Previous two graphs covered this
timespan 3/2004-3/2005
21Common Barriers To Success
- Managerial
- Driving the wrong behaviors
- Migrating to a new lifecycle (from Waterfall to
Lean) - Org Charts Architecture
- Decomposing features into chunks and then not
driving all chunks of a feature to completion
before moving to the next feature - The roles titles assigned to people cause
artificial boundaries to crop up. - Testing organization separated from the
development organization
22Field Guide FAQ
- Measure only Items Done
- Features or feature chunks that are in process at
the end of an iteration continue on into the next
iteration, until it is - Variability is natural.
23The New Current State
- Me (The Change Agent)
- Leading Successful Change
- Creating a better place to work
- Making my boss look good
- Future Career Opportunities
24QA