Growth Case Study Interview: A Worked Activation Problem

Growth cases are usually lost by candidates who are good at growth.

The pattern is consistent: they read the funnel, find the biggest drop-off, and start designing experiments against it. That's the correct instinct in most jobs and it's the wrong one here, because the case is almost always built so that the biggest drop is not the biggest opportunity.

Here's one worked end to end, with the numbers on the page. Try it before the walkthrough — the turn is about halfway down.

The brief

A B2B tool. 1,000 signups a week and flat revenue. Activation is the problem. What would you do?

Step	Rate
Signup → created first project	62%
Created project → invited a teammate	24%
Day-28 retention, users who invited someone	71%
Day-28 retention, users who didn't	24%

That's the whole brief. Most candidates have a plan within forty seconds, and that's the problem.

The first move: turn rates into people

Percentages hide magnitude. Before anything else, convert.

"So we keep about 216 of 1,000 a week. And the population that everyone's attention goes to — the inviters — is 149 people. The group nobody has mentioned is the 471 who created something and then worked alone."

Two numbers that weren't in the brief and change what the case is about. Doing this conversion out loud is worth a lot on its own.

The obvious answer, and why it's a trap

The invite step drops from 620 to 149 — a 76% loss, by far the biggest in the funnel. And inviters retain at 71% against 24%. Every arrow points at "get more people to invite a teammate."

Saying it out loud: "The obvious read is that inviting is the activation event and we should push it harder. Before I take that, I want to check one thing — whether inviting causes the retention or just identifies the people who were going to stay anyway. A solo consultant who signs up has no teammate to invite. If I push invites at them I'm optimising a step they can't complete, and the 71% is telling me about who they are rather than what they did."

This is the moment the case is scored. Correlation between an action and retention is the single most common trap in growth interviews, and it's built into almost every brief.

The segmentation that changes the answer

"So I'd split the 471 who didn't invite by what they did do."

There it is. The 24% average was hiding two completely different populations: a group retaining at 58% — not far off the inviters' 71% — and a group at 8%.

Saying it out loud: "This says there are two activation paths, not one. Collaborating is one. Connecting your own data is another, and it works nearly as well. The team has been treating invites as the activation event, which means every solo user has been pushed toward a path they may not have available, while the path that does work for them gets no attention at all."

Sizing the two levers

This is the part candidates skip, and it's what turns an insight into a recommendation. Compare in the same unit — retained users per week — not in percentage-point improvements, which aren't comparable.

"The second lever is worth roughly twice the first. And I'd argue it's also the easier one, because we'd be helping people do something they can already do, rather than persuading people to invite colleagues who may not exist. A six-point move on invites is a hard win; a fourteen-point move on a step that's currently unprompted is a product prompt away."

Then the honesty:

"Both numbers assume the relationship holds when we intervene, and it might not. The 146 who connect a data source today do it without being asked — they're self-selected, and the ones we prompt into it will be less motivated by definition. I'd expect the real number to land below 33. I'd still take it over the invite lever, and I'd want the first test sized to tell me how much of that 58% survives prompting."

The recommendation

Three parts, in the order someone can act on them.

What I'd do first. Make the data connection a prompted step for users who haven't invited anyone by day two. Not a blocking step, a prompt — because the 8% group may include people for whom neither path is right, and forcing them through a funnel they don't want produces activations that don't retain.

What I'd measure. Day-28 retention of prompted solo users against unprompted ones. Not connection rate — connection rate will go up, that's what prompts do. The question is whether prompted connections carry the same retention as voluntary ones, and if they don't, by how much.

What would make me stop. If prompted connectors retain at 20% rather than 58%, the action wasn't causing anything and we've learned that the whole framing is wrong — which is worth knowing in three weeks rather than after a quarter of building.

The whole answer, in four sentences

Case rounds don't end when you stop talking. They end when the interviewer repeats your answer to someone who wasn't in the room, and whatever survives that retelling is your actual score. So compress it yourself:

"We keep 216 of every 1,000 signups. The obvious lever is getting more people to invite a teammate — but 471 users a week work alone, many of them have nobody to invite, and the third of them who connect their own data instead retain at 58%. That unattended path is worth roughly twice the obvious one: +33 retained users a week against +18. I'd prompt it at day two, measure retention rather than connection rate, and drop it if prompted connectors come in at 20% instead of 58%."

Four sentences holding the sizing, the reframe, the comparison and the kill criterion. If your answer won't compress to something this length, it won't be repeated accurately — and an answer nobody can repeat loses to one they can.

Practising this

Take any funnel you've worked with and do two things to it. Convert every rate into people per week, and then find the largest step and ask who is structurally unable to complete it. Those two moves produce most of the insight in most growth cases, and both take about a minute.

Then practise the sizing sentence out loud — "lever A is worth roughly X, lever B roughly Y, and here's why I'd discount B" — because that's the sentence candidates know how to think and not how to say.

For the rest of the loop — experiment design, reading a result that came back flat, what you do with a loss — the question set is here: Growth Manager Interview Questions. And when the case is framed around a metric that moved rather than a funnel, the data-side version of this reasoning is worked through here: Data Scientist Interview Case Study.