A growth engineering interview should test whether the candidate can turn a product hypothesis into a measurable, maintainable change. Use a fictional scenario and evaluate the plan, not a promised uplift.
A short assessment cannot establish that someone will produce a particular growth result. It can reveal how they reason about implementation and evidence.
Provide a fictional activation problem
Describe a fictional collaboration app where new users start setup but often leave before completing a useful action. Provide a simple flow diagram and synthetic event definitions, including one ambiguous event.
Ask the candidate to identify what they would investigate before changing the flow. Do not give access to real user records or request work on a live commercial problem.
Request a bounded proposal
The output should include:
- A hypothesis about the friction.
- A small product change that could test it.
- The event definitions needed to interpret behavior.
- Quality checks and rollout considerations.
- A decision rule for keeping, revising, or removing the change.
If coding is included, constrain it to a small component or event-handling example. Explain the time boundary and permitted tools.
Use an evidence rubric
| Dimension | What to look for |
|---|---|
| Problem framing | Distinguishes observed behavior from its possible causes |
| Measurement | Checks event meaning and instrumentation before interpretation |
| Implementation | Chooses a small change with clear ownership |
| User experience | Avoids misleading prompts or friction hidden elsewhere |
| Interpretation | Recognizes inconclusive evidence and alternative explanations |
| Maintenance | Plans cleanup and ownership after the experiment |
Do not reward a confident forecast unsupported by the fictional packet.
Introduce a measurement failure
Tell the candidate that an event fires more than once for some users. Ask how this changes the analysis and what they would inspect.
The useful signal is whether they revisit the evidence before declaring success. They should be able to explain the issue to a nontechnical stakeholder.
Discuss production responsibility
Ask how the change would interact with accessibility, privacy, reliability, and the product’s broader design. A short-lived experiment can still create lasting problems.
Use the growth engineer versus marketer guide to set scope and the bounded work-sample guide to keep the assessment proportionate. Discuss an engineering search with Refery.