Analytics Strategy Creator

What is an analytics strategy?

An analytics strategy is a short document that names the specific decision you're trying to make, then picks the few metrics that actually inform it. It's the thing you write before you build a dashboard, set up tracking, or argue about KPIs — so you don't spend the next quarter staring at numbers that don't change anyone's mind.

A good strategy answers four questions:

• What's the decision? (e.g., “invest in onboarding redesign vs. paid acquisition”)
• What 3–5 metrics inform it? (paired with counter-metrics to catch gaming)
• How do we segment each metric? (e.g., traffic_source, plan_tier, new_vs_returning)
• Which vanity metrics do we explicitly skip? (named, not generic)

Without a strategy, you end up with a 30-metric dashboard nobody looks at, instrumentation work that doesn't map to any decision, and a team that argues about gut feel because the data doesn't resolve anything. A 30-metric dashboard is not a strategy; it's a graveyard.

Why founders skip analytics strategy (and pay for it later)

Most early-stage founders skip the strategy step and jump straight to “let's set up Mixpanel.” Six months in, they have:

• A dashboard with DAU, MAU, signups, and page_views — none of which change what they ship next
• Metrics that go up while the business goes sideways (no counter-metric to catch it)
• Aggregate numbers that hide the segment where the real problem lives
• Tracking debt built on top of a strategy that was never written down

By the time you realize the numbers aren't actually helping you decide anything, you've already shipped a dashboard the team doesn't trust — and you're back to gut feel.

What this tool does

Once installed in Claude, you tell it the decision you're trying to make and give it your product URL. Then Claude:

1. 01Pins down the specific decision before recommending any metric — and pushes back firmly if you answer with a vanity count or a vague category
2. 02Researches your product by fetching your marketing site and relevant sub-pages so it understands what you actually do
3. 03Asks a few sharp questions about stage, business model, and existing instrumentation
4. 04Picks 3–5 primary metrics, each paired with a counter-metric and segmentation dimensions — and names the vanity metrics you're explicitly skipping
5. 05Outputs a markdown strategy doc you can hand to your PM, eng team, or data person

What you get

A markdown strategy doc with these sections, ready to share or paste into Notion / Linear / your wiki of choice:

The doc is intentionally short — roughly 1–2 printed pages. If you want full instrumentation specs to hand to engineering, the output chains naturally into the Event Tracking Plan Generator.

Installation

Steps:

1. 01Download the .skill file from this page
2. 02Open Claude → Settings → Capabilities → Skills
3. 03Click “Upload skill” and select the file
4. 04Open a new conversation. The skill activates automatically when you ask Claude what to measure for a product decision.

How to use it

Start a new chat and say something like:

“I want an analytics strategy for https://app.yourproduct.com — I'm trying to decide [the specific decision].”

Claude will then walk you through four short steps:

1. 01Decision first — Claude pins down the specific decision you're making. If you answer with a vanity count (“I want to track DAU”) or a vague category (“engagement”), it'll push back until you have a real decision. The single test: what would you do differently if the metric is good vs bad?
2. 02Product context — share your URL and a few sentences on what the product does, who uses it, and the core action. Claude fetches your site to fill in the rest, then recaps so you can correct anything off.
3. 03Business context — stage (pre-PMF / early growth / scale), existing instrumentation (nothing yet / GA4 / Mixpanel / PostHog / Amplitude / full stack), and anything else worth knowing.
4. 04Scope recap — Claude confirms the decision, product, stage, and stack before writing. Then it produces the markdown strategy doc.

The whole run typically takes 10–20 minutes, depending on how sharp your decision is going in. The markdown doc is delivered at the end — short, opinionated, and ready to share.