Sales Funnel Drop-Off Calculator

Setting it out

Drop-off at a step is the users lost divided by the users who reached it: (entered − continued) ÷ entered × 100. It is the complement of the step conversion rate.

Expressed in absolute numbers rather than percentages it is frequently more useful, since a 30% drop-off at the top of a funnel costs far more people than a 60% drop-off at the bottom.

Lost value = users dropped × the probability they would have converted × contribution per order.

A concrete case

From the funnel above: 33,600 lost at product view, 3,067 at add to cart, 2,000 at checkout start, 333 at checkout completion.

In percentage terms the worst step is product view at 84%. In value terms the picture differs, because those 33,600 were far less likely to buy than the 333 who abandoned at payment.

Valuing each: the 333 checkout abandonments were worth perhaps 60% of a conversion each: 200 orders, $6,380 of contribution. The 33,600 product-view drop-offs at maybe 1% each are 336 orders, $10,718.

Both are large and they require completely different work, which is why the percentage view alone misdirects effort.

What the number hides

The largest percentage drop-off is usually at the top of the funnel and is usually the least fixable, because it includes everyone who was never going to buy.

Assigning a conversion probability to dropped users is also an estimate rather than a measurement, and the whole value calculation is sensitive to it.

Where to go from here

Weight each drop-off by how close the user was to purchasing, so the ranking reflects value rather than volume. Late-stage losses are worth far more per person than early ones.

Then look at the trend for each step rather than the level. A step whose drop-off has worsened is a specific, recent problem with a likely cause; one that has always been high may simply be the shape of the funnel.

Diagnosing a step that has worsened

A sudden increase in drop-off at a specific step almost always has a technical or a content cause with a date attached: a deployment, a price change, a payment provider issue, a third-party script slowing the page.

Cross-referencing the change against the deployment log and the analytics annotations usually locates it in minutes, which is far faster than treating it as a design problem.

That is the strongest practical argument for step-level instrumentation. A site-wide conversion rate falling 8% gives no starting point; a checkout step falling from 75% to 61% on a specific Tuesday gives a very short list of things to check.

Annotating the analytics with deployment dates makes a sudden change in any step immediately traceable, which turns a week of investigation into a few minutes of checking.

Most tools support annotations and almost nobody uses them, which is why a step-level deterioration is so often diagnosed long after the change that caused it.

Comparing drop-off between new and returning visitors usually shows the steps where unfamiliarity is the barrier, which points at explanation rather than at interface changes.