Product Learning Lab
Skill · Analytical

Metric drop diagnosis

Locate where a number moved before you decide what to do about it.

A metric that dropped tells you something changed — not what, where, or why.

Scenario

Weekly active users on a B2B tool fell 9% last week. Leadership wants a fix by Friday.

What do you do first?

Why the aggregate lies

One number, many stories

Weekly active users is a sum across enterprise and self-serve, new and tenured, mobile and web. A 9% dip could be one large account offboarding, a tracking change that stopped logging events, a seasonal lull, or a genuine product regression.

Reacting to the aggregate treats all of those as the same problem — and each needs a completely different response. The skill is refusing to act on the top-line number until you've found the slice it moved in.

The method

How to localize a drop

  1. 01

    Split by segment

    New vs returning, plan tier, account size, geography. Which cohort moved?

  2. 02

    Split by surface

    Which feature, page, or step lost usage — not the product as a whole?

  3. 03

    Split by time

    A cliff on one day points to a release or outage; a slow slope points to something structural.

  4. 04

    Check instrumentation

    Rule out a tracking or pipeline change before you trust the number at all.

  5. 05

    Form one hypothesis

    Name the single most likely cause the slices point to — then test that, not everything.

  6. 06

    Then act

    Now the fix is aimed at a cause, not a symptom.

Reflect

The last time a metric moved on a product you know, what was the first cut you made — and what did it let you rule out?

Takeaway

Strong PMs don't fix the metric that moved. They find the slice it moved in, then fix the cause there.

Keep going
Skill · Analytical

Problem vs Symptom

Separate what you observed from what caused it.

Skill · Analytical

Funnel thinking

See exactly which step loses users.

Practice

Try a case

Diagnose a real activation drop end to end.