1. Tell me about a time when a customer requested a feature and you solved the underlying problem in a different way than they had requested; how did…
The full question
Tell me about a time when a customer requested a feature and you solved the underlying problem in a different way than they had requested; how did you decide whether the request belonged on the roadmap and manage expectations?
Model answer
- Situation — At a B2B infrastructure company, a customer operating a production data pipeline asked us to add a one-click button to replay an entire change stream after a sink outage. Their team had spent hours working out whether data was missing or merely delayed, and the request came from a technically influential account. A full replay seemed simple in the request, but it could create duplicate downstream records and add operational risk for every customer.
- Task — I owned understanding the actual recovery problem, deciding whether the request represented a broader product need, and giving the customer a credible path forward without promising a risky implementation.
- Action
- I started with a joint incident review rather than debating the proposed button. We mapped what the customer knew at each step and found that the main pain was uncertainty about progress and recovery position—not a need to replay all historical data.
- I checked support cases and spoke with other customers and our field team. The same confusion appeared elsewhere, but the requested “replay everything” behavior did not; it would also have required customers to handle duplicates safely.
- I worked with engineering and support to compare options. We prioritized clearer progress and checkpoint visibility, plus guidance for resuming safely, over a broad replay control. That addressed the common diagnosis gap while preserving explicit recovery semantics.
- I explained the decision directly to the customer: we agreed on the operational problem, but I described why the proposed mechanism could create more risk than it removed. I shared what we could deliver, what remained manual, and when we would review the outcome.
- I kept the broader replay capability as a discovery item, with evidence we’d need—such as repeated demand, clear semantics, and a safe idempotency story—before putting it on the roadmap.
- Result — The customer used the improved visibility and recovery guidance during a later sink interruption and was able to identify where processing had stopped without triggering a blind full replay. We also gave support a repeatable way to diagnose similar cases, while avoiding a feature that could silently duplicate data. I learned to separate customer urgency from the first solution proposed: take the pain seriously, validate how widely it occurs, and be transparent about both the trade-off and next step.