Tell me about a time a customer demanded a feature that would not solve their real problem. What did you do?
Answer it out loud first, then compare. That's what makes it stick.
Situation: a support lead insisted we add a chatbot widget to cut ticket volume. Task: keep the relationship while solving the real issue. Action: I pulled 300 tickets, showed that 40 percent were password resets and order-status lookups, and proposed a two-day pilot of automated order-status replies instead. I framed it as a faster route to their goal, not a refusal. Result: ticket volume fell 28 percent in a month, and they later funded the wider chatbot with evidence.
- Restate their goal, not their solution
- Gather evidence from their own data
- Show the gap between the request and the goal
- Offer a small pilot that tests your idea cheaply
- Keep the door open for their original ask
A general prep source lists 'a customer demands a feature that will not solve their real problem' as a customer-case prompt, not tied to one company. It is also a classic FDE situation: a sponsor arrives with a solution already in mind. The scoring is about whether you can redirect without damaging trust.
The core move: serve the goal, not the request
- Ask what success looks like. 'If this feature shipped tomorrow, what number would move?' This surfaces the goal and often reveals that the request is a guess.
- Go to their data. Pull a sample of tickets, logs or transcripts. 300 items is usually enough to see the distribution. Evidence turns an opinion fight into a shared look at facts.
- Show the gap. 'Your goal is fewer tickets. 40 percent of tickets are two question types; a widget would not touch them.'
- Propose the smallest test. A two-day pilot of a narrower fix. A cheap experiment is far easier to accept than a no.
- Do not erase their idea. Park it with a trigger: 'If the pilot reaches X, we revisit the full chatbot.'
Language that works
- 'Here is the fastest way to hit your number.'
- 'I want to make sure we build the thing that moves it.'
- Avoid 'that will not work' and 'the customer is wrong'.
When the customer insists anyway
- Make the risk explicit in writing and agree what evidence would change the decision.
- If it is cheap and reversible, build it, but add a measurement so the result speaks for itself.
- If it is costly or harmful, escalate to your lead with a recommendation, not a complaint.
Numbers to state
- Sample size you reviewed (300 tickets)
- Share the request would have affected (for example 0 percent of the top issues)
- Result of the pilot (28 percent fewer tickets)
Pitfalls
- Winning the argument and losing the account.
- Building the wrong thing silently to keep the peace.
- Presenting it as a heroic override rather than a shared discovery.
A good closing line: 'They trusted my next recommendation faster because the first one came with their own data.'
- What if they still insisted after seeing the data?
- How did you decide the pilot was cheap enough to run?
- How did you word it so it did not sound like a refusal?
- Tell me about a time you built what they asked for and were right to.
- Forward Deployed Engineer Interview Questions (2026, by Round) (prep, 2026-10-16) Lists a customer-demands-a-feature-that-will-not-solve-the-problem scenario as not company-specific.
- Forward Deployed Engineer Interview: The Role and Case Round (prep, 2026-10-27) Advises stories about a request that hid a different problem and a user who pushed back, ending with what changed for users.