Funnel analysis

Funnel analysis is the practice of measuring how users progress through an ordered sequence of steps towards a goal, counting how many arrive at each step and how many are lost between one step and the next. The shape is a funnel because the count narrows as users advance. A sign-up flow might start with ten thousand visitors and end with five hundred paying customers, and funnel analysis shows exactly where the other nine and a half thousand left.

The value is in the steps, not the endpoints. Anyone can see that a flow starts wide and ends narrow. Funnel analysis names the specific step where the steepest drop happens, which is where a fix returns the most. A flow that loses half its users at the email verification step has a very different problem from one that loses them at pricing.

The analysis depends on a clearly defined, ordered set of steps and clean event data for each one. If steps overlap, or if users can skip ahead, the counts blur and the drop-offs lose meaning. Strong funnel analysis starts with a sequence that genuinely reflects the journey and events logged consistently at every step.

Funnel analysis produces two numbers per step. The step conversion is the count reaching that step divided by the count reaching the previous one. The drop-off is one minus the step conversion. Multiply the step conversions together and you get the overall conversion from the top of the funnel to the goal.

For example, if 10,000 users start, 4,000 reach step two, and 1,200 reach step three, the first step conversion is 40 percent and the second is 30 percent. The overall conversion to step three is 12 percent. Reading the step conversions in order tells you which single step is leaking the most. The steps below give a repeatable way to build the analysis for any flow.

1. 1

   Define the ordered steps

   List the stages a user passes through, in strict order, from entry to goal. Keep each step a distinct, observable event so the counts do not overlap.

2. 2

   Count users at each step

   Count the distinct users who reach each step within the same window. Counting events instead of users double counts anyone who repeats a step and distorts the funnel.

3. 3

   Compute step conversion and drop-off

   Divide each step count by the one before it for the step conversion, then subtract from one for the drop-off. This is where the leakiest step becomes visible.

4. 4

   Compute overall conversion

   Multiply the step conversions together to get the end-to-end conversion from the top of the funnel to the goal, so you can track the whole journey alongside the parts.

A metric tree extends funnel analysis from a flat sequence into a causal model. The overall conversion sits at the top, each step conversion forms a branch, and the factors that drive each step form the leaves. Reading the tree downward shows not just where users drop off but why, and which lever might lift the step.

This is the difference between a dashboard and a decision. A funnel chart shows that checkout conversion fell. A metric tree shows that the fall came from a longer load time at the payment step, names the team that owns that step, and connects it to the goal it ultimately affects.

Funnel benchmarks depend heavily on the funnel type, the traffic source, and the commitment each step asks for, so the figures below are orientation rather than fixed targets. A free trial funnel and an enterprise sales funnel cannot share a target. What matters most is whether each step holds steady and whether your end-to-end conversion trends in the right direction.

Better funnel analysis comes from sharper step definitions and from acting on the one step that leaks most, not from optimising every step at once. The analysis points you at the bottleneck. Improvement is then a matter of fixing that step and confirming the overall conversion moved.

1. 1

   Counting events instead of users

   A user who retries a step is counted twice, inflating that step and hiding the true drop-off. Count distinct users at each step.

2. 2

   Mixing windows across steps

   Comparing this week at step one with last month at step two breaks the funnel. Hold the same time window across every step.

3. 3

   Ignoring users who skip steps

   If users can reach a later step without the earlier one, the drop-off between them is fiction. Define a strictly ordered sequence.