Every formula from the book, gathered in one place. Each entry names the
topic with the full discussion, including its gaming risk and guardrail.
Use this as a quick lookup, not a substitute for the topic itself.
Flow metrics (Part 2)
Metric
Formula
Topic
Flow velocity
Count of flow items completed per unit time
2.3
Flow distribution
(Completed items of one flow item type) / (Total completed items) x 100%
2.3
Flow time
Time from a flow item entering the value stream to its delivery
2.4
Flow load
Count of flow items currently active or waiting in the value stream
2.4
Little’s law
Flow load (work in process) = Arrival rate x Flow time (cycle time)
2.4, 2.7
Flow efficiency
Active work time / Total elapsed time x 100%
2.5
Cycle time
Sum of stage durations: coding + pickup + review + test + deploy
2.6
Utilization
Arrival rate / Service rate
2.7
Percent complete and accurate (%C/A)
(Units usable downstream without rework) / (Total units) x 100%
2.8
Rolled throughput yield
%C/A of stage 1 x %C/A of stage 2 x … x %C/A of stage N
2.8
Takt time
Available working time / Customer demand over that period
2.8
Time to first review
Time from pull request opened to first substantive reviewer response
2.9
Deployment frequency
Count of successful production deployments per unit time
2.10
Lead time for changes
Time from first commit to successful production deployment (report median and 90th percentile)
2.10
Change failure rate
(Deployments causing a failure) / (Total deployments) x 100%
2.10
Failed deployment recovery time
Time from failure detection to genuine service restoration
2.10
Developer experience (Part 3)
Metric
Formula
Topic
Focus time
Count and duration of uninterrupted two-hour-plus blocks per week, from calendar data
3.6
Response rate
(Survey responses received) / (Survey invitations sent) x 100%
3.7
Code and quality (Part 4)
Metric
Formula
Topic
Cyclomatic complexity
Independent paths through control flow (edges − nodes + 2, per McCabe)
4.1
Test coverage
(Lines/branches executed by tests) / (Total lines/branches) x 100%
4.2
Mutation kill rate
(Mutants killed by test suite) / (Total mutants introduced) x 100%
4.2
Code churn
Lines added + modified + deleted per file over a time window
4.3
Hotspot score
Churn x Complexity, ranked per file
4.3
Debt carrying cost
Estimated ongoing cost of not fixing an item (slower related work, elevated defect risk)
4.5
Product and business (Part 5)
Metric
Formula
Topic
Escaped defect rate
(Severity-weighted escaped defects) / (Unit of delivery or time)
5.1
Initial adoption
(Users who tried the feature at least once) / (Target audience) x 100%
5.2
Retained adoption
(Users still using the feature after N weeks) / (Users who initially tried it) x 100%
5.2
Unit cost
Total cost (people + infrastructure + tooling) / Meaningful unit (customer, transaction)
5.4
ROI
(Total benefit − Total cost of ownership) / Total cost of ownership, presented as a range
5.5
Reliability, operations, and security (Part 6)
Metric
Formula
Topic
Error budget
(1 − SLO target) x Time window (e.g., 0.1% of 30 days ≈ 43 minutes)
6.1
Error budget burn rate
Error budget consumed / Error budget allotted, over a given window
6.1
MTTD
Time from incident onset to detection
6.2
MTTA
Time from incident notification to acknowledgement
6.2
MTTR (incident)
Time from acknowledgement to genuine service restoration
6.2
On-call page distribution
Pages received per individual, over a rolling window (not team average)
6.3
Vulnerability time-to-remediate
Time from discovery to genuine remediation, tracked by severity
6.4
Notes on using these formulas
Always pair a speed or output formula with its guardrail (topic 1.2): change failure rate with deployment frequency and lead time; escaped
defect rate with delivery speed; error budget burn with deployment
activity.
Use medians and percentiles, not averages, for time-based formulas (topic 1.6) unless a formula explicitly calls for a mean.
Every formula needs a documented source system and collection method (topic 1.5) alongside its mathematical definition; two teams computing
the same formula from different sources will not produce comparable
numbers.
Severity weighting is not shown explicitly in every formula above but
applies wherever “severity-weighted” appears; see the relevant topic for
the full classification scheme.
