How funnel analysis works
A funnel is an ordered list of steps toward a goal. Funnel analysis counts the users who reach each step and computes the conversion between steps:
step conversion = users at step N / users at step N-1 x 100
A worked example for a shopping app over a 7-day window:
- Install: 10,000 users
- Sign-up: 6,000 users (60%)
- Product view: 4,800 users (80%)
- Add to cart: 1,440 users (30%)
- Purchase: 576 users (40%)
Overall conversion from install to purchase is 576 / 10,000 = 5.76%. The biggest drop is from product view to add to cart, so that is where a product change pays off most.
Building a useful funnel
- Keep it short. Four to six steps that map to real decisions.
- Use events that mean something. "Viewed paywall" and "started checkout" beat generic screen views.
- Set a window. Decide how long a user has to complete the funnel.
- Choose strict or open ordering. Strict for diagnosing a flow, open for a broad picture.
- Segment it. By platform, country, app version and, most of all, acquisition source.
Funnels by acquisition source
The same funnel split by campaign often tells a different story than the blended view. Imagine a TikTok campaign and a Google campaign with similar install volume. TikTok users sign up at 70% but purchase at 2%; Google users sign up at 50% and purchase at 6%. The blended funnel hides that TikTok's problem is late in the journey while Google's is early. Each needs a different response: a different offer or audience for TikTok, a shorter sign-up for Google.
Funnels in Linkrunner
Linkrunner's Events tab includes funnels built from the in-app events you send, and cohorts can be turned into funnels as well. Because every event is attributed to the campaign behind the install, you can read funnel drop-off by source and compare it with cost per event. Funnels can also be created through the MCP server from an AI assistant. See Linkrunner analytics.