Measure Twice, Cut Once: Using Team Operation Metrics to Optimize a Scaling Software Shop

VMware Tanzu
VMware TanzuVMware Tanzu
Confidential │ ©2021 VMware, Inc.
Krista Luna & Daniel Witenberg
Measure Twice, Cut
Once: Using Team
Operation Metrics to
Optimize a Scaling
Software Shop
VMware Tanzu Labs
2021-09-02
Confidential │ ©2021 VMware, Inc. 2
Who Might Find This Valuable?
Core audience, but not limited to:
Studio Leads Practice
Leads/Coaches
Business Owners
Confidential │ ©2021 VMware, Inc. 3
Program Health is an assessment
of the collective health of the set
of teams in a lab, studio or larger
grouping
What is Program Health?
By Program, we mean any collection of teams—Portfolio, Lab, Studio
4
Confidential │ ©2021 VMware, Inc.
A common challenge we find when
joining clients on site, is the client’s
desire to scale and mature their
Agile software org, without scaling
the human capital investment to
managing that growth.
Confidential │ ©2021 VMware, Inc. 5
Organization are hungry for an
approach that allows for a large
and growing number of teams to
be managed and optimized by a
small team of program-oriented
practitioners—which grows
comparatively slowly.
Confidential │ ©2021 VMware, Inc. 6
With healthy quantitative
measurement of program
optimization, practice leadership at
large enterprises can manage
increasing scale more effectively
than with qualitative tactics alone.
Confidential │ ©2021 VMware, Inc. 7
Practice Coaches can manage far
more team analytics than they
can team-level qualitative
intervention, leading to cost
savings around Practice
Management.
Teams’ progress on their Agile
maturity path will reflect naturally
and organically in their metrics
Coaching teams scale slower
than application teams
Benefits of Quantitative Approach
Measuring Team Health Maps to Program Health
Agile Maturity can be clearly
measured
Confidential │ ©2021 VMware, Inc. 8
Benefits of Quantitative Approach
Creating a common language
A common and credible language in which
program health can be described and
discussed with stakeholders who don’t
come into regular contact with the program
teams is crucial.
That common language needs to be
meaningful to the business, consistent
across the organization, and provide
objective verifiability.
Confidential │ ©2021 VMware, Inc. 9
Benefits of Quantitative Approach
Manage by Exception as a Scaling Growth Model
Respond Only Where
there is need
Using this approach, we
‘don’t fix what ain’t broke’
Scale Exponentially,
but Sustainably
Fewer Practices Coaches
needed for more teams
A quantitative approach allows us to Manage by Exception, which is to say, focus
on only those things that fall outside of our accepted standards.
Confidential │ ©2021 VMware, Inc. 10
What this isn’t!
Our approach will enhance other tactics, not replace them
DORA
DORA measures DevOps
lifecycle health.
Team Health
Checks
Provides valuable leading
and lagging feedback
indicators.
OKR/KPI/JTBD
Our approach doesn’t
replace team’s own work
to track their optimization
and set goals for
improvement.
11
Confidential │ ©2021 VMware, Inc.
Dashboarding a Program
Team performance transparency is the key to
sustainable scaled Program optimization
Confidential │ ©2021 VMware, Inc. 12
Anyone with an interest in the
Program should have a clear line
of sight into the metric and
understand that it’s credibly
connected to the thing it’s
supposed to measure.
To be useful, program metrics are just like the metrics we teach our PMs about for
product dev. They need to be:
A change in a metric means a
change in how optimized the
program is, so any metric we use
needs to be something we can
act against when things are
off-kilter
It needs to be easy to
understand, track and synthesize
meaning from a metric. They
shouldn’t create more work for a
practice lead to explain, than
they are saving in other areas.
How Do We Measure?
Actionable Accessible Auditable
13
Confidential │ ©2021 VMware, Inc.
To keep it simple, actionable and
understandable, we like to use RAG!
Red/Amber/Green is a quick visual
way to demarcate health/severity of
an exception, so it’s highly effective
for quickly triaging response actions.
14
Confidential │ ©2021 VMware, Inc.
Which core metrics inform us
about Program Health?
Don’t flood the zone—we want a few
measurements to tell us a lot
Confidential │ ©2021 VMware, Inc. 15
● Generating about as much
product value as they are able
to, with little resource waste on
non-value generating tasks
like fixing bugs or maintaining
environments and pipelines
● Routinely delivering about the
same increments of value over
a given regular iteration
period, making forecasting
future delivery volume more
precise.
Defining Optimization
We think of a program being optimized when it is highly efficient and highly reliable. In practice
this means that nearly all teams are:
16
Confidential │ ©2021 VMware, Inc.
Efficiency Metrics are used to
determine how much of a team or
a program’s overall available effort
is being expended on value
generating activity, versus other
activities.
Statistical Category: Efficiency
Confidential │ ©2021 VMware, Inc. 17
Mean Percentage of Accepted Stories that were defect corrections
Defect correction does not deliver new
value, so although fixing bugs can be
valuable insofar as it enables
delivery/deployment of defective code
associated with previously delivered value,
the preference is for a team to spend its
capacity on delivering high quality,
defect-free code associated with novel
and new value, that won’t require defect
correction.
Efficiency Metric
Confidential │ ©2021 VMware, Inc. 18
When an engineer or pair delivers a story ready to be accepted or rejected, they
pick up another story and move on.
Acceptance criteria tests should be run as soon as possible, because:
● Newer stories may depend on successful delivery of earlier ones;
● Leaving acceptance to the end of an iteration of work may lead to rejection of
multiple stories, strongly affecting velocity.
This metric may also indicate chronic unresolved bottlenecks or
blockers—removing those can greatly increase throughput.
Mean Time in Delivery State in days
Efficiency Metric
Mean Time in Delivery is a
Cycle Time segment!
19
Confidential │ ©2021 VMware, Inc.
Reliability tells us how well we can
forecast future delivery outcomes
for the program—or a given team.
Teams with regular and disciplined
work habits are more likely to
deliver a predictable volume of
new-value stories.
Statistical Category: Reliability
Confidential │ ©2021 VMware, Inc. 20
Volatility is simply the measure of how
different a team or program’s velocity is,
iteration to iteration. High volatility
indicates that a team or program is
unpredictable and will only allow for
imprecise forecasting.
Low volatility leads directly to high
reliability, by narrowing the band between
best and worst case scenarios when
forecasting future work completion
windows.
Typically, a team with high volatility is
either facing challenges related to story
writing discipline, or is facing delivery
impediments related to organizational
complications not entirely within the
control of the team.
Mean Volatility
Reliability Metric
Confidential │ ©2021 VMware, Inc. 21
Visualizing a Forecast
● A low volatility program has a
relatively narrow band of
output scenarios even
projecting out fairly far into the
future, making forecasting the
business outcomes more
precise and more strategically
useful
22
Confidential │ ©2021 VMware, Inc.
Although not relevant at the team
level, spreads at the program level
help us to determine whether
lower-health indicators are
associated with the program as a
whole, or with outlier teams that
are performing differently to the
rest of the program.
Statistical Category: Spread Measurements
Confidential │ ©2021 VMware, Inc. 23
Tight spreads are a sign that whatever is
going on, it’s something that is being widely
practiced across the program. Coaching
strategies for non-optimal indicators in this
case, should revolve around
mass-intervention at the program level.
Wide spreads, on the other hand, would
suggest that a small number of teams are
outliers from the program. Examining the
team-level metrics will show which teams
deviate from the mean.
A strategy of specific team-level interventions
will likely be the more effective direction to
take. It is often the case that a team has
unique parameters to cope with, so, action
specific to that uniqueness is required to
unblock or re-orient the team.
Spread Measurements
Confidential │ ©2021 VMware, Inc. 24
Put it all together and you have a concise program dashboard
25
Confidential │ ©2021 VMware, Inc.
Practical Application
Making measurement meaningful and useful
Confidential │ ©2021 VMware, Inc. 26
When to start
It’s best to start at the beginning of an
engagement or as soon as possible
You will need:
● Planning tools that can surface the relevant raw data
● The data are obtainable in a sustainable,
easy-to-access manner
Confidential │ ©2021 VMware, Inc. 27
Set achievable goals for a Program
Gather your baseline
Now
Apply a liberal initial
baseline, drive the
Program toward realistic
and achievable
expectations
Next
As teams go more and
more green, we raise the
baselines to determine our
ceiling
Later
Eventually, we reach a
point where raising a
baseline turns nearly all
teams from green to red,
indicating peak
optimization
Confidential │ ©2021 VMware, Inc. 28
Balancing Quantitative with Qualitative
There’s value in learning...
●The local culture
●The operational model
...so that you can create the playbook
of tactics used to manage by
exception and enabling the
community of practice
Confidential │ ©2021 VMware, Inc. 29
Maturity Path
What metrics will do at each step
Now
Enable teams to gather &
understand their metrics
Next
Teams begin to understand
and self correct based on
their metrics
Later
Teams improve themselves
and the program around
them with little outside
enablement
Confidential │ ©2021 VMware, Inc.
Thank You
What questions are there?
1 of 30

More Related Content

What's hot(20)

Agile project managementAgile project management
Agile project management
Bhawani N Prasad897 views
DevOps in the EnterpriseDevOps in the Enterprise
DevOps in the Enterprise
Nauman Noor851 views
What is DevOps? What is DevOps CoE? What is DevOps? What is DevOps CoE?
What is DevOps? What is DevOps CoE?
7Targets AI Sales Assistants12.7K views
Webinar - Devops platform for the evolving enterpriseWebinar - Devops platform for the evolving enterprise
Webinar - Devops platform for the evolving enterprise
DBmaestro - Database DevOps360 views
Continuous delivery best practices and essential toolsContinuous delivery best practices and essential tools
Continuous delivery best practices and essential tools
DBmaestro - Database DevOps403 views
DevOps: What, who, why and how?DevOps: What, who, why and how?
DevOps: What, who, why and how?
Red Gate Software381 views

Similar to Measure Twice, Cut Once: Using Team Operation Metrics to Optimize a Scaling Software Shop(20)

Agile challengesAgile challenges
Agile challenges
David Yancey555 views
rally-business-agility-survival-guide-digitalrally-business-agility-survival-guide-digital
rally-business-agility-survival-guide-digital
Phillip McKenzie Smith681 views
Product Development Journey MarkersProduct Development Journey Markers
Product Development Journey Markers
VMware Tanzu237 views
7 Steps To Developing A Cloud Security Plan7 Steps To Developing A Cloud Security Plan
7 Steps To Developing A Cloud Security Plan
Envision Technology Advisors918 views
Making sense of value - Michel ThiryMaking sense of value - Michel Thiry
Making sense of value - Michel Thiry
PMIUKChapter314 views
Chrysalis Cleariew Introductory V3Chrysalis Cleariew Introductory V3
Chrysalis Cleariew Introductory V3
Robbie Hazlehurst167 views
TDI GPS + Deloitte DevOps JourneyTDI GPS + Deloitte DevOps Journey
TDI GPS + Deloitte DevOps Journey
Fawzy Manaa132 views
Benefits-led decision making drives value maximisation white paperBenefits-led decision making drives value maximisation white paper
Benefits-led decision making drives value maximisation white paper
Association for Project Management 964 views

Recently uploaded(20)

Measure Twice, Cut Once: Using Team Operation Metrics to Optimize a Scaling Software Shop

  • 1. Confidential │ ©2021 VMware, Inc. Krista Luna & Daniel Witenberg Measure Twice, Cut Once: Using Team Operation Metrics to Optimize a Scaling Software Shop VMware Tanzu Labs 2021-09-02
  • 2. Confidential │ ©2021 VMware, Inc. 2 Who Might Find This Valuable? Core audience, but not limited to: Studio Leads Practice Leads/Coaches Business Owners
  • 3. Confidential │ ©2021 VMware, Inc. 3 Program Health is an assessment of the collective health of the set of teams in a lab, studio or larger grouping What is Program Health? By Program, we mean any collection of teams—Portfolio, Lab, Studio
  • 4. 4 Confidential │ ©2021 VMware, Inc. A common challenge we find when joining clients on site, is the client’s desire to scale and mature their Agile software org, without scaling the human capital investment to managing that growth.
  • 5. Confidential │ ©2021 VMware, Inc. 5 Organization are hungry for an approach that allows for a large and growing number of teams to be managed and optimized by a small team of program-oriented practitioners—which grows comparatively slowly.
  • 6. Confidential │ ©2021 VMware, Inc. 6 With healthy quantitative measurement of program optimization, practice leadership at large enterprises can manage increasing scale more effectively than with qualitative tactics alone.
  • 7. Confidential │ ©2021 VMware, Inc. 7 Practice Coaches can manage far more team analytics than they can team-level qualitative intervention, leading to cost savings around Practice Management. Teams’ progress on their Agile maturity path will reflect naturally and organically in their metrics Coaching teams scale slower than application teams Benefits of Quantitative Approach Measuring Team Health Maps to Program Health Agile Maturity can be clearly measured
  • 8. Confidential │ ©2021 VMware, Inc. 8 Benefits of Quantitative Approach Creating a common language A common and credible language in which program health can be described and discussed with stakeholders who don’t come into regular contact with the program teams is crucial. That common language needs to be meaningful to the business, consistent across the organization, and provide objective verifiability.
  • 9. Confidential │ ©2021 VMware, Inc. 9 Benefits of Quantitative Approach Manage by Exception as a Scaling Growth Model Respond Only Where there is need Using this approach, we ‘don’t fix what ain’t broke’ Scale Exponentially, but Sustainably Fewer Practices Coaches needed for more teams A quantitative approach allows us to Manage by Exception, which is to say, focus on only those things that fall outside of our accepted standards.
  • 10. Confidential │ ©2021 VMware, Inc. 10 What this isn’t! Our approach will enhance other tactics, not replace them DORA DORA measures DevOps lifecycle health. Team Health Checks Provides valuable leading and lagging feedback indicators. OKR/KPI/JTBD Our approach doesn’t replace team’s own work to track their optimization and set goals for improvement.
  • 11. 11 Confidential │ ©2021 VMware, Inc. Dashboarding a Program Team performance transparency is the key to sustainable scaled Program optimization
  • 12. Confidential │ ©2021 VMware, Inc. 12 Anyone with an interest in the Program should have a clear line of sight into the metric and understand that it’s credibly connected to the thing it’s supposed to measure. To be useful, program metrics are just like the metrics we teach our PMs about for product dev. They need to be: A change in a metric means a change in how optimized the program is, so any metric we use needs to be something we can act against when things are off-kilter It needs to be easy to understand, track and synthesize meaning from a metric. They shouldn’t create more work for a practice lead to explain, than they are saving in other areas. How Do We Measure? Actionable Accessible Auditable
  • 13. 13 Confidential │ ©2021 VMware, Inc. To keep it simple, actionable and understandable, we like to use RAG! Red/Amber/Green is a quick visual way to demarcate health/severity of an exception, so it’s highly effective for quickly triaging response actions.
  • 14. 14 Confidential │ ©2021 VMware, Inc. Which core metrics inform us about Program Health? Don’t flood the zone—we want a few measurements to tell us a lot
  • 15. Confidential │ ©2021 VMware, Inc. 15 ● Generating about as much product value as they are able to, with little resource waste on non-value generating tasks like fixing bugs or maintaining environments and pipelines ● Routinely delivering about the same increments of value over a given regular iteration period, making forecasting future delivery volume more precise. Defining Optimization We think of a program being optimized when it is highly efficient and highly reliable. In practice this means that nearly all teams are:
  • 16. 16 Confidential │ ©2021 VMware, Inc. Efficiency Metrics are used to determine how much of a team or a program’s overall available effort is being expended on value generating activity, versus other activities. Statistical Category: Efficiency
  • 17. Confidential │ ©2021 VMware, Inc. 17 Mean Percentage of Accepted Stories that were defect corrections Defect correction does not deliver new value, so although fixing bugs can be valuable insofar as it enables delivery/deployment of defective code associated with previously delivered value, the preference is for a team to spend its capacity on delivering high quality, defect-free code associated with novel and new value, that won’t require defect correction. Efficiency Metric
  • 18. Confidential │ ©2021 VMware, Inc. 18 When an engineer or pair delivers a story ready to be accepted or rejected, they pick up another story and move on. Acceptance criteria tests should be run as soon as possible, because: ● Newer stories may depend on successful delivery of earlier ones; ● Leaving acceptance to the end of an iteration of work may lead to rejection of multiple stories, strongly affecting velocity. This metric may also indicate chronic unresolved bottlenecks or blockers—removing those can greatly increase throughput. Mean Time in Delivery State in days Efficiency Metric Mean Time in Delivery is a Cycle Time segment!
  • 19. 19 Confidential │ ©2021 VMware, Inc. Reliability tells us how well we can forecast future delivery outcomes for the program—or a given team. Teams with regular and disciplined work habits are more likely to deliver a predictable volume of new-value stories. Statistical Category: Reliability
  • 20. Confidential │ ©2021 VMware, Inc. 20 Volatility is simply the measure of how different a team or program’s velocity is, iteration to iteration. High volatility indicates that a team or program is unpredictable and will only allow for imprecise forecasting. Low volatility leads directly to high reliability, by narrowing the band between best and worst case scenarios when forecasting future work completion windows. Typically, a team with high volatility is either facing challenges related to story writing discipline, or is facing delivery impediments related to organizational complications not entirely within the control of the team. Mean Volatility Reliability Metric
  • 21. Confidential │ ©2021 VMware, Inc. 21 Visualizing a Forecast ● A low volatility program has a relatively narrow band of output scenarios even projecting out fairly far into the future, making forecasting the business outcomes more precise and more strategically useful
  • 22. 22 Confidential │ ©2021 VMware, Inc. Although not relevant at the team level, spreads at the program level help us to determine whether lower-health indicators are associated with the program as a whole, or with outlier teams that are performing differently to the rest of the program. Statistical Category: Spread Measurements
  • 23. Confidential │ ©2021 VMware, Inc. 23 Tight spreads are a sign that whatever is going on, it’s something that is being widely practiced across the program. Coaching strategies for non-optimal indicators in this case, should revolve around mass-intervention at the program level. Wide spreads, on the other hand, would suggest that a small number of teams are outliers from the program. Examining the team-level metrics will show which teams deviate from the mean. A strategy of specific team-level interventions will likely be the more effective direction to take. It is often the case that a team has unique parameters to cope with, so, action specific to that uniqueness is required to unblock or re-orient the team. Spread Measurements
  • 24. Confidential │ ©2021 VMware, Inc. 24 Put it all together and you have a concise program dashboard
  • 25. 25 Confidential │ ©2021 VMware, Inc. Practical Application Making measurement meaningful and useful
  • 26. Confidential │ ©2021 VMware, Inc. 26 When to start It’s best to start at the beginning of an engagement or as soon as possible You will need: ● Planning tools that can surface the relevant raw data ● The data are obtainable in a sustainable, easy-to-access manner
  • 27. Confidential │ ©2021 VMware, Inc. 27 Set achievable goals for a Program Gather your baseline Now Apply a liberal initial baseline, drive the Program toward realistic and achievable expectations Next As teams go more and more green, we raise the baselines to determine our ceiling Later Eventually, we reach a point where raising a baseline turns nearly all teams from green to red, indicating peak optimization
  • 28. Confidential │ ©2021 VMware, Inc. 28 Balancing Quantitative with Qualitative There’s value in learning... ●The local culture ●The operational model ...so that you can create the playbook of tactics used to manage by exception and enabling the community of practice
  • 29. Confidential │ ©2021 VMware, Inc. 29 Maturity Path What metrics will do at each step Now Enable teams to gather & understand their metrics Next Teams begin to understand and self correct based on their metrics Later Teams improve themselves and the program around them with little outside enablement
  • 30. Confidential │ ©2021 VMware, Inc. Thank You What questions are there?