Funnel Drop-off Calculator

Four numbers come out of a funnel, and each answers a different question.

• Step conversion = users at this step ÷ users at the previous step. How well one transition works on its own.
• Drop-off rate = (previous step − this step) ÷ previous step, which is 1 − step conversion. The calculator also shows the absolute number of people lost.
• Cumulative conversion = users at this step ÷ users at the first step. How much of the original audience is left.
• Overall conversion = last step ÷ first step, the cumulative figure for the final step.

Here is the sample funnel the calculator opens with, worked through:

Overall conversion is 4.8%. The highest drop-off rate is at “Added to cart” (64.3%), but the biggest leak is “Visited site → Viewed product”, which loses 5,800 people. The calculator ranks leaks by people lost, because that is where the most room for improvement sits.

Overall conversion is the product of the step conversions: 42.0% × 35.7% × 60.0% × 53.3% = 4.8%. That has a useful consequence. A 10% relative improvement at any step lifts the end-to-end rate by the same 10%.

In the sample funnel, raising product-to-cart from 35.7% to 39.3% takes purchases from 480 to 528. Raising checkout completion from 53.3% to 58.7% also takes them to 528. Same gain, very different work: one might be a better product page, the other a payment method you were missing.

So the question is not which step has the worst rate, but where a given lift is cheapest and most believable. The biggest absolute leak shows where most of the people are; the cost of fixing each step decides the order.

• Drop-off is measured between two steps of a funnel you defined: people who reached one step and not the next.
• Cart abandonment is a drop-off with a conventional name: 1 − purchases ÷ carts created. In the sample funnel that is 1 − 480 ÷ 1,500 = 68%.
• Bounce rate is not a funnel metric at all. It is the share of visits that saw a single page, wherever they started.

• Count people or visits, not page views. Someone who reloads the cart five times is one person at that step. Page views inflate early steps and make drop-off look worse than it is.
• Use one period for every step. Counts pulled from different date ranges produce a funnel that never happened.
• Steps should be nested. Each step should only contain people who passed the one before. If a later step has more users than the previous one, people are skipping a step or entering midway, and the calculator shows a negative drop-off to flag it.
• Small numbers move a lot. Forty people at a step is too few to compare one week against the next.

A funnel in PulsePanda counts these for you from the pages and events it already records, splits desktop from mobile, and links the replays of the visits that stopped, which is the part a calculator cannot do: it shows where people leave, not why. For the why, watch a handful of those session replays before you change the step.