# Lean Analytics > Startup metrics and the One Metric That Matters — an AI agent skill based on *Lean Analytics* by Alistair Croll & Benjamin Yoskovitz. Published 2026-06-11 · Updated 2026-08-17 · Product & Innovation · Free and MIT-licensed Canonical: https://skills.wondel.ai/skills/lean-analytics/ Source: https://github.com/wondelai/skills/tree/main/lean-analytics **Install:** `npx skills add wondelai/skills/lean-analytics --global` ## What is the Lean Analytics skill? Lean Analytics gives your agent the measurement half of the lean toolkit, from Croll and Yoskovitz’s book: choose the One Metric That Matters for your business model and stage, separate numbers that change a decision from ones that only look good in a board deck, and draw a line in the sand with a target, a date and a consequence. ## Key concepts from Lean Analytics your agent applies - **Good vs Vanity Metrics** — A good metric is comparative, understandable, a ratio or rate, and changes behavior. Cumulative up-and-to-the-right charts are the classic vanity tell. - **One Metric That Matters** — Track the single number that tells you whether the riskiest part of the business is working — paired with a counter-metric so it can't be gamed. - **Metrics by Business Model** — Six archetypes — e-commerce, SaaS, mobile app, media, UGC, marketplace — each with its own metric tree and definition of "working." - **The Five Stages** — Empathy, Stickiness, Virality, Revenue, Scale — each has a gate, and working on a later stage's metric too early is the canonical startup mistake. - **Lines in the Sand** — A metric without a target is trivia: set a number, a date, and a pre-committed response to missing it. ## When to use Lean Analytics — and when not to Reach for it when: - Your dashboard carries forty numbers and the weekly meeting is spent explaining which of them went up. - The headline metric is cumulative total signups, which by construction has never once gone down. - You are optimizing revenue while a third of new accounts never reach the core action a second time. - Someone asks whether four percent is a good conversion rate and nobody can say what good would look like here. Reach for something else when: - You need the experiment designed around the metric rather than the metric chosen — Lean Startup runs Build-Measure-Learn. - A checkout funnel is leaking and you want it diagnosed step by step — CRO Methodology is the sharper instrument. - Retention is the number and what you actually need is a behavioral intervention — Improve Retention designs that. - You are setting a price and need willingness-to-pay evidence, not a revenue metric — Monetizing Innovation covers it. ## About Alistair Croll & Benjamin Yoskovitz Serial founders, authors of the Lean Series data handbook. Alistair Croll is an entrepreneur and analyst who co-founded web performance company Coradiant and chairs Startupfest. Benjamin Yoskovitz is a founding partner at venture studio Highline Beta and a serial founder and startup investor. They wrote Lean Analytics for Eric Ries's Lean Series. The book: https://openlibrary.org/books/OL26184439M/Lean_Analytics ## Example prompts - Pick the One Metric That Matters for our SaaS at our current stage using lean-analytics skill _(Metrics strategy)_ - Audit our dashboard for vanity metrics and rewrite them as ratios using lean-analytics skill _(Dashboard audit)_ - Build the metric tree for our marketplace with liquidity measures using lean-analytics skill _(Instrumentation)_ - Set a line in the sand for churn with a pre-committed miss response using lean-analytics skill _(Target setting)_ ## Frequently asked questions ### How do I install the Lean Analytics skill? Run `npx skills add wondelai/skills/lean-analytics --global`. It takes about 30 seconds and needs no account. The Lean Analytics skill then works in Claude, Claude Code, Claude Cowork, Codex, Cursor, OpenClaw and Hermes Agent — anything that reads the open agentskills.io format — and your agent loads it on its own when a task calls for it. It is free and MIT-licensed, and the source is at https://github.com/wondelai/skills. ### Which book is the Lean Analytics skill based on? It packages *Lean Analytics* by Alistair Croll & Benjamin Yoskovitz — serial founders, authors of the Lean Series data handbook. The skill distils the book's method into instructions your agent follows while it works, covering good vs vanity metrics, one metric that matters and metrics by business model. It sits in the Product & Innovation part of the library. ### Does One Metric That Matters mean ignoring everything else? No. It means one metric gets the team’s attention at a time, chosen because it measures the riskiest unproven part of the business right now — and it is expected to change as you move stages. Croll and Yoskovitz pair it with a counter-metric so it cannot be gamed: chase signups alone and you will buy junk traffic, pair signups with activation and you cannot. Everything else stays instrumented. It simply does not get to run the weekly meeting. ### How is this different from the Lean Startup skill? Ries gives you the loop; Croll and Yoskovitz give you the instrument panel. The Lean Startup tells you to measure what proves your leap-of-faith assumption and to prefer cohort behavior over cumulative totals — but not which numbers matter for a marketplace as opposed to a SaaS or a media site. Lean Analytics does: six business-model archetypes with their own metric trees, five stages with gates between them, and the classic mistake of optimizing a later stage’s metric before you have passed the earlier gate. ### The benchmarks are from the early 2010s — is the book still useful? Yes, with a caveat the skill will repeat back to you. Treat the specific numbers as shapes rather than targets; the structure has aged far better than the data — the four tests of a good metric, the stage gates, the per-model metric trees. Where you need genuine comparators, your own cohorts beat a decade-old industry median every time, and the skill will decline to hand you a published benchmark as though it were current. ### Will it pull my actual numbers? Only what you bring it, or what your agent can reach with the tools you have connected. The skill has no privileged access to your warehouse or product analytics. Where it earns its place is the argument before the query: which metric, which segment, which denominator, over what window, and what result would change the decision. A metrics framework applied to numbers nobody has instrumented yet is still just a plan — a much better plan, but a plan. ## Related - All Product & Innovation skills: https://skills.wondel.ai/skills/product/ - Jobs to Be Done: https://skills.wondel.ai/skills/jobs-to-be-done/ - Hooked UX: https://skills.wondel.ai/skills/hooked-ux/ - Improve Retention: https://skills.wondel.ai/skills/improve-retention/ - Lean Startup: https://skills.wondel.ai/skills/lean-startup/ - Design Sprint: https://skills.wondel.ai/skills/design-sprint/ - Continuous Discovery: https://skills.wondel.ai/skills/continuous-discovery/ - Inspired: https://skills.wondel.ai/skills/inspired-product/ - The Mom Test: https://skills.wondel.ai/skills/mom-test/ - The 37signals Way: https://skills.wondel.ai/skills/37signals-way/