If you work in ecommerce growth or conversion rate optimisation, your experiments are your portfolio. A designer brings a case study; you bring tests. And in almost every growth, CRO or ecommerce marketing interview, there is a moment where the hiring manager leans in and says some version of: "Walk me through a test you ran."
It sounds like the easiest question in the room. You ran the test. You know the result. But this is where a surprising number of strong candidates lose the interview, not because their work was weak, but because of how they described it.
The problem is that the people interviewing you for a growth role have usually run hundreds of experiments themselves. They know what a fragile result looks like. They know how often a "+27% conversion lift" turns out to be a novelty effect, a tracking bug or a test that was stopped the day it hit significance. So they are not just listening to your result. They are listening to how you think about evidence.
Here is how to talk about your experiments so that an experienced interviewer comes away trusting your judgement, not just your numbers.
What the interviewer is actually scoring
When a hiring manager asks about a test, they are quietly checking four things:
- Did you start from a real problem? Was the test driven by data, user research or a clear hypothesis, or was it "we tried a green button because someone suggested it"?
- Do you understand the mechanics? Sample size, test duration, the primary metric, guardrail metrics, segments. Not at a statistician's level, but enough to know when a result can be trusted.
- Can you separate signal from noise? Do you know the difference between a winner and a lucky week?
- Did anything change because of it? Was the winner rolled out? Did the learning shape the next test? Did revenue actually move?
Notice that "how big was the lift" is not on the list. A big number with no reasoning behind it worries an experienced interviewer more than a small number with clean reasoning.
A structure that works: problem, hypothesis, design, result, decision
Most candidates tell experiment stories in chronological order and run out of time before the interesting part. A tighter structure, borrowed from how good experimentation teams write up their tests, keeps you on track:
- Problem. What was wrong, and how did you know? "Mobile visitors were adding to cart at roughly the same rate as desktop, but completing checkout at about half the rate."
- Hypothesis. What you believed and why. "We thought surprise shipping costs at the final step were causing drop-off, because exit surveys kept mentioning shipping."
- Design. What you changed, who saw it, what you measured. "We showed the shipping threshold on the product page for mobile visitors only, with checkout completion as the primary metric and average order value as a guardrail."
- Result. What happened, stated honestly, including how confident you were.
- Decision. What you did next, and what you learned that you did not expect.
That last step is the one most people skip, and it is often the most impressive. "We rolled it out, and then realised the same message was hurting average order value for returning customers, so we split it by segment" shows far more maturity than the original win.
Keep the whole story to about two minutes. The interviewer will ask follow-ups on whatever part interests them, and that is where the real conversation happens.
Say the uncomfortable part before they ask
The fastest way to earn trust in a growth interview is to mention the weaknesses of your own test before the interviewer has to dig for them.
Compare two answers to the same question.
Answer A: "We changed the add-to-cart button copy to 'Ships tonight' and conversion went up 27%."
Answer B: "We changed the add-to-cart copy to 'Ships tonight' and saw about a 27% lift in add-to-cart over two weeks. We were fairly confident in it, but two things made me cautious. The test ran during a promotion week, so traffic was unusually high-intent, and add-to-cart isn't revenue. So we kept it running a further two weeks outside the promotion and tracked completed orders. The lift held, but it settled closer to 10% on purchases, which is the number I'd actually stand behind."
Answer B has a smaller headline number and is far more convincing. It tells the interviewer that you understand seasonality, that you know the difference between a proxy metric and the business outcome, and that you will not walk into a leadership meeting with a number that collapses a month later.
A few honest caveats that interviewers respect:
- "The sample was smaller than I'd like, so I treated it as directional."
- "It won overall, but when we looked by device the effect was almost entirely on mobile."
- "We stopped it early because of a stock issue, so the result is inconclusive."
- "In hindsight, we tested two changes at once, so we couldn't say which one drove it."
None of these make you look weak. They make you look like someone who has run real experiments.
Have a losing test ready
Almost every senior growth interviewer will eventually ask about a test that failed. Some candidates treat this as a trap and offer a disguised win: "It failed, but we learned so much that it was really a success." Interviewers see through that immediately.
Instead, pick a test that genuinely lost or came back flat, and focus on what it taught you about your customers. Most experiments in mature programmes do not produce a clear winner, and people who have run many tests know it. A candidate who can talk calmly about a flat result signals experience. A candidate whose every test was a winner signals either luck, a very young programme or selective memory.
A good losing-test story sounds like this: "We were sure a countdown timer on the product page would create urgency. It lost on conversion and we saw an increase in support tickets asking whether the offer was real. That changed how we thought about urgency for that brand. Our customers were buying considered, higher-priced products, and pressure tactics made them trust us less. The next three tests we ran were about reassurance instead, and two of those won."
That answer shows hypothesis, evidence, a real customer insight and a change in strategy. It is a stronger story than most wins.
Know your numbers, and know their limits
You do not need to derive a p-value on a whiteboard, but you should be comfortable with the basics, because a sharp interviewer will test them in passing:
- Statistical significance and confidence. What it means in plain language, and why reaching significance on day three of a two-week test is not a reason to stop.
- Sample size and duration. Why you decide how long a test runs before you start it, and why tests should usually cover full weekly cycles.
- Primary versus guardrail metrics. Why a test that lifts conversion but lowers average order value or increases returns might not be a win.
- Segments. Why a flat overall result can hide a strong effect in one segment, and why slicing results after the fact can also produce false winners.
- Revenue per visitor. Why many teams prefer it to conversion rate alone when judging an ecommerce test.
If your tool used Bayesian statistics and reported a "probability to beat control", be ready to explain that in one sentence too. You are not being examined on maths. You are being checked for whether you would make a costly decision based on a misread dashboard.
Bring the business back into the story
Growth and CRO roles sit close to revenue, so the interviewer wants to know that you think beyond the test itself. Where you can, translate results into business terms: "That change was worth roughly this much a month in additional revenue at our traffic levels", or "It reduced checkout support tickets, which mattered to the operations team."
If you do not have exact revenue numbers, say so, and explain how you would have estimated them. An honest "I don't have the final revenue figure, but at our average order value and traffic it was in the range of a few thousand a month" is fine. A suspiciously precise number you cannot explain is not.
It also helps to show that you understand experiment prioritisation. Mention how you chose which tests to run, such as by expected impact, confidence and effort, and what you deliberately did not test. Hiring managers for growth roles are often hiring someone to decide where limited traffic and development time should go, not just someone to run whatever is next on the list.
Prepare for the follow-ups, not just the first answer
The first answer is the easy part. The interview is decided by the second and third questions:
- "How did you decide the test had run long enough?"
- "What would you have done if it had been flat?"
- "Who disagreed with this test, and why?"
- "How do you know the lift wasn't seasonal?"
- "What did you test next?"
The only reliable way to handle these is to rehearse your stories out loud and have someone push back on them. Write down your three or four strongest experiments, plus one honest failure, using the problem, hypothesis, design, result and decision structure. Then practise them with a colleague who will ask awkward questions, or use an AI mock interview tool such as InterviewAce to rehearse answers and follow-up questions based on your own resume and the job description. The goal is not to memorise a script. It is to know your own work so well that any follow-up feels like a normal conversation.
A quick checklist before your next growth interview
Before you walk in, make sure you can do each of these without notes:
- Tell one winning test in about two minutes, from problem to decision.
- Tell one losing or flat test, and what it changed about your thinking.
- Name the primary and guardrail metrics for each story.
- Explain how long each test ran and why.
- State at least one honest limitation of each result.
- Translate at least one result into revenue or another business outcome.
- Explain how you decided which tests to run first.
Growth interviews reward the same thing good experimentation programmes do: clear hypotheses, honest measurement and decisions that follow the evidence. If you talk about your tests the way you would want a colleague to report a result to you, the interviewer will hear exactly what they are hiring for.
Ready to Build Better Experiment Stories?
CustomFit.ai helps growth and CRO teams plan focused experiments, validate results and turn evidence into decisions stakeholders can trust.
Book a CRO strategy call
About the author: Iliya is the founder of InterviewAce, an AI interview coach that helps candidates practise with realistic, resume-based mock interviews and get real-time support during live interviews.
ab testing
ab testing
ab testing