Pulse 2.0 launch brief
Where we stand. Pulse 2.0 launches on Tuesday, November 17. Today it is a No-Go: readiness is 40 of 100, and two Sev1 blockers are still open, either of which stops the launch on its own. The Go / No-Go call is on November 13 at 10:00 PT. The live score, the task list, and the blocker log are on the Stage, where owners tick off tasks, sign off their workstreams, and log blockers as they happen; every figure here is what it showed when this brief was written.
What is launching
Pulse 2.0 is cut from the release branch v2.0.0, frozen on October 24, so only P0 fixes merge from here. It ships behind feature flags in four cohorts, 1%, 10%, 50%, and 100% of customers, with the mobile app submitted to the app stores alongside it. Around the product, the launch brings a reworked onboarding flow, an updated pricing page, a launch blog post and changelog, press briefings under embargo, and new help center articles. Six workstreams carry it: Engineering, Design, Marketing, Support, Legal, and Ops, each with a lead who signs off.
The readiness criteria
Readiness is one score out of 100, built from three measures and weighed as below. Tasks count by priority: a P0 counts 3, a P1 2, and a P2 1, so finishing the critical work moves the score most. Blocker health starts at 100 and loses 15 points for each Sev1 not yet resolved, 5 for each Sev2, and 1 for each Sev3.
| Measure | Weight | Now |
|---|---|---|
| Tasks done, by priority | 60% | 32% (15 of 47 points; 6 of 20 tasks) |
| Workstreams signed off | 25% | 50% (3 of 6) |
| Blocker health | 15% | 59% (5 open blockers cost 41 points) |
| Readiness | 40 of 100 |
The call follows from the score, with one rule above it:
- Go at 85 or more, with no Sev1 open.
- Caution from 70 up to 85: the leads decide on the call whether the remaining work can land after launch.
- No-Go below 70, or whenever any Sev1 is open, whatever the score.
Marketing, Support, and Ops have signed off. Engineering, Design, and Legal have not, and their open work is where the score is lost: 14 tasks are still open, 5 of them P0.
What holds the launch
Two Sev1 blockers, both in Engineering:
| Blocker | Owner | Status |
|---|---|---|
| Payment webhook retries double-charge in the EU region | Priya | Open: reproduced in staging, idempotency key missing |
| Production snapshot restore fails on a 2 TB volume | Kenji | Mitigating: splitting the restore by tablespace |
The second also blocks a P0 task, the database migration dry run on the production snapshot. Three more blockers are open below Sev1: the editor's keyboard trap for screen readers (Sev2, Design), the iOS build rejected for a missing privacy manifest (Sev2, Engineering), and help center search not indexing new articles (Sev3, Support).
Escalation
- Anyone who finds a launch risk logs it on the Stage as a blocker, with a severity, a workstream, and an owner, the moment they find it.
- A new Sev1 goes to the owning workstream's lead and the on-call rota at once, through the escalation runbook. It makes the launch a No-Go until it is resolved, so it is raised before the call, never at it.
- A Sev2 that will not be resolved before launch day needs its lead's sign-off to launch with it, and a note on the blocker saying why.
- Customer-facing incidents are announced on the status page with the prepared incident templates.
Rollback
The staged rollout is the first line of defence. Each cohort runs before the next is opened; a Sev1 found at any stage stops the rollout where it is, and the flags turn the new release off for everyone in it. If turning the flags off is not enough, the release is rolled back in full. Ops rehearses that rollback before launch, with a target of a full rollback in under 15 minutes; until the rehearsal has passed, it is one of the five open P0 tasks.