What Is a CRO Hypothesis? A Working Template
A CRO hypothesis names the change, the reason it should work, and the metric that proves it — anything looser is a guess wearing a hypothesis's clothes.
A CRO hypothesis is a written prediction with three parts: the change you will make, the reason you expect it to work, and the metric that will prove or disprove it. Writing it down before the test runs is what separates a test from a guess dressed up afterward as one.
It is the starting point for every test inside the A/B testing guide — nothing downstream of it, from sample size to the final read, means much if the hypothesis itself was vague.
What makes something a hypothesis instead of a guess?
The reason attached to it. "Moving the CTA higher will increase signups" is a guess. "Because most visitors do not scroll past the first screen on mobile, moving the CTA above the fold will increase signup completion rate" is a hypothesis — it names a mechanism, which means the test result can actually teach you something about visitor behavior, not just about one button's position.
That distinction matters more than it looks. A guess that happens to win tells you nothing you can reuse elsewhere. A hypothesis that wins — or loses — confirms or rules out a belief about why visitors behave the way they do, and that belief informs the next ten tests, not just this one.
Losing is where a real hypothesis earns its keep. A guess that fails just gets discarded, and the team moves to the next idea with nothing learned. A hypothesis that fails tells you the stated mechanism was wrong — visitors were not, in fact, missing the CTA because of scroll position — which rules out an entire category of future tests built on that same assumption and points the next hypothesis somewhere more promising.
What is the standard hypothesis template?
Most teams converge on a version of the same three-part structure, because it forces the same three decisions every time regardless of who is writing it.
State the change
Exactly what will differ between the control and the variant — one change, not a bundle of them, or the result will not tell you which part mattered.
State the reason
Why you expect this specific change to move behavior — a mechanism, not a hope. If you cannot state a reason, you do not have a hypothesis yet.
State the metric and the direction
The one number that decides the test, and which direction counts as success, decided before you see any data.
What separates a testable hypothesis from an untestable one?
Whether the result could come back "no." A hypothesis phrased so broadly that almost any outcome can be read as support is not testable in any useful sense, because it never risks being wrong.
Run a quick check before any hypothesis goes into the backlog: write down, in one sentence, what a losing result would look like. If you cannot picture a version of the test data that would count as a clear loss, the hypothesis is too vague to run yet, no matter how confident the team feels about the underlying idea.
| Too weak to test | Testable |
|---|---|
| "A shorter form will convert better" | "Cutting the signup form from 6 fields to 2 will raise completion rate" |
| "Users want more trust signals" | "Adding a specific review count near the CTA will raise click-through" |
| "The page feels slow" | "Reducing hero image size will raise mobile conversion, based on load-time data" |
| "This copy is more persuasive" | "Naming the price in the headline will raise scroll-to-CTA rate" |
How specific should the predicted outcome be?
Specific enough that two people reading the result afterward would agree on whether the hypothesis held. "Improve conversion" fails that test, because almost any change to almost any metric could be squinted into a win. "Raise free-trial signup completion rate, measured over the same traffic mix, by an amount large enough to be visible against normal week-to-week variation" passes it, because it names the metric, the comparison and roughly what would count.
This is the same discipline behind reading page speed and conversion rate correctly: a vague belief that "faster is better" is not a hypothesis until it names which page, which metric and roughly how much movement would count as proof.
What happens after you write the hypothesis?
It goes into a backlog, ranked, not straight into a test. Most teams have more hypotheses worth testing than traffic to test them with, so the write-up doubles as a prioritization tool: a hypothesis with a clear mechanism, a cheap build and a metric tied to revenue outranks one that is expensive to build and only loosely connected to what the business cares about.
The same structure carries a test's result once it finishes — CRO results covers how to read and report what a test actually found, which only works cleanly when the hypothesis it is measured against was written down clearly in the first place.
Questions people also ask
What is a CRO hypothesis, in one sentence?
A stated prediction that changing X will affect metric Y, for a named reason, checked against a specific result — written down before the test runs, not explained afterward.
What's wrong with just testing an idea without writing a hypothesis first?
Without a written prediction, any result can be read as a win after the fact. A hypothesis forces you to say what would count as success before you know the answer, which is the only thing that makes the later result meaningful.
Does every AB test need a full written hypothesis?
For anything that will inform a decision beyond one page, yes. A quick, low-stakes test on internal tooling can skip the formality; a test whose result will change the homepage or the checkout flow should not.
How specific does the predicted metric need to be?
Specific enough that two people looking at the result afterward would agree on whether it succeeded. "Improve conversion" is not specific; "raise signup completion rate by a measurable amount" is.
Sources
- A/B Testing — Optimizely (checked 22 September 2026)
- A/B Testing — Nielsen Norman Group (checked 22 September 2026)
- Statistical Significance — Optimizely (checked 22 September 2026)
Turn your test ideas into hypotheses worth running
Book a free 30-minute consultation with Ibrahem and Jeremy. We look at your site while we talk, and you leave with a prioritised plan.
Who wrote this
Ibrahem HelalCo-founder
Leads strategy and delivery.
Works on: SEO, GEO
Reviewed before publishing by Jeremy Zava, Co-founder.
- A/B Testing: The Complete GuideWhat A/B testing is, when your traffic can support it, how to run a split test that holds up, and how to read the result without fooling yourself.
- SaaS Signup Optimization: Cut Trial Drop-OffSaaS signup optimization means shortening the form, deferring the credit card ask, and testing one change at a time instead of redesigning the whole flow.
- Page Speed and Conversion Rate, ExplainedPage speed and conversion rate move together because every added second gives a visitor one more reason to leave before they buy, sign up or fill in a form.
- How to Read and Report CRO ResultsCRO results mean nothing without the hypothesis, the sample size and the confidence level attached — here is how to read them honestly and report them well.