Funnel Conversion Analyzer: Find Where Visitors Drop Off

A conversion funnel is a simple idea with powerful implications: visitors arrive, take a series of steps, and only some complete the final action, whether that is a purchase, a sign-up, a booking, or a download. At each step, some visitors leave, and the number who continue gets smaller. The Funnel Conversion Analyzer turns raw counts at each step into conversion rates between steps and overall, so you can see where the largest losses happen. This guide explains how to build a funnel, calculate its rates with checked arithmetic, interpret the results without overclaiming, and decide what to improve first. It is general business and marketing information, not a guarantee of any traffic, conversion, or revenue outcome.

Funnels are useful because they replace a single overall conversion rate, which hides where the problem is, with a stage-by-stage picture. An overall rate of two percent could mean a healthy funnel with a small audience, or a leaky funnel where most visitors leave at one broken step. Only the stage rates distinguish those cases, and only the stage rates tell you where to work. A funnel that is not broken down by stage is a summary, not a diagnosis.

Building a funnel that means something

A funnel needs clearly defined steps in the order visitors actually take them. A typical e-commerce funnel might be: visited product page, added to cart, started checkout, entered payment, completed purchase. A sign-up funnel might be: visited landing page, started form, submitted form, confirmed email, completed onboarding. The steps must be events you can count reliably from analytics or product data, with the same definition over time. If the definition of "started checkout" changes halfway through a comparison period, the rates before and after cannot be compared, and any conclusion drawn from them is unreliable.

Each step's count should be the number of visitors or sessions that reached that step, counted once per visitor or session as appropriate, and consistent across steps. Mixing visitors at one step with sessions at another, or counting repeat visits as new visitors at one step but not another, distorts the rates. Also decide whether your funnel is for a specific segment or for all traffic. Overall funnel rates mix segments with very different behaviour, such as mobile and desktop, new and returning visitors, or paid and organic traffic, and the mix can hide a problem in one segment behind good performance in another. Analyse the overall funnel first for an overview, then break it down by the segments that matter for your decisions, such as device, traffic source, or geography, provided each segment has enough visitors for its rates to be stable.

Time windows matter too. A funnel measured over a day with fifty visitors will have noisy rates, where a handful of visitors changes a percentage by several points. A funnel measured over a month with thousands of visitors will have steadier rates, but it may hide a change that happened mid-month. Choose a window with enough volume for stability, and note any changes during the window, such as a site update, a campaign launch, or a pricing change, that could explain a shift. Do not compare funnels across different windows without noting those changes.

The calculations, step by step

There are two kinds of conversion rate in a funnel, and both are useful. Step conversion rate is the number reaching the next step divided by the number reaching the current step, multiplied by one hundred. Overall conversion rate is the number reaching the final step divided by the number reaching the first step, multiplied by one hundred. Drop-off at a step is one hundred percent minus the step conversion rate, or in counts, the current step count minus the next step count. Overall conversion rate also equals the product of all step conversion rates, which is a useful check: multiply the step rates as decimals, and you should get the overall rate as a decimal. If the product does not match, an arithmetic or counting error has crept in.

A common misreading is to treat a low step rate as automatically the most important problem. Importance depends on both the rate and the volume at that step, and on how fixable the step is. A step with a fifty percent drop-off but only one hundred visitors loses fifty visitors, while a step with a twenty percent drop-off but ten thousand visitors loses two thousand visitors. The larger absolute loss usually deserves attention first, unless the small step is a critical bottleneck for a high-value segment or the fix is trivial. Always look at counts and rates together, not rates alone.

Worked example with sample values

Take a sample e-commerce funnel with the following counts, clearly labelled as samples: visited product page, ten thousand; added to cart, two thousand; started checkout, one thousand two hundred; entered payment, nine hundred; completed purchase, seven hundred and twenty. Step rates are: product to cart, two thousand divided by ten thousand, multiplied by one hundred, which equals twenty percent; cart to checkout, one thousand two hundred divided by two thousand, multiplied by one hundred, which equals sixty percent; checkout to payment, nine hundred divided by one thousand two hundred, multiplied by one hundred, which equals seventy-five percent; payment to purchase, seven hundred and twenty divided by nine hundred, multiplied by one hundred, which equals eighty percent. Overall conversion is seven hundred and twenty divided by ten thousand, multiplied by one hundred, which equals seven point two percent. Check by multiplying step rates as decimals: zero point two multiplied by zero point six equals zero point one two; multiplied by zero point seven five equals zero point zero nine; multiplied by zero point eight equals zero point zero seven two, which is seven point two percent, matching the overall rate.

Drop-offs in counts are: product to cart, ten thousand minus two thousand, which equals eight thousand visitors lost; cart to checkout, two thousand minus one thousand two hundred, which equals eight hundred; checkout to payment, one thousand two hundred minus nine hundred, which equals three hundred; payment to purchase, nine hundred minus seven hundred and twenty, which equals one hundred and eighty. The largest absolute loss is at the first step, eight thousand visitors who viewed a product but did not add it to cart. That step has the lowest rate, twenty percent, and the largest count, so it dominates the funnel. Improving product-to-cart from twenty percent to twenty-five percent, while other steps stay the same, would give two thousand five hundred cart additions instead of two thousand. Applying the later step rates, checkout starts would be two thousand five hundred multiplied by zero point six, which equals one thousand five hundred; payment entries would be one thousand five hundred multiplied by zero point seven five, which equals one thousand one hundred and twenty-five; purchases would be one thousand one hundred and twenty-five multiplied by zero point eight, which equals nine hundred. That is nine hundred purchases instead of seven hundred and twenty, an increase of one hundred and eighty, from a five-point improvement at the first step alone. This sample shows why stage analysis matters: the same five-point improvement at the payment step, from eighty to eighty-five percent, would give nine hundred multiplied by zero point eight five, which equals seven hundred and sixty-five purchases, an increase of only forty-five. The first step's improvement is worth four times as much in this sample, because it acts on a larger volume.

That arithmetic is a projection, not a promise. Real improvements depend on why visitors leave, which the numbers alone do not explain. A low product-to-cart rate could reflect price, product relevance, traffic quality, page clarity, or a technical problem. The funnel tells you where to investigate; user research, session observation, and testing tell you why. Never present a funnel projection as a guaranteed outcome, to yourself or to anyone else.

Common mistakes to avoid

The first mistake is defining steps that do not match the real journey, such as omitting a step visitors must actually take, which makes the funnel look better than reality. The second is mixing counting units across steps, as discussed above. The third is drawing conclusions from tiny samples, where a rate change of several points is just random variation. The fourth is ignoring segments: a funnel that looks healthy overall may be broken on mobile, or for one traffic source, and the overall rate hides it. The fifth is attributing a funnel change to a single cause without checking for other changes in the same period, such as a campaign, a price change, or a site outage. The sixth is optimising a step rate by making the step easier to pass but less meaningful, for example by pre-selecting options that inflate a step count without increasing real completions. That improves the funnel's appearance while harming the business.

Another mistake is treating funnel analysis as a one-off report. Funnels drift as products, prices, audiences, and competitors change, so the value comes from regular measurement with consistent definitions, and from tracking whether a change actually moved the step it was meant to move. If you change a checkout page and the checkout-to-payment rate does not move, the change did not work as intended, regardless of how good it looked in review. If the overall rate moves but no step rate moves clearly, check for a mix change, such as a different traffic source share, before claiming the change worked.

Deciding what to improve first

Rank steps by absolute drop-off in counts, not just by rate, and consider fixability and value. A large drop-off at a step with a clear, fixable cause, such as a confusing form, a slow page, a missing payment option, or an error message, is a strong candidate. A large drop-off caused by low purchase intent, such as visitors browsing without any intention to buy, may be harder to fix and may be better addressed by attracting better-qualified traffic rather than by changing the page. For each candidate, form a specific hypothesis: what do you think is causing the drop-off, what change would address it, and what step rate would you expect to move if you are right? Then test the change in a controlled way where possible, measure the step rate before and after with sufficient volume, and keep or discard the change based on the measurement, not on opinion.

Document the funnel, the segment, the window, the step definitions, and the counts alongside the rates, so anyone reviewing the analysis can reproduce it. That documentation is what turns a funnel chart from a persuasive picture into a trustworthy analysis. Revisit the funnel after each significant site or campaign change, and review whether the steps still match the journey, because journeys evolve and funnels that do not evolve with them become misleading.

For related calculations, the Marketing Budget Allocator helps you plan spend across channels that feed the funnel, and the Percentage Calculator helps with rate changes and comparisons. You can also read the related guide on cost per click and related metrics at /blog/cpm-cpc-and-cpa-calculator-guide/ for how traffic costs relate to funnel outcomes. Use the funnel to find the largest, most fixable loss, investigate why it happens, test a change, and measure honestly. That discipline, repeated, is what improves conversion over time, without any guarantees about any single change.