How to calculate conversion loss in euros

Putting a euro figure on a conversion problem takes four numbers and about ten minutes. The arithmetic is easy. The honest part is admitting that one of the four is a guess, and that the size of your answer depends almost entirely on it.

How do you calculate conversion loss?

Multiply your traffic by the conversion gap you could close, by your average order value. Revslip uses one published formula: traffic × CVR gap × AOV × mobile weight. Traffic and order value come from your analytics. The mobile weight scales a defect that only appears on phones. The gap is the problem.

Three of the four are measurements. One is a forecast.

Traffic you can read off a dashboard. Average order value is arithmetic on last quarter. Mobile weight is the share of sessions the defect actually touches. None of these are in dispute, and none of them decide the size of your answer.

Where does the conversion gap come from?

The gap is the difference between your rate now and your rate after the fix, and nobody can measure the second half in advance. Most calculators let you type in whatever you hope for. Type 2% instead of 1% and you have doubled your loss figure without learning anything about your site.

So do not type in a hope. Find a gap you already have.

Your own traffic is already segmented into groups converting at different rates, and the distance between them is a measured number rather than a wish. Four places to look for one.

1. The device gap. Desktop at 2.1% and mobile at 0.9% is a 1.2 point gap you can see today, not one you invented. Closing mobile to two thirds of desktop is a defensible target, for the reasons in mobile conversion rate lower than desktop.
2. The source gap. If one channel converts at half the others on the same landing page, that difference is your ceiling. Splitting it properly is covered in is it my traffic or my website.
3. The step gap. The pass rate between two checkout events, measured against the step before it. The steepest drop is the one worth pricing.
4. The returning visitor gap. People who have seen the site before almost always convert better. If first-time visitors are far behind them, the problem is something a stranger hits and you no longer can.

Working from a gap you already have? Get yours measured free.

How big a lift is realistic?

Smaller than the calculators suggest, and less likely to happen at all. Ron Kohavi and Roger Longbotham report that only one third of ideas tested at Microsoft improved the metric they were built to improve, and they cite roughly 10% at Google. Most changes do nothing at all.

The ones that do land are modest.

Georgi Georgiev re-analysed 115 publicly published A/B tests and found 70% of them underpowered. After pruning the compromised ones, the remaining 85 averaged a 3.77% relative lift. Even the statistically significant winners averaged 6.78%.

Note what those two sources have in common, which is nothing. Kohavi and Longbotham are describing an enterprise experimentation platform in an academic reference work. Georgiev sells A/B testing software, and his finding argues that the wins his customers publish are small. They arrive at the same place from opposite directions.

Multiply by the odds it works

A loss figure is a best case unless you weight it. If a third of fixes land and the typical winner moves the number by a few percent, then the expected value of a change is a fraction of the headline. Report both: the size if it works, and the size times the probability that it does.

The number you should defend in a meeting is not what the fix could be worth. It is what it is worth multiplied by the chance you are right.

A worked example you can copy

Take 22,000 sessions a month, 64% of them on mobile. Desktop converts at 1.8%, mobile at 0.7%. Average order value is €59. Every number there is measured rather than assumed, because both rates came off the same dashboard this morning and neither is a target anyone picked.

Here is the arithmetic.

Closing mobile to two thirds of desktop means 1.2%, a 0.5 point gain on 14,080 mobile sessions. That is 70 more orders at €59, or €4,130 a month. Weight it by the one in three chance the fix lands and you get an expected €1,377 a month. Both numbers are true. They answer different questions.

The first is what to chase. The second is what to budget against.

Price your leaks in one pass, free.

What this number is not

It is a hypothesis with a euro sign on it. It tells you which of your problems is worth attention first, which is a genuinely useful thing, and it does not tell you what your revenue will be next quarter. Anyone presenting it as a forecast is overselling arithmetic. Turning it into a measured number afterwards is a separate job with its own arithmetic: how to track whether a fix worked.

That includes us.

Every euro figure Revslip puts on a finding inherits exactly this uncertainty, because it runs the same formula on the same kind of assumed gap. We publish the formula so you can check the working and disagree with the input. A number you cannot audit is not evidence, it is decoration, and the whole point of showing our maths is that you get to argue with it. We take that apart input by input in are website audit revenue estimates real.

One more caveat on our side of the ledger. Our frequency claims come from 134 audited sites, and those sites arrived because their owners already suspected something was wrong, so problems appear more often in our data than in the wild.

Questions people ask about pricing a conversion problem

These come up when someone has run the arithmetic once and needs to defend it to a finance team or a client. The recurring theme is that the formula is not the hard part, and the assumption behind the gap always is.

What if I have no idea whether my rate is bad?

Start there rather than with the formula. Sector rates vary widely enough that your number may already be normal, which is set out in is my conversion rate bad.

How do I know which problem to price first?

Rank before you calculate, because most sites have dozens of issues and pricing all of them is busywork. The severity model is in the conversion leak index.

Can I do this for my checkout specifically?

Yes, and the step gap is usually the cleanest one to measure because the events are already there. The five events and what each drop means are in why is my checkout abandoning.

Is the number big enough to justify hiring someone?

Compare it against the fee before you sign anything. The thresholds where an agency, a consultant or a tool pays back are in what a CRO audit actually costs.

How do I run this for a free trial instead?

Same formula, but pick the stage first, because a trial loses people in three places and only one of them is about price. The worked trial example is in where a free trial actually breaks.