Hypothesis statement examples (with an epic hypothesis template)
Short answer
A hypothesis statement is one or two sentences that say what will change, for whom, what result is expected, why, and which metric will prove it right or wrong. For a single test or feature, a simple form works: Because [evidence], changing [what] for [who] will [move metric], measured by [metric and success rule]. For a large initiative, the SAFe epic hypothesis statement adds an elevator pitch (For, who, the, is a, that, unlike, our solution), business outcomes and leading indicators.
Key takeaways
- A hypothesis statement is testable: someone else can read it and agree in advance what result would prove it wrong.
- Every good statement has the same core: evidence, change, audience, expected result and the metric that decides it.
- The epic hypothesis statement is SAFe’s version for big initiatives. It pairs an elevator pitch with business outcomes and leading indicators.
- Pick the metric that decides the statement before any work starts, and match it to the size of the bet.
- A weak statement teaches nothing when it fails. A strong one tells you what to try next.
What is a hypothesis statement?
A hypothesis statement is a short, testable claim about what will happen if you make a specific change, and how you will know. In product, CRO and growth work, it is the sentence written before anything is designed or built. It turns “let’s try a new onboarding” into something a team can measure, argue about and learn from.
Optimizely’s guide to hypotheses says the difference between a well-formed hypothesis and a guess is data, and that a strong one has three parts: the problem, the proposed solution and the result. I add two more parts on top of that: who the change is for, and the metric that decides it. Without those two, people can read the same result and walk away with opposite conclusions.
The same idea shows up at every level of work:
- A/B test hypothesis: one change on one page, judged by one primary metric. I cover this one in depth in my A/B test hypothesis template.
- Product or feature hypothesis: a new feature or flow, judged by adoption, retention or revenue. See the sibling post, product hypothesis template.
- Epic hypothesis statement: a large initiative that spans many teams and months, judged by business outcomes and early leading indicators. This is the SAFe format explained below.
What goes into a good hypothesis statement?
Five parts: evidence, change, audience, expected result and the metric that decides it. The order can change. The parts can’t be missing.
- Evidence. What you saw that makes you think the problem is real: funnel data, recordings, survey answers, support tickets, sales calls.
- Change. Exactly what will be different, specific enough to build.
- Audience. Who will see it: new visitors, mobile users, trial accounts, lapsed customers.
- Expected result. The direction of the change in behavior, in plain words.
- Deciding metric. One primary metric, plus any guardrail that must not get worse, and the rule for calling it.
What is an epic hypothesis statement?
An epic hypothesis statement is the template the Scaled Agile Framework (SAFe) uses to describe an epic, a large initiative, as a hypothesis instead of a fixed plan. SAFe defines an epic as a significant solution development initiative, and says each epic needs a minimum viable product (MVP), a Lean business case and a cost estimate. The epic hypothesis statement is where the epic is first written down so that stakeholders can agree on what it is and how its success will be judged.
A note on sources: the full SAFe article on epics now sits behind a login. What I describe here comes from the parts of Scaled Agile’s site I could read, including the public text of the epic hypothesis statement template figure. The template has these fields:
- Funnel entry date, epic name and epic owner. The admin header: when the epic entered the portfolio funnel, a short name, and who owns it.
- Epic description. An elevator pitch, or value statement, written in this pattern: For [customers] who [do something], the [solution] is a [something: the “how”] that [provides this value]. Unlike [competitor, current solution or no existing solution], our solution [does something better: the “why”].
- Business outcomes. The measurable benefits the business can expect if the epic hypothesis proves correct.
- Leading indicators. The early measures that help predict the business outcomes, so the team doesn’t have to wait months to learn whether it is on track.
- Nonfunctional requirements (NFRs). Requirements such as performance, security or compliance that come with the epic.
Scaled Agile’s older public version of the epic article explains how the pieces connect. The analysis of an epic produces a Lean business case. The epic includes an MVP, described there as an early and minimal version of the solution used to prove or disprove the epic hypothesis. Teams then run a build, measure, learn cycle. If the hypothesis proves true, the epic continues. If it proves false, the owner can pivot by creating a new epic or drop it.
Here is an epic hypothesis statement for a made-up online store, written in the SAFe pattern. The elevator pitch uses the template’s own field names, so “our solution” is the store team’s voice, not mine.
Example
Epic name: Guided gift finder.
Epic description: For shoppers buying a gift for someone else, who don’t know which product to pick, the guided gift finder is a short quiz on the home page and category pages that recommends three products with a reason for each. Unlike the current category filters, which assume the shopper knows the product type, our solution starts from the person receiving the gift.
Business outcomes: Say the business expects revenue per visitor from gift shoppers to rise by 5% and returns from gift orders to fall, within two quarters of launch.
Leading indicators: Say the team watches quiz start rate, quiz completion rate, clicks from the recommendations to product pages, and add-to-cart rate from those pages in the first few weeks.
NFRs: The quiz must not slow the home page, must work with a keyboard and a screen reader, and must not store personal data about the gift recipient.
The useful part of this format, even outside SAFe, is the split between business outcomes and leading indicators. A big initiative can take months to move revenue. Leading indicators let the team see early whether the hypothesis is heading the right way.
What do hypothesis statements look like in practice? 10 examples
Here are ten hypothesis statement examples across A/B tests, features, onboarding, email, growth and epics. They show the format, not results from real tests. Swap in your own evidence.
| Type | Example statement | Metric that decides it |
|---|---|---|
| A/B test: landing page | Because the ad promises free delivery and the landing page opens with a brand slogan, repeating the ad’s promise in the headline for visitors from that ad will raise purchases. | Revenue per visitor, with refund rate as a guardrail |
| A/B test: checkout | Because the biggest drop happens at the step where shipping cost first appears, showing shipping cost on the cart page for all visitors will raise completed orders. | Revenue per visitor, with average order value as a guardrail |
| A/B test: pricing page | Because sales calls keep hearing “I don’t know which plan to pick”, marking one plan as the recommended choice and saying who it is for will raise paid sign-ups from the pricing page. | Paid sign-ups per pricing page visitor, with downgrade rate as a guardrail |
| Product feature | Because support tickets show users exporting data to a spreadsheet every week to share it, adding a shareable report link for account admins will raise weekly active use. | Share of admins who create a report link, then weekly active accounts |
| Product feature | Because recordings show users searching for past orders by scrolling, adding search to order history for returning customers will raise reorders. | Reorder rate among returning customers |
| Onboarding | Because most new trial users never connect their data in the first session, replacing the product tour with one guided setup step will raise the share of trials that reach first value. | Share of trials that complete setup in the first session |
| Because cart abandoners who come back mostly return within a day, sending the first reminder sooner, with the product image and price, will raise recovered carts. | Recovered revenue per abandoned cart, with unsubscribe rate as a guardrail | |
| Growth experiment | Because happy customers mention the product in reviews but rarely share it, asking for a referral right after a positive review, with a reward for both sides, will raise referred sign-ups. | Referred sign-ups per active customer |
| Growth experiment | Because most search visitors leave the blog without seeing the product, adding a relevant free tool to the top posts will raise sign-ups from organic search. | Sign-ups from organic search per blog visitor |
| Epic level | For gift shoppers who don’t know what to buy, the guided gift finder is a short quiz that recommends products with reasons. Unlike category filters, our solution starts from the person receiving the gift. | Business outcome: revenue per visitor from gift shoppers. Leading indicators: quiz completion and add-to-cart from recommendations |
Notice the pattern. The smaller the change, the closer the metric sits to money and the shorter the wait. The bigger the bet, the more the statement needs leading indicators to show progress before the final outcome arrives. For more landing page ideas to plug in, see A/B test ideas for landing pages.
What makes a hypothesis statement good or weak?
A good statement is specific, based on evidence and can be proven wrong. A weak one is vague, based on taste, or impossible to fail.
| Part | Weak | Good |
|---|---|---|
| Evidence | “Everyone on the team thinks onboarding is too long.” | “Funnel data shows most new trials drop off during the product tour.” |
| Change | “Improve onboarding.” | “Replace the product tour with one guided setup step.” |
| Audience | “Users.” | “New trial accounts on desktop.” |
| Metric | “Engagement will improve.” | “Share of trials that complete setup in the first session.” |
| Success rule | “Let’s see how it goes.” | A minimum effect, a guardrail and a sample or time limit, agreed before launch |
| If it fails | Nothing learned, the team moves on | The problem stays on the list and the next idea attacks it differently |
The last row is the real test. If a failed result would teach the team nothing, the statement was too vague to begin with.
What is a hypothesis statement template I can copy?
Use one sentence for the idea and one for how it will be judged. This works for tests, features and growth experiments. For epics, wrap it in the SAFe fields above.
Copy this template
Because [evidence: data, recordings, surveys, tickets or calls],
changing [what] to [new version]
for [audience]
will [raise or lower] [primary metric].
It is right if [primary metric] moves by at least [minimum effect] and [guardrail] does not get worse, within [sample size or time limit].
Say a trial product sees most new users skip data setup. Filled in: Because most new trials never connect their data in the first session, changing the product tour to one guided setup step for new desktop trials will raise first-session setup. It is right if first-session setup rises by at least 15% and trial-to-paid rate does not fall, within four weeks.
How do you turn a hypothesis statement into a test?
Rank it, size it, run it, and read it on the rules you wrote down.
- Rank it against other ideas. A simple score for impact, confidence and ease keeps the strongest statements at the top. See ICE prioritization.
- Size it. Work out how many visitors or users it needs to detect the minimum effect. See A/B test sample size.
- Pick the cheapest way to learn. Not every statement needs a full A/B test. A product idea can start with a fake door or a prototype, which I cover in the product hypothesis template.
- Read it once, at the planned end. See when to call an A/B test, and what to do when a test is flat.
Across six years and more than 300 A/B tests designed, the statements that paid off were rarely the cleverest ideas. They were the ones with clear evidence behind them and a metric everyone agreed on before launch. If you want help turning research into hypotheses and test designs, see A/B testing or book a call.
FAQ
What is an example of a hypothesis statement?
Because the biggest checkout drop happens where shipping cost first appears, showing shipping cost on the cart page for all visitors will raise completed orders. It is right if revenue per visitor rises by the planned minimum effect and average order value does not fall.
What is an epic hypothesis statement in SAFe?
It is SAFe’s template for describing a large initiative as a hypothesis. It includes an elevator pitch in the pattern For, who, the, is a, that, unlike, our solution, plus business outcomes, leading indicators and nonfunctional requirements. It feeds the Lean business case and the MVP.
What is the difference between business outcomes and leading indicators?
Business outcomes are the measurable benefits expected if the hypothesis is right, such as revenue or retention. Leading indicators are earlier signals that predict those outcomes, such as adoption or completion rates, so the team can see progress before the outcome arrives.
How is a hypothesis statement different from an A/B test hypothesis?
An A/B test hypothesis is one kind of hypothesis statement, for a single change judged by one primary metric in a controlled test. Hypothesis statements also cover features, onboarding, email, growth experiments and large epics, which may be tested in other ways.
Sources
I’m George (Djordje) Peric, a CRO designer, CXL certified in Conversion Optimization and Ecommerce Marketing. I design websites, landing pages, funnels and ads, and A/B test them.