measuring the agile process improvement
TRANSCRIPT
11th Central and Eastern European Software Engineering Conference in Russia - CEE-SECR 2015
October 22 - 24, Moscow
Konstantin Savenkov
Measuring the agile process improvement
ContextThere’s no universal solution, everything matters in its
contextBookmate is a B2C content service:
lots of stakeholderslots of external partieslots of urgent stufflots of experimentsrapid team growth
A way different from: MVP, B2B, outsourcing
The goalsSubscribers
New Recurring Reactivated
Traffic Conversion Retention Reactivation
Engi-neer-ing
Marketing Product Sales Business Development Content & Editorial
Stakeholders and engineeringProduct
Marketing
BD & Sales
Content
Engineering
iOS Android Web
Backend
OpsContent Ingestion
Bookmate Publisher
Windows2-weekcycle
How things ideally would happen
ROADMAPS DESIGN DEVELOPMENT QA SHIPPED
PLANNING
How things actually happenINFLATING EFFORT
ROADMAPS DESIGN DEVELOPMENT QA SHIPPED
BACKLOG
NEW STUFF
TECH. DEBT
PLANNING
BUGS
UX
SOFTWARE OPS
URGENTSTUFF
UNCLEAR DESIGN
UNCLEAR TECH
ITERATING
Continuous improvementOverall productivity KPI is clear: shipped/committed ratioWithout actionable metrics, improvement involves too
much emotion and causal attributionThe metrics should be:
end-to-end (any change in KPI reflected by metrics)invariant to story size and story point valueallow for measuring external factors (stakeholders)
What’s out there? stories committed / stories completed technical debt management team velocity story cycle time CAT/QA cycle time estimation accuracy defects per release cycle defects per story point test cases run
DEPEND ON STORY SIZEDEPEND ON STORY POINTA LOT OF GAPS
What do we use
PRODUCTIVITY
Estimated effort on
planned and doneAll available effort
ACCURACY
Estimated effort on
planned and done
Actual effort on
planned and done SHIPABILITY
Actual effort on
planned and doneEffort spent
on plannedPLANABILITY
Effort spent on planned
Actually spent effort
UTILISATION
Actually spent effort
All available effort
Issues we discover
ACCURACY
Estimated effort on
planned and done
Actual effort on
planned and done SHIPABILITY
Actual effort on
planned and doneEffort spent
on plannedPLANABILITY
Effort spent on planned
Actually spent effort
UTILISATION
Actually spent effort
All available effort
Unexpected vacations, OOO etc
Feasibility of all sort of buffers (management,
hot bugs etc)
Insufficient lookahead
Improper communication
Unclear design
Unclear tech Software bugs
Team cohesionUnclear tech
Design bugs
What’s good about thatAlmost free (given you estimate effort)Don’t depend on story size (measure effort, not stories)Don’t depend on the story point size (it cancels out)Any issue from the messy picture above is reflected by
those metricsEasily projected back to stakeholders to find external bottlenecks, e.g. BUSINESS DEVELOPMENT
Effort spent on planned
(BD)Actually spent effort (BD)
Track stuff that is easily swept under the rugs:
Next batch
QUALITY
Estimated effort onall bugs
All available effort
TECH DEBT
Estimated effort on
tech. backlogAll available
effort
Be sane about thatIt doesn’t remove bottlenecks and issuesIt just shows themSometimes it’s enough to solve the issueMost of the times it isn’t