LAB QUOTA · OK
[ kpi-define:// ] experimental
cat: ai model: @cf/meta/llama-3.1-8b-instruct

Describe what you want to measure → get 3 KPI candidates with leading / lagging / composite breakdown + how to instrument.

// system prompt
You define KPIs from a measurement goal. User describes what they want to know + context. Output 3 KPI candidates.

For each:

  ### KPI N: <name>

  Type: <Leading / Lagging / Composite>
  Definition: <how it's calculated — exact formula>
  Healthy range: <what "good" looks like for this kind of product>
  Instrumentation: <what events / data sources you need>
  Gameable how: <the failure mode where the metric improves but the underlying thing doesn't>
  Pair with: <a second metric that protects against the gaming>

After the 3 KPIs:

  Pick which one if:
  • You can only have one metric → <which> because <reason>
  • You're trying to spot trouble early → <which leading>
  • You're reporting to the board → <which lagging>

Rules:
- One Leading, one Lagging, one Composite (built from a ratio or rate over time).
- "Gameable how" is mandatory. Every metric is gameable; surface the failure mode.
- "Healthy range" is rough — say "depends on product maturity" if truly unclear, but give a starting bracket.
- Don't invent specific numbers. Give ranges + the reasoning behind them.
⚡ powered by Cloudflare Workers AI · quota deducted on success
// sample output
### KPI 1: Time-to-First-Value (TTFV) for enterprise self-serve

Type: Leading
Definition: Median time from "enterprise signup" to "first meaningful action" (definition of meaningful action depends on product — e.g. first dataset uploaded, first integration connected, first 5 users invited).
Healthy range: < 1 hour is great for self-serve; 1-6 hours is OK; > 24 hours is broken.
Instrumentation: Signup timestamp + per-account event for the chosen "first meaningful action". Both already exist in product analytics for most stacks (Amplitude / Mixpanel / Snowplow).
Gameable how: Lower TTFV by relaxing what counts as "first meaningful action" — e.g. counting "viewed homepage" as activation.
Pair with: Day-7 retention of self-serve enterprise accounts — confirms TTFV improvements stick.

### KPI 2: Self-Serve Enterprise Conversion Rate

Type: Lagging
Definition: % of self-serve enterprise signups that convert to a paid plan within 30 days.
Healthy range: 3-8% is typical for mid-market B2B self-serve; > 10% is exceptional. Heavily depends on signup quality and pricing.
Instrumentation: Signup events + paid-plan-activated events tagged with self-serve channel. Most billing platforms (Stripe / Chargebee) emit the second; you may need to tag the first manually.
Gameable how: Drive conversion by qualifying out top-of-funnel — fewer signups, same conversions, looks like a higher rate.
Pair with: Total paid-enterprise count (absolute number) — guards against shrinking the funnel.

### KPI 3: Activation-to-Stuck Ratio

Type: Composite
Definition: For each cohort, (% of accounts that complete activation by day 7) / (% of accounts that complete signup but show no activity by day 7). Higher is better.
Healthy range: > 3 is healthy. < 1.5 means more accounts are getting stuck than activating.
Instrumentation: Cohort signup events + activation event + a "stuck" event (heuristic: no events for 7 days post-signup).
Gameable how: Inflate the numerator by counting trivial actions as activation; deflate the denominator by quietly excluding "low-quality" cohorts.
Pair with: Raw counts of both segments — surfacing the absolute number prevents the ratio from masking a shrinking population.

Pick which one if:
• You can only have one metric → KPI 3 (Activation-to-Stuck Ratio) because it captures both the success path and the failure path in a single number.
• You're trying to spot trouble early → KPI 1 (TTFV) — moves before conversion does.
• You're reporting to the board → KPI 2 (Conversion Rate) — the lagging, "did it work" number that executives recognise immediately.
// powered by cloudflare workers ai · quota deducted on success ← back to catalog