Skip to content

Quarterly KPI review

Per reference/kpis.md, 961tech tracks a north-star (Weekly Qualified Outbound Clicks), secondary metrics (build-completion rate, search-result-clickthrough rate, retailer-coverage breadth), and health metrics (scraper success rate, P75 LCP, error rate).

The quarterly review is when we look at the trend, decide what to change, and write it down.

Cadence

First Monday of January, April, July, October. Solo for now; expand when there's a co-maintainer.

Agenda (timeboxed, 60 min)

  1. Trend snapshot (10 min) — pull the per-day rollup from MetricRollup for each KPI. Plot QoQ delta. No commentary yet.

  2. North-star walk (15 min) — Weekly Qualified Outbound Clicks. What drove the QoQ delta? If positive: was it organic growth, a new retailer, a category bump, or a ranking-algo change? If negative: which segment fell?

  3. Secondary metric audit (10 min) — build-completion rate, search-CTR, retailer-coverage breadth. Anything moving > 20% needs a hypothesis.

  4. Health metric audit (10 min) — scraper success rate per retailer, P75 LCP per route, error rate. Anything red flips to a fix-this-quarter ticket.

  5. Decisions (15 min) — what changes for next quarter. Three formats:

  6. Re-shape a KPI — e.g. WQOC → unique-clicks-per-builder if click-spam emerges. Land as an ADR.
  7. File implementation tickets — every "we should do X" becomes a GitHub issue.
  8. Defer — if the metric is moving in the right direction and we can't articulate why we'd intervene, don't.

Source queries

Stored in docs/runbooks/observability-queries.md (filed when #43 implementation lands per ADR-0015). Each KPI has a named SQL block.

Decision log

Append to docs/plans/YYYY-MM-DD-quarterly-review.md per session. Each decision: title, options considered, chosen, rationale, follow-up issue (if any). One-page max — long discussions become ADRs, not notes.

When to skip

  • The first month of a quarter where data volume is tiny (< 50 builds/week, < 100 outbound clicks/week). Skip; pick up next quarter.
  • A quarter where the only meaningful signal is "we shipped feature X and now traffic increased" — that's a release retrospective, not a KPI review. Use the postmortem template instead.

See also