How do you build and prioritise a growth experimentation backlog?

The backlog starts with understanding the growth model: which metric is the north star, what are the inputs that move it, and where in the funnel are the biggest drop-offs? I map the funnel from awareness through activation, retention, and revenue, then use data to identify the stages with the largest improvement opportunity. For each opportunity I generate hypotheses with the team, then score them using ICE (Impact, Confidence, Ease) or PIE (Potential, Importance, Ease), depending on what the team already uses. The output is a stack-ranked list of experiments where each item has a clear hypothesis, a measurable success metric, a minimum detectable effect, and a required sample size. I run high-confidence, easy-to-implement experiments first to generate quick wins and build stakeholder trust, while also scheduling at least one high-impact, lower-confidence experiment every quarter to learn from. I review and re-prioritise the backlog every two weeks, because both the data and the business context change fast enough that a static backlog becomes stale quickly.

How do you design a statistically valid A/B test?

A valid A/B test requires four things to be decided before you start: the primary metric you are testing, the minimum detectable effect you care about, the required sample size based on current baseline and desired effect size, and the pre-determined test duration. I use a sample size calculator to determine how many users I need in each variant to achieve 80% statistical power at a 95% confidence level. I then calculate how long the test needs to run to reach that sample size based on current traffic to the test surface. I never stop a test early based on interim results, because of the peeking problem: early significance that disappears when you reach full sample size leads to false conclusions. I also check for novelty effects on tests involving significant UI changes by extending the test long enough to capture user behaviour after the initial surprise wears off. After the test concludes, I look at the primary metric, any secondary metrics that could reveal unintended effects, and segment the results by user cohort to understand whether the effect was uniform.

How do you think about user activation and what makes an activation model effective?

Activation is the moment when a new user first experiences the core value of the product, and defining it precisely is one of the most important things a growth team can do. A good activation metric is behaviourally specific (a concrete action, not a feeling), predictive of long-term retention, and achievable by a meaningful proportion of new users. To define it, I look at cohort retention data: which early actions in the first session or first week are most strongly correlated with users who are still active 30 or 90 days later? That action or sequence of actions becomes the activation event. Once I have the activation event defined, I look at the funnel leading to it: what proportion of new users reach it, where do users drop off, and what does the median time to activation look like? I then run experiments that reduce friction on the critical path and eliminate steps that do not contribute to value realisation.

How do you approach retention analysis and what interventions work best for improving it?

Retention analysis starts with cohort curves: I look at the percentage of users from each acquisition cohort who are still active at 7, 30, 60, and 90 days. The shape of the retention curve tells me whether I have a retention problem at all and where in the lifecycle it occurs. I then segment retention curves by acquisition channel, user persona, onboarding path, and feature usage to identify which segments retain well and which do not. The goal is to understand what retained users do differently in their first week, then find ways to get more new users to do those things. Common interventions include onboarding improvements that drive users to the activation event faster, in-product nudges during the critical first 30 days, email and push campaigns triggered by usage patterns rather than fixed time intervals, and habit-forming features like streaks or weekly summaries. I track retention improvements at the cohort level, not just in aggregate, because aggregate retention can improve through composition effects (better acquisition mix) even when per-cohort retention is flat.

Tell me about a growth experiment you ran that did not achieve the expected result. What did you do?

I ran an experiment testing a simplified onboarding flow for a B2C app. The hypothesis was that reducing the number of setup steps from seven to three would increase activation rate. The experiment reached statistical significance after three weeks and showed no meaningful difference in activation rate between the control and variant. Before calling it a failure, I dug into the data. The segment-level analysis revealed something interesting: users who signed up on mobile showed a 12% improvement in activation on the shorter flow, while desktop users showed a slight decline. The null result was masking a meaningful mobile win. I followed up with a mobile-only deployment of the shorter flow while keeping the longer flow on desktop, and activation on mobile improved by 9% in the following cohort. The lesson was that experiment analysis should always include segmentation, because aggregate results can hide meaningful signals in subgroups.

Describe a time you had to influence a product or engineering team to prioritise a growth initiative.

I wanted to add a referral programme to our product, which required three weeks of engineering time to build. The engineering team had a full backlog of feature work and the referral programme was not on their roadmap. I prepared a business case that modelled the expected impact: using data from comparable products in our category, I estimated a k-factor of 0.15, meaning every 100 users would bring in 15 additional users from referrals. At our current growth rate, that would compound to roughly 800 additional users per month within six months, worth approximately $24,000 in monthly recurring revenue at our average contract value. I proposed a phased approach: a lightweight email referral mechanism first (one week of engineering time) to test the hypothesis before building the full in-product experience. The engineering team agreed to the lightweight version. It generated a k-factor of 0.11, which was enough to justify the full build.

Give an example of a growth loop you identified and built into a product.

At a document collaboration tool I worked on, I identified a viral loop that was underexploited. When a user shared a document with a non-user, the non-user received an email with a preview and a prompt to sign up to edit. The conversion rate from that email to a signed-up user was 18%, but the activation rate for those users was only 31%, far below our overall activation rate of 54%. The issue was that sign-up from the share email dropped the new user on a generic onboarding flow rather than taking them directly to the document they had been invited to edit. I redesigned the post-signup redirect to send new users directly to the invited document, pre-filling their profile from the invitation context. Activation rate for share-invitee signups rose to 61%, and 30-day retention for these users was 22 percentage points higher than average.

How do you instrument a product for growth analytics?

Good instrumentation starts with an event taxonomy that maps to the growth model, not just to what is technically easy to track. Before writing a single tracking call, I define the events that represent meaningful user actions at each stage of the funnel: account creation, first key action, feature usage, sharing actions, and any event that serves as a leading indicator of retention or revenue. I use a structured event schema with consistent naming conventions so that queries are predictable and new events follow the same pattern. I implement tracking at the server side where possible, particularly for revenue events and critical conversion events, because client-side tracking can be blocked by ad blockers and creates data consistency issues. For A/B testing, I ensure that variant assignment is logged as a property on subsequent events, so I can segment any downstream metric by experiment variant without a separate attribution join. I also build data quality checks that alert on sudden drops or spikes in event volume.

How do you use SQL to do growth analysis? Walk me through a query you would write.

SQL is the foundational tool for growth analysis because it lets me ask questions directly against the event data without waiting for a dashboard to be built. A query I write frequently is a cohort retention analysis. I start by identifying the cohort: all users who completed their first session in a given week. I then join that cohort to their subsequent activity events to find which users were active in each subsequent week. The query structure uses a CTE to define the cohort with their first activity date, then a LEFT JOIN to the events table to find activity in subsequent periods, grouped by weeks since first activity. I calculate retention as the count of active users in week N divided by the total cohort size. I also regularly write funnel queries: for each step in the funnel, count the distinct users who completed that step, then calculate the step-to-step conversion rates. I use query results to identify hypotheses and validate experiment results, not to make decisions directly, because SQL queries on event data can have subtle counting issues that need to be understood before acting on a number.

How do you measure the impact of growth initiatives on revenue, not just on engagement metrics?

Connecting growth metrics to revenue requires two things: a revenue model that links user behaviour to monetisation, and the discipline to track experiments against revenue-linked metrics, not just engagement proxies. For a subscription product, I track the full chain from activation rate through free-to-paid conversion rate, initial plan selected, average revenue per paying user, and 12-month retention rate. An improvement in activation rate is only valuable if it converts into paying users at a similar or better rate: sometimes an experiment improves activation by attracting less qualified users who never pay. I build a revenue attribution model that assigns expected lifetime value to each cohort at activation based on historical conversion and retention curves, which lets me estimate the revenue impact of an experiment without waiting 12 months for actual revenue to materialise. I present growth results to senior stakeholders in revenue terms wherever possible, because engagement metrics can be gamed and do not translate directly to business outcomes.

Growth Manager

Growth Manager interviews are fundamentally about experimentation: can you design a test that'll actually answer the question you're asking, interpret the results without fooling yourself, and ship changes that move metrics in a way you can attribute? Interviewers are looking for analytical rigour, product intuition, and the ability to work across engineering, data, and design without having direct authority over any of them. This guide covers the questions that come up most often and what the better answers actually look like.

For general interview preparation tips, read our guide to common interview questions.

Prepare further

How to ace the behavioural interview (STAR method)How to answer competency-based interview questions

Common Growth Manager Interview Questions

Behavioural Interview Questions for Growth Manager Roles

Technical Questions for Growth Manager Candidates

What Hiring Managers Look for in Growth Manager Interviews

What hiring managers really look for in Growth Manager candidates:

Real statistical literacy. Growth is built on experiments, and candidates who can't explain statistical power, minimum detectable effect, and the peeking problem will run experiments that produce confident but wrong conclusions. This is worth testing directly in the interview rather than taking on faith.
Full-funnel thinking. Candidates who can only talk about acquisition or only about retention are optimising a slice of the problem, and the best growth managers understand the whole system and know which lever is worth pulling at any given stage.
A clear line to revenue. Engagement metrics without any connection to business outcomes are a warning sign, and it's worth asking candidates explicitly how they connect their work to things the business actually cares about, not just product metrics.
A track record of shipping things through other teams. Growth depends on engineering, product, design, and data, and candidates who've only run experiments in isolation haven't done the harder part of the job. Look for specific examples of getting changes shipped in collaboration with teams they didn't manage.
Intellectual honesty about experiments that didn't work. Growth managers who run a lot of experiments will have a lot of failures, and candidates who only describe successes are either not running enough tests or not learning from the ones that don't pan out.

Questions to Ask Your Interviewer

→What is the current north star metric for growth and how has it evolved over the past 12 months?
→How mature is the experimentation infrastructure and what is the typical time from experiment idea to a test running in production?
→How does the growth team work with the product and engineering teams and who owns the roadmap for growth-related changes?
→What are the biggest retention challenges right now and what has been tried so far?
→How is growth performance reported to senior leadership and what is the primary metric the business holds this team accountable to?

Practise These Questions Before Your Interview

The mock interview tool builds a practice session around a specific job posting and your background, so you rehearse the questions most likely to come up.

Start Practising

Free to start. No commitment.

Related Roles

SEO Manager

Business10 questions

Performance Marketing Manager

Business10 questions

Product Manager

Tech10 questions