The metric a team is graded on is the only strategy document that gets read every week. This skill is about choosing that number well — how to weight it, where to set its bar, what counter-metric sits beside it, and what caveats must travel with it — so that moving the number means users actually got value.
Scope note: this is the dashboard your own team argues over. Scores you publish to users or outsiders — ratings, certifications, proficiency levels, trust scores — have to survive gaming by the people being scored and belong elsewhere.
This skill is a graph. Scan the descriptions, follow only the [[wikilinks]] you need.
Shaping the number itself
- [[references/put-the-weight-in-the-formula-not-the-memo]] — when some actions are worth more than others, encode the ratio as a coefficient; a stated priority changes nothing.
- [[references/threshold-percentage-not-aggregate-total]] — grade on the share of users clearing a meaningful bar, not the sum, so a power-user cohort cannot absorb the whole gain.
- [[references/pair-every-growth-metric-with-a-quality-proxy]] — a growth metric alone drifts toward whatever is cheapest to repeat; the counterweight has to be readable daily.
Deciding what is measurable at all
- [[references/refuse-to-measure-the-vague-goal]] — decline to score the unfalsifiable word in your positioning; decompose it into named sub-competencies and publish the weak one too.
Getting numbers to the people who can act on them
- [[references/leading-indicators-for-the-people-doing-the-work]] — a metric that arrives after the work is finished cannot change the work; move it into the authoring surface.
- [[references/per-unit-quality-map-not-aggregate-score]] — an overall score says something is wrong; a color-coded per-unit map says which part to open.
- [[references/score-the-feedback-queue-do-not-drain-it]] — past the volume you can review, rank by predicted validity and mine the rejected majority for comprehension failures.
Reading order
Start with [[references/pair-every-growth-metric-with-a-quality-proxy]] if you are choosing a north star from scratch — it explains the failure mode that every other node here is a response to. Start with [[references/score-the-feedback-queue-do-not-drain-it]] if the immediate problem is a backlog rather than a definition.
Sibling skills
- [[../duo-measurement-validity/SKILL]] — whether a number means what you think it means, before you grade anyone on it.
- [[../duo-efficacy-measurement/SKILL]] — the slow, rigorous outcome study that a daily proxy is standing in for.
- [[../duo-experimentation/SKILL]] — the A/B machinery your metric has to be readable inside.
- [[../duo-product/SKILL]] — "take the long view" as the stated rule that keeps a fast lever from beating a slow mechanic.
Sources
blog.duolingo.com — time-spent-learning-well, product-principles, can-duolingo-make-me-fluent, how-were-improving-duolingos-course-creation-process, how-user-reports-improve-course-content, chess-points-of-pieces, duolingo-leagues-leaderboards, duolingo-english-test-readiness, what-languages-do-they-study-in-france