What is SEO testing worth on your site?
Tell us your traffic and what a session is worth, then pick the testing pace that fits your team. We’ll model a year against the counterfactual where you ship nothing and organic search quietly depreciates.
Nothing you enter leaves your browser.
Organic revenue ÷ organic sessions. Use gross revenue, or margin if that’s the number your CFO cares about.
Takes 10 seconds. Nothing you enter leaves your browser.
Each one is a fixed bundle of assumptions, priced on your numbers.
Not sure which scenario fits your site?
Counting revenue from winning tests plus revenue rescued from losing tests you’d otherwise have shipped, net of platform cost, measured against a do-nothing baseline where organic revenue depreciates.
No sign-up, no email wall. Every assumption below is visible and editable, and every figure recalculates in your browser.
What’s inside the Balanced scenario
These three numbers drive the whole model. They’re fixed values from SearchPilot’s field experience across enterprise ecommerce, travel and listings programmes — not sliders you can push until the answer looks good. You can nudge each one a little to stress-test the result.
Set by the scenario, not typed in. To change the pace, pick a different scenario.
Share of tests that produce a statistically significant positive result.
Lift in non-branded organic sessions to the tested page type, against the control group.
Why a losing test still pays
A 10–20% win rate sounds low until you look at what the other 80% do for you. Every test lands in one of four boxes, and three of them create value — which is why the model counts rescued revenue alongside wins, and why the win rate in your scenario isn’t the whole story. SearchPilot CEO Will Critchlow makes the full case for this in We’re Doing Business, Not Science.
Wins • counted
Pan for gold
Speculative changes nobody would sign off without proof. Most fail, but the winners are pure upside you’d never have captured. This is the “from wins” number.
Wins • counted
Insurance
Changes you were shipping regardless — a redesign, a migration, a brand requirement. The test proves the release didn’t cost you traffic, and quantifies what it gained.
Loses • cheap to learn
Safety net
Ideas that lose and were never going to ship anyway. No revenue rescued, but the roadmap gets cheaper: you stop paying dev time for tactics that don’t work on your site.
Loses • counted
Seatbelt
A change you’d have shipped that turns out to hurt. Caught in test, never released. This is the “rescued” number — driven by the surprising loss rate above.
Share of tests that turn out significantly negative on changes you’d have shipped anyway. Testing catches these before they cost you.
How much a shipped-but-harmful change would have depressed traffic to that page type.
Without a control group, how long before you’d notice and roll it back.
This is an estimate, not a quote. It is subtracted from the headline rather than hidden below it, so replace it with your own figure and the whole model moves with it.
The counterfactual, made visible
What happens to organic revenue on this page type if you test, versus if you don’t. Both lines carry the same depreciation assumption — the gap between them is where the headline number comes from.
Organic revenue run-rate: testing vs. doing nothing
Annualised monthly revenue from the tested page type.
Across all three scenarios — Conservative, Balanced and Aggressive — we assume you have at least two templated page types you can test on (for example PDPs and PLPs). That’s what the traffic figures below are sized around.
If you ship nothing, organic revenue doesn’t hold flat — it decays as competitors, SERP features and AI answers take share. Applied to both scenarios, so it never inflates the headline; it just means uplift compounds on a shrinking base. See how SearchPilot builds this counterfactual into a CFO-ready business case in The One-Page SEO Testing Plan.
Brand searches don’t move when you change a template, so they’re excluded from the base.
We can only run tests on templated pages that repeat across your site — think PLPs and PDPs. One-off pages like your homepage, About page or checkout don’t count here, however much traffic they get.
The most this page type is allowed to grow in a single year, however many tests you run. It stops the model from claiming an unbelievable jump — like traffic doubling in 12 months — just because you ran a lot of tests.
Year-by-year breakdown
| Year | Revenue from wins | Rescued revenue | Gross impact | Annualised increase | Platform cost est. | Net return | Cumulative ROI |
|---|
Gross impact is measured against the do-nothing baseline in the same year, so the depreciation assumption is already netted out of every figure. Annualised increase is that same gross impact expressed as a percentage of what this page type would have earned that year if you’d done nothing — i.e. how much bigger your testing revenue is than the do-nothing baseline. Cumulative ROI = cumulative net return ÷ cumulative platform cost.
- Base.
annual sessions x non-branded share x page-type share x revenue per session= revenue at risk and in play on the tested template. - Counterfactual. That base decays at the annual depreciation rate in both scenarios. Uplift is applied to the decayed base, never the original one.
- Scenario. Tests/month, win rate and average uplift come from the Conservative / Balanced / Aggressive preset you picked.
- Wins. Each month,
tests x win ratewinners are called and start earning one month later. Every live win then decays on the model you chose. - Diminishing returns. One template has finite improvable surface, so each successive win is worth less than the last — marginal value halves roughly every
12winning tests. This is what keeps test velocity from compounding without limit. - Rescued. Each month,
tests x surprising loss rateharmful changes are caught. Each avoids its loss for as long as it would otherwise have run, and the number running concurrently is capped. - Net. The Annualised Increase Cap backstops the total — a hard limit on how much this page type can grow in a year — then platform cost is subtracted month by month. The headline is Year-1 net, which is by definition the gap versus doing nothing.
Want this model built on your actual data?
Send us your site and we’ll replace every default on this page with real numbers: your non-branded traffic by template, your revenue per session, and the test velocity your traffic can actually support.