AddThisFeature

Funnel Analytics

Show where people move forward in a journey and where they drop out.

involved Analytics

What it adds

Definable multi-step funnels with conversion windows, identity stitching, and versioned definitions so historical results stay stable.

What your agent is told to do

6
  1. 1

    Let a user define a funnel as an ordered list of steps, each backed by an existing tracked event, with optional property filters per step.

  2. 2

    Make ordering an explicit choice: strict ordered funnels require steps in sequence, unordered ones only require all steps to occur. State which mode a funnel is in on the chart.

  3. 3

    Require a conversion window per funnel — a fixed span from first step to last. Without one, every funnel eventually converges on 100% and means nothing.

  4. 4

    Stitch anonymous activity to the authenticated identity when a user signs in, so the top of the funnel is not permanently orphaned from the bottom.

  5. 5

    Version funnel definitions. Editing steps must create a new version; historical results stay attached to the definition that produced them.

  6. 6

    Do NOT define, rename, or capture events here — event capture, naming, and schema are owned by Event Tracking. Funnels read from it.

Edge cases it handles

8
  • Bots and internal traffic inflate the first step and deflate every rate below it. Exclude them by default and say so on the chart.
  • Repeated events from one person must count once per funnel entry, not once per event.
  • A user who completes step three without step two must be handled by the ordering rule, not quietly counted as converted.
  • Users who enter near the end of the reporting range have not had a full conversion window yet; either exclude them or label the recent buckets as incomplete.
  • Stitching must not merge two people who shared a device — bind to the sign-in, and never retroactively merge on a weak signal like IP address.
  • A renamed or deprecated event must break the funnel loudly rather than silently reporting zero at that step.
  • Small cohorts produce unstable percentages; show the underlying counts next to every rate.
  • Deleted user data must disappear from funnel results too, which means recomputing or excluding rather than trusting a cached aggregate.

Definition of done

9
  • A funnel is defined as ordered steps over existing tracked events with per-step filters.
  • Ordered versus unordered mode and the conversion window are both explicit and shown on the chart.
  • Anonymous sessions are stitched to identities on sign-in without merging distinct people.
  • Bot and duplicate events are excluded, and the exclusion is disclosed.
  • Editing a funnel creates a new version and leaves historical results intact.
  • Absolute counts are shown alongside every conversion rate.
  • Event capture and naming remain owned by Event Tracking.
  • The feature matches the existing design system.
  • No existing functionality is broken.

Related features

How it works

  1. 1

    Copy the link

    Grab the Markdown instruction URL for this feature.

  2. 2

    Give it to your AI

    Paste it into Claude Code, Cursor, v0, Lovable — whatever you build with.

  3. 3

    It inspects, then implements

    Your agent reads your existing app first, then adds the feature to fit it.

Works with your stack

These instructions are written to adapt. They tell the agent to detect your framework, match your existing design system, and reuse what you already have — rather than assuming a particular stack.

Need it tighter than that? Customize the feature and tell it exactly what you're running.