← Back to the examples
Interactive · Growth systems

What is the funnel trying to tell you?

A metric can show where the journey broke. It usually cannot tell you why. Three launch situations. You start with the dashboard, choose which signals to inspect, and diagnose when you think you know enough.

The scenarios and figures are illustrative composites built to demonstrate the decision method. They are not customer or employer data.

Public exampleBuilt to show the method in use.

The method

Observe Segment Hypothesize Run the smallest useful test Route the action

0 of 3 situations diagnosed

Three situations, one discipline

In real launches, evidence is not free: every signal costs a meeting, a query, or a week. Inspect the sources in any order, or diagnose early and see what you had not read yet. Every option offered is defensible in a meeting. The question is which one the evidence best supports, and what the smallest useful test would be.

The method, in full

The signals disagree. That is normal.

No single dashboard held any of these answers. The diagnosis lived across product behavior, audience cuts, documentation searches, developer questions, route-to-market performance, and lifecycle response. The work was separating what the evidence establishes from what it merely suggests, naming what remains unknown, and choosing the smallest test able to change the decision.

Boundaries hold here too. Product Analytics owns the instrumentation and the trusted definitions. Product marketing connects those signals to audience, message, route-to-market, launch, and adoption hypotheses, then helps the relevant teams choose what to test next. The operating goal is learning velocity: replacing plausible explanations with decision-changing evidence, quickly and cheaply.

Original public model  ·  West of Obvious

Design and build notes

Strategic premise. Aggregate metrics locate a break; audience, behavioral, and qualitative signals diagnose it. Evidence is not free in real launches, so the model makes you spend for it, lets you commit early, and shows what the shortcut skipped. Segmentation helps interpret the evidence. Sometimes the cut reveals the story; sometimes it proves the problem is shared. The adoption journey maps where a journey can leak; this piece interprets what the leak means.

Public example. The situations and figures are fictional composites created to demonstrate the decision method. They contain no employer or customer information, and each includes at least one signal that is real but deliberately not yet a pattern.

Build. Original concept, writing, interface, and vanilla JavaScript. One file, no framework, no build step. Keyboard navigation, a screen-reader status line, and reduced-motion preferences are supported.