Accelerate - book notes

previous attempts - the flaws

- focused on outputs, not outcomes
- focused on individual, not team/global outcomes

measuring performance

- delivery lead time
- deployment frequency
- time to restore service
- change fail rate

impacts: delivery perf -> org perf

driving change





how to change it?



impact of CD on quality

CD: what does/does not work

CD adoption

system types

focus on deployability & testability

loose coupling = scalability

allowing devs to choose their own tools

focus on devs & outcomes - not tools or technologies

"shift left" (built-in, not separate phase)

"rugged" devops

lean mgmt practices

implementing a lightweight change mgmt process

lean PD practices

- small-batch workflows
- workflow visibility
- customer feedback
- team experimentation

deployment "pain"

developer burnout

common burnout problems

- work overload
- lack of control
- insufficient rewards
- community breakdown
- lack of fairness
- value conflicts

how to reduce burnout

employee loyalty

- measuring NPS (net promoter score)

changing organizational culture & identity

job satisfaction & org performance

diversity matters

transformational leadership

- vision
- inspirational communication
- intellectual stimulation
- supportive leadership
- personal recognition

the role of managers

tips to improve culture/support

- build trust with your counterparts on other teams
- encourage people to move between departments
- seek/encourage/reward collaborative work
- create a training budget. use it.
- provide informal learning resources
- make it safe to fail
- ID opportunities to share info
- demo days/forums
- making monitoring a priority

primary/secondary research

qualitative/quantitative research

analysis types

- descriptive
- exploratory
- inferential
- predictive
- causal
- mechanistic

classification analysis

trusting data with latent (hidden) constructs (LCs)

LCs --> what are you measuring?

LCs --> multiple data "views"

LCs --> help safeguard against rogue data

LCs --> system data

quick data collection/analysis

measuring full stack with system data is hard

complete measurement with system data is hard

trustworthy data

some things can only be measured via surveys


transforming leadership, mgmt & team practices

continuous delivery

- strict version controls
- automated deployments
- continuous integration
- trunk-based devt
- test automation
- test data mgmt
- "shift left" on security
- continuous delivery


- loose coupling
- empowered team choices

product & process

- customer feedback - gather, implement
- workflow visibility throughout the value stream
- small batches
- foster team experiments

lean mgmt & monitoring

- lightweight change approvals
- monitoring tools
- proactive system health checks
- work-in-process (WIP) limits
- work visualization / communication


- generative (Westrum) characteristics
- learning encouraged
- team collaboration
- correct tools
- transformational leadership (see

organizational performance

SW delivery performance

- deploy frequency
- lead time
- mean time to restore
- change fail%


- unplanned work/rework
- manual work

burnout & deployent pain

technical skill

- trunk-based devt
- continuous delivery
- ...


lean mgmt

lead prod mgmt

org culture

identity, employee net promoter score, job satisfaction



other measures

survey prep & data collection

bias testing

- (chi-square, t-test, harman's single-factor, marker variable

relation testing

- (PCA, AVE, correlation, reliability)

relationship testing

- (correlation - pearson's, PLS & linear regression)

classification testing

(clustering (5 methods), ANOVA)