Engineering Performance · PI 10 brief
PI 10 · Cloud
PI 10 ran 28 April to 5 August 2026. This brief summarises Cloud's PI 10 results across the five active sprints. The IP sprint (S6) is excluded from headline metrics per the standard convention.
Cloud closed PI 10 with every sprint goal hit: 16 of 16 across the five active sprints, up from 13 of 15 in PI 9. Velocity vs Capacity recovered to 96% from 84% in PI 9. Commitment came down to 117% of capacity, the most realistic plan the team has set in the trend window, and merged MR throughput finished 26% above the PI 8 baseline.
Completion Ratio
50%
▲ +8pp vs PI 9 (41.8%)
Committed Done vs Capacity
59%
▼ -2pp vs PI 9 (61.0%)
Velocity vs Capacity
96%
▲ +11pp vs PI 9 (84.4%)
Sprint Goals Hit
100%
16 of 16 goals
Creep Ratio
42%
▼ +11pp vs PI 9 (31.0%)
Bugs Completed
21
▼ +4 vs PI 9 (17)
Merged MRs
312
▲ +26% vs PI 8 · 4.4/day
Capacity, Commitment, Velocity, Creep · PI-level
Story points per PI · velocity = commit_done + creep_done · creep shown as absolute SP · active sprints only
Capacity Commitment Velocity Creep
Capacity, Commitment, Velocity, Creep · PI 10 by sprint
All six sprints actual. S6 is the IP sprint and is excluded from the headline metrics above.
Capacity Commitment Velocity Creep
Creep Ratio · PI-level
creep / commitment · 20% ceiling shown
Creep ratio 20% ceiling
Bugs Completed · PI-level
Count per PI
Bugs
Merged MRs and commits · PI-level activity
Accepted merge requests and commits per PI, attributed by team member, active sprints only. PI 9 was a broad-based activity peak across the whole organisation.
Merged MRs Commits (right axis)
Platform releases deployed · PI-level
cloud-platform-environment production deployments per PI (the actual release event) · active window only
Platform deployments
Deployments per PI · by tier
Active-period count (excludes IP sprint)
Production Staging
Deployment frequency · by tier
Deploys per active day · 0.5/d is the DORA target
Production Staging 0.5/d target
Releases this PI
All Cloud project releases in PI 10 through 7 July 2026, the end of S5. Platform releases are 3.10.15.0 (deployed 7 May) and 3.11.0 (deployed 30 June), and the release cadence metric is scoped to cloud-platform-environment only. The remaining rows are CI-driven per-commit tags on rest-api, beam-processors and flink-maps-stream-inspector, which is why the raw tag count is high. cloud-platform-environment rows use production deployment dates (the actual release event); other projects use GitLab tag dates (no deployment tracking). 3.10.14.0 was deployed 7 Apr (PI 9) despite its May tag date, so it appears in the PI 9 brief. 3.10.16.0 was tagged 3 June but never deployed to production and was superseded by 3.11.0.
| Date | Project | Version |
|---|
| 29 Apr | rest-api | 1.205.10000 |
| 6 May | rest-api | 1.206.10000 |
| 7 May | beam-processors | 1.28.0 |
| 7 May | cloud-platform-environment | 3.10.15.0 |
| 7 May | rest-api | 1.207.10000 |
| 8 May | rest-api | 1.208.10000 |
| 12 May | beam-processors | 1.28.1 |
| 14 May | rest-api | 1.209.10000 |
| 22 May | beam-processors | 1.29.0 |
| 22 May | rest-api | 1.210.10000 |
| 22 May | rest-api | 1.211.10000 |
| 22 May | rest-api | 1.211.10001 |
| 22 May | rest-api | 1.211.10002 |
| 22 May | rest-api | 1.211.10003 |
| 22 May | rest-api | 1.211.10004 |
| 22 May | rest-api | 1.211.10005 |
| 25 May | rest-api | 1.211.10006 |
| 25 May | rest-api | 1.211.10007 |
| 25 May | rest-api | 1.211.10008 |
| 25 May | rest-api | 1.211.10009 |
| 25 May | rest-api | 1.211.10010 |
| 26 May | rest-api | 1.211.10011 |
| 26 May | rest-api | 1.211.10012 |
| 26 May | rest-api | 1.212.10000 |
| 27 May | beam-processors | 1.30.0 |
| 27 May | beam-processors | 1.30.1 |
| 28 May | beam-processors | 1.30.2 |
| 28 May | beam-processors | 1.31.0 |
| 2 Jun | beam-processors | 1.31.1 |
| 8 Jun | beam-processors | 1.32.0 |
| 9 Jun | flink-maps-stream-inspector | 7.7.0 |
| 9 Jun | flink-maps-stream-inspector | 7.7.1 |
| 12 Jun | beam-processors | 1.32.1 |
| 12 Jun | beam-processors | 1.32.2 |
| 12 Jun | rest-api | 1.213.10000 |
| 22 Jun | beam-processors | 1.33.0 |
| 23 Jun | beam-processors | 1.34.0 |
| 24 Jun | beam-processors | 1.34.1 |
| 24 Jun | rest-api | 1.214.10000 |
| 30 Jun | cloud-platform-environment | 3.11.0 |
| 1 Jul | beam-processors | 1.34.2 |
| 2 Jul | beam-processors | 1.35.0 |
| 2 Jul | flink-maps-stream-inspector | 7.8.0 |
Goal progress
Final PI 10 position against the 2026 team-level engineering goals applicable to Cloud.
Sprint & PI goal achievement
100% · 16 of 16 goals (S1-S5)
Target: ≥ 80% (shared, all teams)
Pass
Release cadence
2 platform releases · 5 active sprints · 0.4/sprint
Target: ~1 release per active sprint (Nick) · measured by production deployment date
Miss
What stood out
Every sprint goal hit across all five active sprints (16 of 16). PI 9 was 13 of 15. PI 9 was 13 of 15, so this is a clear improvement on an already strong base.
Velocity vs Capacity finished at 96%, up from 84% in PI 9. Total resolved work (157.3 SP) came close to measured capacity (164.1 SP) for the first time since PI 8. Completion Ratio improved to 50% from 42%.
Commitment discipline improved markedly. Cloud committed 117% of capacity in PI 10 versus 146% in PI 9 and 193% in PI 8. The plan is now much closer to what the team can absorb.
Merged MRs finished at 312 against 248 in PI 8, up 26%. PI 9's 666 was an organisation-wide activity peak rather than a Cloud-specific one, so the fall from that figure is a return toward baseline rather than a decline. Review depth held: 437 approvals against 312 merges, roughly 1.4 per merge.
Where the data is mixed
Creep Ratio finished at 42%, up from 31% in PI 9 and the highest in the trend window. 81.3 SP of unplanned work landed against 191.7 SP of commitment. Commitment came down and unplanned work rose over the same period. Whether the two are connected is not something these figures can settle, but the net effect is that better commitment discipline did not produce a more predictable PI.
Platform release cadence missed the goal: 2 production deployments (3.10.15.0 on 7 May, 3.11.0 on 30 June) against a target of roughly one per active sprint. PI 9 delivered 5. A 54-day gap separated the two PI 10 releases.
3.10.16.0 was tagged in GitLab on 3 June but never reached production; it was superseded by 3.11.0 on 30 June. Worth a look at what stalled it.
Bug count rose to 21 from 17 in PI 9, continuing a three-PI climb from 10 in PI 8.