A funnel describes a sequence of steps a visitor takes to reach a goal: viewing a product, adding it to the basket, starting payment, completing the order. At every step some people stop. The funnel exists to show where people are lost, so that effort can be focused where it pays off most.
Defining the steps
The first job is conceptual rather than technical: deciding which steps matter. Some guidance:
- few steps: three to six are usually enough; a twelve-step funnel is hard to read;
- observable steps: each must correspond to a page view or an event recorded reliably;
- logical order: each step must be a prerequisite, at least in practice, of the next;
- one unit: count users or sessions, not a mix of the two.
An example for a site selling online courses:
- course page viewed;
- click on "Enrol";
- details form completed;
- payment started;
- payment confirmed.
Reading the numbers
Suppose you have this data for one month:
| Step | Users | From previous step | From first step |
|---|---|---|---|
| Course page | 4,000 | — | 100% |
| Enrol click | 600 | 15% | 15% |
| Form completed | 420 | 70% | 10.5% |
| Payment started | 380 | 90% | 9.5% |
| Payment confirmed | 250 | 66% | 6.25% |
The two columns tell different stories. The rate from the first step shows how well the whole path performs. The rate from the previous step shows where to intervene. Two points stand out here: only 15 per cent of people who view the course click to enrol, and a third of those who start payment do not finish it.
The second figure deserves immediate attention. Someone who starts payment has already decided to buy: losing a third of them suggests a concrete problem, such as missing payment methods, technical errors or extra costs appearing at the last moment.
Closed and open funnels
Analytics tools generally offer two modes:
- closed funnel: only people who enter at the first step are counted. Anyone who arrives directly at step three, for example through a saved link, is ignored;
- open funnel: anyone who reaches a step is counted from that point on.
A closed funnel is more rigorous for evaluating a specific path. An open one better reflects reality when people enter from different points. Using the wrong mode produces inflated or deflated rates without anyone noticing.
Time window
You also need to decide how quickly the steps must happen. Within the same session? Within seven days? For an impulse purchase the session is enough. For an expensive service many people come back days later, and a session-limited funnel would underestimate conversions. With tools that lack persistent cookies, analysis is usually limited to a single visit, which must be kept in mind when interpreting the numbers.
Segmenting
An aggregate funnel hides important differences. Break it down at least by:
- device: a form that is awkward on a phone creates a collapse visible only in the mobile segment;
- traffic source: visits from paid campaigns may behave differently from organic ones;
- new and returning visitors, where the tool allows it.
In my work this is the most frequent case: a funnel that looks healthy overall and, once split by device, reveals a step that is almost unusable on a smartphone.
From number to cause
The funnel tells you where, not why. To understand why you need other sources:
- form error events, with the name of the failing field;
- server logs for technical errors during payment;
- a few test sessions with real users, watching them complete the path;
- a short question for those who abandon, when it can be asked without being intrusive.
Common mistakes
- Steps defined by events that also fire in other contexts, inflating the numbers.
- Comparing funnels from different periods after changing the site without noting it.
- Optimising the first step, the largest in absolute terms, instead of the one with the most abnormal loss.
In short
- An effective funnel has a few observable steps, in logical order and with a single counting unit.
- The rate from the previous step shows where to act; the rate from the first measures the whole path.
- Choose deliberately between closed and open funnels and define the time window.
- Segment by device and channel: averages hide problems.
- The funnel shows where people drop out; to understand why you need errors, logs and user testing.
