Quarterly update · engineering

Two of four are landing, one is late, and one we should stop

Nine weeks in, three to go. The honest read is below, including the objective I would cancel today rather than carry into Q4.

Period Q3 2026, week 9 of 12 For the board, 4 September Headcount 14, unchanged

If you read one paragraph

Reliability and the billing rewrite are on track and will land. Self-serve onboarding is two weeks late for a reason we understand and have fixed. The partner API has consumed nine weeks and one engineer and has no committed customer — I am recommending we stop it and move that engineer onto onboarding.

The numbers

Uptime
99.94%
▲ 0.11 · goal 99.9%
p95 latency
210ms
▼ 88ms · goal 250ms
Activation
31%
▼ 2 pts · goal 40%
Ship rate
18/wk
▲ 3 · no goal set

Scope, planned against delivered

Solid is what shipped. Dashed is where the plan said we would be.

Q3 burn-up, in shipped scope Delivered Planned
48 36 24 12 0 9 behind today wk 1 wk 4 wk 7 wk 10 wk 12

Objective by objective

ObjectiveStatusProgressThe read
Reliability 99.9% uptime, p95 under 250ms On track
92%
Both targets already met and held for six weeks. The remaining work is the runbook, not the system.
Billing rewrite Replace the invoice engine On track
78%
Migration dry run passed on a copy of production last Thursday. Cutover is scheduled for week 11 with a rollback that we have tested.
Self-serve onboarding 40% activation Two weeks late
54%
We built the flow before we had the funnel data, then rebuilt it once the data arrived. Instrumentation now lands before design on every step. Activation will not reach 40% this quarter; 35% is realistic.
Partner API Three integrations live Recommend stopping
22%
Nine weeks, one engineer, no partner has committed to build against it. The design is sound and will keep. This is the recommendation the meeting should be about.

Why I am recommending we stop the partner API rather than slow it down

A slowed project still holds an engineer, still needs review, and still shows up as red every month. The design work is done and written down, so restarting it costs a week rather than nine. The reason to stop is not that it is going badly — it is that we have no evidence anyone is waiting for it, and onboarding has evidence stacked against it.

What I need from this meeting