How do you adapt contextual inquiry for internal API and tool design?

WHAT THIS TESTS: The interviewer wants to see if you can translate qualitative user research methods into a technical domain where the user is another engineer. Specifically, they are looking for your ability to adapt contextual inquiry, an ethnographic method built on observing users in their natural environment while they work, to the world of APIs, SDKs, and developer tools. The core signal is whether you can connect invisible workflow details, like habitual workarounds or trust issues, to concrete design decisions in API contracts and documentation architecture.

A GOOD ANSWER COVERS: First, describe adapting the four grounding principles of contextual inquiry to an internal engineering context. You would conduct the session in the engineer's natural environment, which means their IDE, terminal, documentation browser, or incident-response channel, not a conference room. Second, explain that you would observe them performing real tasks, such as integrating an existing API, debugging a failed call, or onboarding to a new service, while asking probing questions about why they chose a particular pattern. Third, show how you map findings to API design. If you observe engineers habitually wrapping your SDK in a custom abstraction layer, that signals your resource model does not match their mental model. If they repeatedly hit retry buttons or manually verify saves, that signals a trust gap you should close with clear idempotency keys and explicit confirmation responses. Fourth, connect insights to documentation structure. If engineers keep a personal cheat sheet of error codes, your reference docs should foreground troubleshooting workflows, not just list status codes.

COMMON WRONG ANSWERS: A major red flag is claiming that internal users are too similar to you or too technical to require generative research. Another weak pattern is proposing a survey or structured interview as a substitute for contextual observation, since the reference material explicitly states that users omit low-level details like reasoning and mental models when they are removed from the task. A third red flag is giving generic API advice, such as use REST or add pagination, without tying it back to a specific observed workflow or workaround.

LIKELY FOLLOW-UPS: The interviewer may ask how you would recruit participants without creating bias toward power users, how you would balance qualitative depth with the need to scale research across many engineering teams, or how you would validate that a redesigned API contract actually resolved the workflow friction you observed.

ONE CONCRETE EXAMPLE: Imagine you are designing a fleet-management API. During contextual inquiry, you watch an engineer onboard a new vehicle. They copy data from a spreadsheet into your API payload, but then switch to a second screen to cross-reference missing correlation fields, and they manually call GET after every POST to verify persistence. These invisible steps reveal three design needs: first, a bulk-upload endpoint to replace the spreadsheet dance; second, a richer request schema that accepts the correlating fields in one call; and third, a synchronous response or webhook that confirms persistence so they stop polling. The documentation should then lead with a bulk-upload quickstart and a trust section explaining durability guarantees.

Source: nngroup.com