Funnel analysis
Why it matters
A single top-line figure hides the problem. If overall conversion looks weak, the instinct is to redesign the homepage. The funnel might show that signup is healthy and that people fall away after signing up, which points to onboarding. Fixing the wrong step leaves the number you wanted to move almost where it was.
A funnel also tells you where a pound of effort goes furthest. Lifting a step from 20 per cent to 25 per cent adds a quarter more people to everything after it, whereas lifting a step that is already healthy adds far less.
And it gives a team a shared picture. Marketing owns the early steps, product the middle and sales the end, and each can see its own rate move, which makes it easier to agree where to focus.
How to apply it
- Define the steps once, in writing, so everyone measures the same path. Set them up in an event tracking tool.
- Calculate the rate between each pair of steps, not only start to finish.
- Compare each step with its own history. The step that has fallen most is often a better lead than the step with the lowest rate.
- Split the funnel by traffic source or customer type. A drop that appears only for one channel is a targeting problem, not a product problem.
- Look at time as well as counts. A long gap between steps is a drop-off that has not happened yet.
- Re-check after every change, since a fix at one step can move the drop-off to the next.
What it is
A funnel is a fixed sequence of steps, such as visited the site, signed up, activated, paid. Funnel analysis counts how many people reach each step and works out the conversion rate between each pair.
Say 10,000 people visit, 800 sign up, 240 activate and 48 pay. Visit to signup is 8 per cent. Signup to activation is 30 per cent. Activation to payment is 20 per cent. The overall rate from visit to payment is 0.48 per cent, but that one number says nothing about where the loss happens. The step rates do.
Common mistakes
- Too many steps. Five or six that matter beat twenty.
- Mixing groups who joined at different times, which distorts the rates. Cohort analysis keeps them separate.
- Reading small numbers as a trend. A step with 30 people in it moves wildly.
