Drop-off analysis

Drop-off analysis is the practice of measuring where users leave a multi-step flow before they finish it. Instead of reporting a single completion percentage, it splits the flow into ordered steps and calculates how many users are lost between each step and the next. The result is a map of the funnel that shows exactly where attrition happens.

The value of the analysis is precision. A flow might convert 40 percent of the people who start it, but that number says nothing about where the other 60 percent went. Drop-off analysis answers that question. It might reveal that the flow holds users well through three steps and then loses half of them at a single payment or verification screen.

Knowing the where points you straight at the why. A step with an unusually high drop-off is rarely a random event. It usually signals friction: a confusing form, an unexpected requirement, a slow page, or a moment where the value of continuing is not obvious. Drop-off analysis turns a vague sense that users are leaving into a specific, fixable problem.

For each step, take the number of users who entered it, subtract the number who advanced to the next step, divide by the number who entered, and multiply by 100. That gives the drop-off rate for that step.

For example, if 1,000 users reach the payment step and 700 advance past it, the drop-off at that step is 30 percent. Repeat the calculation for every step and you have the full picture of where the flow leaks. To find overall completion, divide the users who finish the final step by the users who entered the first step.

1. 1

   Define the ordered steps

   List the steps of the flow in sequence and decide what counts as entering and completing each one. The steps must be mutually exclusive and consistently ordered so the math holds.

2. 2

   Count entries and exits per step

   For every step, count how many users entered and how many advanced. Use a fixed cohort and a fixed time window so users are not double counted as they move through.

3. 3

   Calculate drop-off at each step

   Apply the step drop-off formula to each step. Rank the steps by drop-off rate to find the single biggest point of loss, which is where attention should go first.

4. 4

   Compute overall completion

   Divide the users finishing the final step by the users entering the first step. This headline number contextualises the per-step rates and tracks whether fixes move the whole flow.

A funnel is already a chain of cause and effect, which makes it a natural fit for a metric tree. Overall completion sits at the top, and each step sits beneath it as a driver. Decomposing the tree further shows the reasons users leave a given step, so the analysis moves from where to why in one structure.

Drop-off benchmarks depend heavily on the type of flow and the effort it demands. A long onboarding asks more of users than a one-page checkout, so higher per-step loss is expected. Use these ranges to judge whether a single step is an outlier rather than to set an absolute target.

Improving drop-off is not a single action but a loop: find the worst step, understand the friction, change it, and confirm the change moved the number. The cards below cover the moves that resolve the most common causes of step abandonment.

1. 1

   Reading drop-off without a fixed cohort

   If users enter the funnel across different days and you measure each step on a different population, the rates do not line up. Use one cohort moving through the flow within a fixed window.

2. 2

   Treating a moved loss as a fixed loss

   A change that reduces drop-off at one step can simply push the abandonment to the next step. Always check overall completion, not just the step you changed.

3. 3

   Ignoring low-traffic steps with high impact

   A step with few users but a very high drop-off can still be worth fixing if those users are valuable. Weigh drop-off rate against the value of the users at that step.