KPI Advisor
The KPI interpretation layer above analytics (guidance-only): a non-technical founder's fractional-CFO conversation. Takes raw metrics — typed OR pulled live from GA4 + PostHog via their read-only MCP servers OR from the sales-pipeline/retention peers — and returns a plain-english read, a persona-lensed investor view (growth-VC / efficiency / bootstrapper / regulated-B2B-distributor), benchmark deltas with cited sources, impact-ranked next moves, and a 5-part board talk-track — anti-sycophantically (no un-anchored soft hedges). Branches on business model FIRST; a compliance-industry branch (REACH/CLP/biocide distributor retention math) reads distributor numbers SaaS benchmarks mis-flag. No callable API — the agent's own LLM runs the decision tree in SKILL.md. Backed by 55 evidence-bundled registry items + 14 learnings + 5 enforcement scripts. Multilingual EN floor (DE/RU/FR/RO/HE/ES planned).
Two founders showed me the same 26-month CAC payback. I told one it was fine and the other that it was why they would not raise. Both reads were right, because a number does not carry a verdict on its own — it carries a verdict for someone. Benchmarkit tracked the median CAC payback drifting from 14 to 18 months across 2023 and 2024; a 2021 growth partner called 26 months "playing long" and a 2024 efficiency fund calls it un-fundable. So kpi-advisor asks which investor lens you want before it interprets anything: growth VC (stares at NRR, walks below 100% — segment medians run 118% enterprise / 108% mid-market / 97% SMB across the 939 companies Optifai measured), efficiency fund, bootstrapper, or regulated B2B distributor. That last lens is the one no other skill ships. Wholesale runs roughly 44% logo churn as normal and about 80% of value creation comes from existing accounts, so SaaS thresholds mis-flag a healthy chemicals distributor as CRITICAL — this branch measures reorder cadence, account-level NRR and registration coverage instead. I made anti-sycophancy the product, not a tone setting: every verdict cites a dated benchmark, fixes are capped at three, and a missing metric produces a question rather than a guess. Guidance-only — no backend, no provisioning. 55 evidence-bundled registry items, 14 learnings, 5 enforcement scripts.
Features5
Growth VC, efficiency fund, bootstrapper or regulated distributor. It asks which one before it reads a thing.
No un-anchored hedges. lint-sycophancy.mjs fails the output when one slips through.
A missing metric produces a question, not a plausible-looking estimate you would act on.
Wholesale runs ~44% logo churn as normal, so SaaS thresholds mis-flag a healthy distributor.
Impact x effort x time-to-signal. A list of eleven priorities is a list of zero priorities.
Videos1 video
DemoKPI Advisor · Demo video
Demo · SpiderMediaArticles about KPI Advisor2 articles
Stop asking if your numbers are good. Ask who's reading them.
opvs.ai22 Jul 2026Stop asking if your numbers are good. Ask who's reading them.
Two founders showed me the same 26-month CAC payback. I told one it was fine and the other it was why they would not raise. Both reads were correct. Here is why a number carries a verdict only for someone, and the KPI skill I built to stop AI telling founders their numbers look solid.
Videos1 video
DemoKPI Advisor · Demo video
Demo · SpiderMediaArticles about KPI Advisor2 articles
Showing all 2
SKILL.mdthe file your agent reads
SKILL.md loads when you open this tab.
What's new4 releases · newest first · from the package's CHANGELOG
v0.1.32026-11-10Latest
Marketplace listing update: added the launch explainer film (`media.demo_video_url`). No code, schema, copy or guidance changes; hero and screenshots unchanged from 0.1.2.
v0.1.22026-08-23
Packaging fix — the Layer-7 coordination gate now actually runs from the shipped bundle. - **`scripts/verify-coordination.mjs` crashed on import in every prior release.** It imported `'../../../scripts/_verify-coordination-shared.mjs'` — a path that resolves outside the bundle — while the helper was never staged beside it. The gate died with `ERR_MODULE_NOT_FOUND` on every run and therefore verified nothing, silently, at 100/100 HEALTHY. - **The wrapper is now self-resolving** and the shared helper is vendored into the bundle. No behaviour change to the skill itself — this release only makes its own quality gate functional.
v0.1.12026-07-22
### Fixed - **Customer-visible.** `tier: beta` + `skills[0].visibility: public` + `type: skill` — 0.1.0 shipped with tier and visibility both `internal`, so the package was live, healthy and invisible on the marketplace for 16 days. No change to skill logic. ### Added - Marketplace card asset kit (SPEC-3): top-level `media.hero_image_url` + 3 `media.screenshots` (four investor lenses · the distributor two-scale comparison · the interpretation pipeline). - `marketing.features[]` — 5 verb-led bullets with icons (the card previously shipped an empty feature list). - First-person `marketing.long_description` rewritten against the locked Di-Atomic voice profile. - Propagated to di-atomic.com: launch blog, changelog entry, `/skills-kpi-advisor` listing page, `/skills/kpi-advisor` collection record, and a card on the `/skills` index.
v0.1.02026-07-06
### Added
- Initial release. Guidance-only KPI interpretation skill (Cluster J — the interpretation layer above analytics).
- **Layer 1** — `SKILL.md` router: business-model branch → mode → plain-english → 1-of-4 investor lens →
benchmark deltas → red flags → ≤3 ranked moves → board talk-track. 6 hard rules incl. anti-sycophancy.
- **Layer 2** — 10 references: metric vocabulary + formulas, 2026-refreshed benchmarks, 4 investor lenses,
red-flag catalog, ranking heuristic, board-brief format, anti-sycophancy playbook, compliance-industry
metrics (the distributor moat), GA4+PostHog data-source ingestion, multilingual vocabulary (v0.2.0 stub).
- **Layer 4** — 55 evidence-bundled registry examples (every item a live source URL + measurable metric).
- **Layer 3** — 14 institutional learnings.
- **Layer 5** — 9 recipes (monthly review · quarterly board brief · fundraising readiness · runway emergency ·
first-90-days · churn diagnosis · compliance-industry review · connect-analytics-sources · multilingual stub).
- **Layer 6** — instructions/AGENTS.md + DEPLOYMENT.md.
- **Layer 7** — 5 enforcement scripts (lint-sycophancy · verify-metric-coverage · lint-benchmark-citations ·
verify-learnings-ranking · verify-coordination).
- **GA4 + PostHog** data-source ingestion via their read-only MCP servers (MCP-composition — guidance-only
preserved; external, optional).
- v2.2 coordination cross-refs (agentboard / agentdocs / agentmemory / opvs-protocol).
- Adapted from 3 MIT-licensed upstream collections (anti-sycophancy + forcing questions; metric formulas,
benchmark tables, board-package structure and GA4/PostHog data mapping; the persona-posture mechanic).
Full attribution, commits and licences in `source/README.md`.
### Notes
- **Bundle shape:** NESTED `skills/kpi-advisor/{SKILL.md,schema.yaml,…}` + `skills:[{id,schema}]`
(2026-06-22 corrected shape; media-generator 1.3.2 precedent). Gated by `verify-packaging.mjs`.
- Guidance-only: methods-less `schema.yaml`; no backend.Showing all 4
