How should I approach a take-home interview assignment with an unclear prompt?

First, translate the prompt into plain requirements without trying to fill in its gaps. Make three headings:

1. Stated: words or requirements the prompt actually gives you.
2. Unclear: choices it does not settle.
3. Assumptions: what you would choose if nobody clarifies.

Then test each unclear point: could a different interpretation change what you deliver, how much work you do, or what a reviewer could reasonably expect? If yes, consider asking. If not, a brief, sensible assumption may be enough. This prevents a vague feeling from turning into a long list of questions.

Assuming that is the full wording, your worksheet might look like this:

Stated: Review the supplied sample data; recommend improvements.

Unclear: Whether they want written recommendations, a presentation, or changes to the data itself; expected length; whether to focus on a particular audience or goal.

Assumptions: I would provide a concise written review with prioritized recommendations, and I would not alter the data unless asked.

Now separate consequential uncertainty from detail you can choose yourself. The output format could affect the whole deliverable, so it is a good clarification candidate. A particular heading style probably is not. The example is only a planning exercise; the actual prompt may imply different requirements.

Ask when the missing detail could cause you to produce the wrong kind of work or an unexpectedly large deliverable. Examples include whether they want a recommendation memo or an edited artifact, what question the work should answer, or whether there is a stated time or scope limit you cannot infer.

Proceed with a stated assumption when the choice is small, reversible, or not central to the task. For example, you can choose clear section headings without asking permission. Aim for one message that groups closely related uncertainties, rather than sending a separate question for every preference. If the prompt already answers something, do not ask it again.

Keep it neutral and specific: thank them, identify the ambiguity, say what you plan to do if that is their preference, and ask which option they intend. For the sample-data example, you could write:

“Thanks for sharing the assignment. To make sure I shape the deliverable as intended, would you prefer a concise written review with prioritized recommendations, or edits to the sample data itself? If there isn’t a preferred format, I’ll provide a concise written review and leave the data unchanged.”

That question names the decision that affects your work and offers a reasonable default. Remove any option that the prompt already rules out. Avoid asking them to define every detail of a successful answer; the assignment is still yours to interpret and complete.

Copy this into your notes and fill it in before you start:

Prompt-analysis worksheet

• Deliverable requested: [Use the prompt’s wording]
• Must-follow requirements: [Format, scope, audience, deadline, or other explicit details]
• Unclear choices: [Only points the prompt leaves open]
• Could this change the work substantially? [Yes/no, and why]
• My default assumption: [Smallest reasonable interpretation]
• Clarification to send, if needed: [One focused question]

Adapt this message with only relevant options:

“Thanks for sharing the assignment. Could you clarify [specific ambiguity]? I’m planning to [reasonable default] unless you prefer [alternative].”

Before sending, check that the question is not already answered, concerns a consequential choice, and makes your proposed default clear. If you do not get a reply before you need to start, use the explicit requirements as your anchor, choose the smallest reasonable interpretation, and note any important assumption in your deliverable. Keep your work within the stated scope and deadline. If the unresolved issue changes the core task so much that you cannot choose responsibly, make one focused follow-up through the appropriate channel. The worksheet helps explain what you understood; it does not guarantee that your interpretation matches an unstated preference.