ericrisco/rsc-harness / skills/kpi-framework
kpi-framework
VerifiedUse when a team must decide what to measure before building anything — picking one north-star metric, separating leading input drivers from lagging outputs, adding guardrails so a number cannot be gamed, and setting a target that is not arbitrary. NOT the live dashboard that displays them (that is `dashboard`), NOT instrumenting the events (that is `analytics`), NOT the recurring board report (that is `reporting`).
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 11.5 KB |
| evals/README.md | markdown | 777 B |
| evals/cases.yaml | config | 3.1 KB |
| references/definition-and-targets.md | markdown | 3.4 KB |
| references/metric-catalog.md | markdown | 4.6 KB |
references/definition-and-targets.md
View on GitHub# Writing definitions, measuring baselines, and calibrating targets
Depth offloaded from `SKILL.md` Steps 1-4. This is where a metric definition stops being a
slogan and becomes a number two analysts compute identically.
## Write an unambiguous definition
A definition that drifts produces arguments instead of decisions. Every metric needs four
components spelled out:
1. **Event** — the exact tracked action. Not "uses the product"; "completes `core_action`
(a saved project with >=1 collaborator)".
2. **Time window** — rolling vs calendar, and its length. "Weekly active" must say *rolling
7-day* or *calendar week*; they differ by ~15% on most products.
3. **Denominator** — what you divide by. "Active teams" vs "all signed-up teams" vs "paying
teams" are three different metrics wearing the same name.
4. **Segment** — who's included. New vs existing, paid vs trial, by plan or region. State it
even when it's "all", so the choice is explicit.
- Bad: `weekly active users`
- Good: `accounts with >=1 member completing core_action in a rolling 7-day window / accounts with an active subscription, all regions`
Rule: prefer a **rate or ratio** over a raw count. Counts grow with time and size and mask
decline; a rate normalizes and exposes health.
## Measure the baseline first
You cannot set a target without a baseline — "get to 50%" is empty until you know you're at
12% or 48%.
- Pull the metric over a stable recent period (typically the last 4-8 weeks, long enough to
smooth noise, short enough to reflect the current product).
- Note **variance**, not just the point value. A metric bouncing 30-46% week to week has a
baseline *range*; a target inside that range is no target at all.
- If the event isn't instrumented yet, the first deliverable is "instrument it and measure
4 weeks" — hand the event to `../analytics/SKILL.md`. Do not invent a target over an
unmeasured metric.
## Calibrate the target magnitude
Targets break in two directions: too ambitious (burnout, shortcuts, gaming) or sandbagged
(no improvement, theatre).
- **Improvement over baseline, not a round number.** Derive the band from (a) the levers you
actually have this period and (b) what comparable improvements have historically yielded.
"38% → 52%" grounded in a named initiative beats "→ 50%" chosen because it's round.
- **Set a realistic band, then commit to a point.** E.g. "expect 48-55% from the onboarding
rework; commit to 52%." The band is your reasoning; the point is the target.
- **Avoid round-number theatre.** 50%, 100%, 1M — round targets are usually chosen for how
they read in a deck, not because the levers support them. Treat a suspiciously round
target as unexamined until proven otherwise.
- **Date + owner are mandatory.** No date → it's a wish. No owner → it's nobody's job.
## Re-validation checklist (run semi-annually)
- [ ] Does the north star still **predict** retention/revenue? Re-check the correlation
against the last two quarters; a metric can quietly stop predicting value.
- [ ] Are the inputs still **controllable** by the owning team, and do they still move the
north star when pushed?
- [ ] Has any guardrail crept toward its harm threshold while the target improved? (silent
gaming signal)
- [ ] Did the product/portfolio change enough to retire or redefine a metric?
- [ ] Version and date the doc; record what changed and why. Transparent evolution keeps a
metric shift from looking like cooking the numbers.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| deprecation-and-migrationaddyosmani/agent-skillsManages deprecation and migration. Use when removing old systems, APIs, or features. Use when migrating users from one implementation to another. Use when migrating a database schema in production, such as renaming or dropping a column without downtime (expand/contract). Use when deciding whether to maintain or sunset existing code. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| performance-optimizationaddyosmani/agent-skillsOptimizes application performance across frontend, backend, queries, and databases. Use when performance requirements exist, when you suspect performance regressions, when Core Web Vitals or load times need improvement, when N+1 query patterns need fixing, or when profiling reveals bottlenecks. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |
| agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |