More screenshots
Works with the AI Agents you already use
Measure What Matters: Analytics Strategy and KPI Design
Settle what counts before anyone wires up tracking: one guiding number, owned drivers, an event contract, and three views — a signed measurement plan your whole team can read the same way.
$5
Measure What Matters: Analytics Strategy and KPI Design
Example session with this skill installed
I run a fictional plant-care subscription called Sprout & Stem. We deliver potted plants and care kits to households. I need a measurement plan before we wire up tracking: what is our one guiding number, what drives it, what events do we track, and how do we know the plan is good? Give me the plan plus the rubric score.
- Read your context and instructions
- Compiled the measure what matters
MEASUREMENT PLAN — Sprout & Stem, invented scenario and numbers for this demo only.
Decisions: whether referral invites are worth the spend, whether the free first-month offer brings in subscribers who stay, and whether late deliveries are the fix that matters most.
Guiding number: households with a delivery in three of the last four weeks. First candidate, total app downloads, failed check one — downloads rise with ads whether or not anyone waters a plant. Target: 900 households by end of June.
Ignore list: total downloads; raw page views; follower counts.
Driver tree: invites accepted (owner: Nadia, 120 a month); first orders (Nadia, 100 a month); first delivery within four days (owner: Sam, 90%); second delivery within three weeks (Sam, 50%); cancellations (Nadia, under 25 a month).
Event contract: invite_accepted, order_placed, delivery_received, plus cancellation and skipped weeks. Identity: household account set at first order. Never collected: payment card details. Retention: twelve months.
Views: leadership (guiding number plus five leaves, weekly); team (owned leaves, daily); exploration (raw events, directional only).
Attribution: last-touch, weakness noted — it over-credits the final email before an order. First-touch runs as secondary. Slices: plan tier, channel, tenure.
Rubric: 12 out of 12. The plan is signed.
Connects securely to your tools. The creator never sees your data.
About this skill
Most teams do not suffer from too little data. They suffer from too many numbers with nobody agreeing which one counts, so each meeting quotes a different figure and the argument about whether things are going well never settles. This skill settles it in writing before a single event is wired up. It starts from the decisions the numbers must support, collected as plain sentences from each stakeholder, and sorts them into value, growth, and fixes. Then it picks one guiding number and tests it against three checks: it rises only when a customer gets something they wanted, it is hard to inflate quietly, and a stranger could guess the business from it. An ignore list names the flattering numbers the team agrees to leave out of reviews. The guiding number breaks into a driver tree of at most twelve leaves, every leaf with a plain definition, an owner by name, and a dated target. Only then comes the event contract: one discrete past-tense event per leaf, shared field names, identity rules, never-collected lines, and retention and removal rules. Three views follow — leadership, team, and exploration — one per audience, every chart compared to something. Attribution, slices, and experiment rules are settled before the first argument about them, with the chosen model's weakness written beside it. A six-line rubric scores the plan, and it signs at ten points from twelve possible with no row at zero. Edge cases cover brand-new products, teams that already track everything, and thin data, and the failure-mode catalog names number theatre, definition drift, the vanity comeback, privacy as an afterthought, and set-and-forget plans.
- It produces a signed measurement plan, not a tracking implementation, a dashboard build, or a data pipeline; those come after the plan is signed.
- It cannot write a plan when the team cannot name five real decisions; that is a strategy conversation, and the skill stops rather than measuring the wrong thing well.
- A plan is only as good as its review date: if nobody re-tests the guiding number, it goes stale like any other document.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 3 days ago
Permissions
No special permissions declared or detected