How we estimate impact

Every figure we show can be checked on screen.

Putting a number on "how much revenue this fix could bring" is the heart of this product. A number you cannot trust is worthless, so we publish exactly how it is calculated. One formula, one assumption.

One formula, used by every card

However the issue was detected, the money is always built the same way. Multiply these four and you get the figure on the card.

measured
Shortfall

How far you fall short of a figure measured on your own site

measured
Pass-through to purchase

The measured share that goes on to buy

measured
Average order value

Measured on your own site

the only assumption
30% recovery rate

You never recover all of a gap in practice

= Estimated recovery

All three multipliers come from numbers measured on your own site. Industry averages and published studies are never used in the arithmetic — see below for why.

A full worked example

A card from a live ecommerce site: the share of product viewers who add to cart has fallen versus the previous period.

Product view → add to cart converts worse than last period (27% → 24%)
Est. recovery ~¥173,953 (assumes 30% recovered)
The numbers on the card
Product view → add to cart conversion
23.6%
prev 26.8%
-3.2pt
Product view → add to cart base
11,859
How this figure is calculated
Missing product view → add to cart steps 11,859 × 3.2pt 379
× Pass-through to purchase 2,795 add-to-carts → 1,395 purchases 49.9%
× Average order value revenue ÷ transactions ¥3,066
× Recovery rate the only assumption 30%
= Estimated recovery ¥173,953

This finding is based on 11,859 events.

Type 379 × 49.9% × 3,066 × 30% into a calculator and you get ¥173,953. The grey annotations are on the card too, so “where does 379 come from?” and “where does 49.9% come from?” are answered one step further back.

We only ever compare you to yourself

A shortfall needs something to compare against. Every comparison we use is measured on your own site — nothing is borrowed from outside.

Against your other devices

"Mobile converts at 4.4%, desktop at 14.0%." Same site, same period, same measurement — the conditions match.

Against your previous period

"It was 26.8% last month and 23.6% now." Getting back to your own past performance is an achievable target.

Against your site average

"Only this campaign converts at half your site average." The average is the target.

Against your repeat customers

"New customers convert at a third of repeat customers." Same products at the same prices, so the gap is real headroom.

Why industry averages never enter the arithmetic

There are well-known studies — "0.1s faster loading lifts revenue by 8.4%", for instance. We deliberately do not multiply your revenue by numbers like these.

You cannot verify it

The moment an outside coefficient enters the formula, there is no way for you to check where the 8.4% came from. Citing the source does not help: nobody can tell you whether that study applies to your store.

When it breaks, nothing tells you

Applying "0.1s → +8.4%" to a 5.4s improvement gives +453%. Relationships measured over a narrow range fall apart when stretched. With your own numbers you would notice; with borrowed ones you cannot.

The definitions do not even match

An "industry average cart abandonment of 70%" and the figure in your GA4 do not share a denominator. The same is true of conversion rate and bounce rate — every tool and company defines them differently. Your own devices and your own previous period match exactly.

This does not mean we ignore industry knowledge. We use it freely to explain why a fix works and to decide what deserves your attention. Google’s Core Web Vitals target (LCP under 2.5s), for example, is standardised down to the measurement method, so we show it as a target. The numbers come from your data; the reasoning comes from the field.

Four kinds of amount, with different meanings

Only one of them claims to be an estimate. The rest are statements of fact, not promises.

Estimated recovery

Revenue you could win back, from the formula above. This is the only estimate. It always carries the 30% assumption, and the derivation opens inside the card.

Est. recovery ~¥318,699 (assumes 30% recovered)

Scale

An actual figure, not an estimate. It tells you how much revenue currently rides on this page, channel or region.

Revenue on this page ¥913,622

Change

The movement versus the previous period. Not an amount you can recover. Whether a drop is a post-sale lull or a real problem cannot be decided from the data, so we hand you the fact.

vs previous period -¥441,000

No amount

When we cannot build a defensible figure, we show no number at all — just the finding and the facts behind it.

(blank)

What we decided not to show

What we leave out matters as much as what we show. Each of the following could be turned into a formula — we do not, because the result could not be justified.

We do not turn a speed score into money

A page speed score is a composite of several lab metrics, not a quantity. Multiplying revenue by it does not yield an amount, so speed cards show the actual revenue on the page instead. The score still drives the priority order.

We do not reuse a "this is a problem" threshold as a target

A line like "fewer than 1.3 items per order suggests cross-sell headroom" exists to draw attention. It is not evidence that you can reach 1.3. Reusing it would leave the figure unsupported, so those cards carry no amount.

We do not present a revenue drop as recoverable

A post-sale lull, a launch settling down, a stock-out, a product rename — the reasons differ, and the data cannot tell you whether it comes back. We show the signed change as a fact.

We stay quiet when the sample is too small

Rates built on few events are unstable. In that case we show no amount, only counts and facts. We never dress up something we could not read.

About the one assumption: a 30% recovery rate

The shortfall, the pass-through and the order value are all measured. What cannot be measured is how much of a gap you actually close in practice, so we apply a single conservative assumption of 30%. Every card states it, so read it however matches your experience — halve it if you are sceptical, raise it if you are confident.

This 30% will not stay an assumption. Because we run our own measurement, we can observe across many stores how much of a mobile-speed gap actually closes after a fix. Once we get there, the estimates carry no assumption at all. That is somewhere borrowed studies can never reach, and it is the strongest thing about how we work.

Today: an assumption (30%, the same on every card) Next: replaced by a figure measured across our own network

Questions we get

Is this amount guaranteed?

No. It is an indication of the size of the opportunity. That is precisely why we publish the whole calculation — judge it by where the number comes from, not by the number itself.

Can I add up all the cards to get my upside?

No. The same visitors are counted by more than one card (a mobile conversion gap and a cart-flow gap overlap, for example), so a sum would double count. Use the amounts to decide the order of work.

The amounts look smaller than I expected

In August 2026 we rebuilt the calculation and stopped showing figures we could not justify. Some cards now show less. In exchange, every figure that remains can be checked from the numbers on screen.

Some cards show no amount at all

Those are cards where no defensible figure exists. The finding is still useful, so we show the facts alone rather than invent a number.

Can I change the 30% myself?

Not yet — it is the same on every card today, and always stated so you can reinterpret it. Making it configurable per site is on the roadmap.

See a real card

Open the demo — no signup — and expand the derivation on any card.

Try the demo

The 7-day free trial needs no credit card.

This page reflects the current implementation. If we change how the figures are calculated, we update this page. Our privacy approach