Research take-homes are graded differently from the way candidates assume. The reviewer is not checking whether you know what a diary study is. They're checking whether you'd be safe to hand a vague question and a deadline to.
Which means the document they want is not an essay about methods. It's a plan someone could actually run on Monday, including the parts that are inconvenient.
Below is a realistic brief and a complete worked response — the whole document, not a description of one.
The brief
Our support team says customers are confused by our billing page. Leadership wants to know if that's true and what to do about it. You have two weeks and a small recruitment budget. Send us a study plan.
That's typical: a secondhand complaint, an unclear decision, and a deadline. Three things in it are traps.
"Support says." The signal is real but it's filtered — support hears from people who contacted support. Whatever is happening to people who didn't contact anyone is invisible in that channel, and it may be the bigger group.
"Confused." Not a research question. Confused about what — what they're being charged, when they're charged, how to change it, or whether they've been charged twice? Those are four studies.
"Is that true and what to do about it." Two different questions with different methods, and a plan that treats them as one will do both badly.
Handling those three in the opening paragraph is most of the grade.
The worked plan
Billing page confusion — study plan
What decision this feeds. Before scoping: I'd want to confirm with the requester whether this is heading towards a redesign of the billing page, a change to the billing model itself, or a support-deflection effort. I'm assuming redesign of the page and writing the plan for that. If it's the billing model, the study changes substantially and I'd want to know before starting — flagged here rather than assumed silently.
The question, narrowed. "Confused" isn't measurable. From the support tickets I'd first establish which of these it is:
Candidate problem How I'd confirm it cheaply Can't predict what they'll be charged Ticket tagging, one afternoon Can't find how to change plan Analytics: search terms on the page, rage clicks Believe they were charged incorrectly Refund and dispute rate Don't understand the invoice itself Tickets citing a specific line item One afternoon in the existing tickets replaces a week of guessing. I'd start there on day one and let it point the rest.
Method, and what I'm giving up.
- Ticket review (day 1–2) — free, already exists, tells me which problem to chase.
- Analytics on the billing page (day 2) — tells me scale. Qualitative can't count; this can.
- Six moderated sessions (days 4–8) — the mechanism. Recruited from customers who contacted support in the last month and customers who visited billing three or more times without contacting anyone. The second group is the one support can't see, and it's where I expect the surprise.
- Not running: a survey. With two weeks, a survey would either go out before I know what to ask or come back too late to matter. If leadership needs a prevalence number afterwards, the sessions tell me what to put in it.
Recruitment reality. Six sessions in that window means screening about 25 people, and the second group is harder to reach because they have no support relationship with us. I'd budget for incentives there and I'd accept four from that group rather than delay. If recruitment slips past day 6, I'd cut to four sessions rather than push the readout — a late finding is worth less than a smaller one.
What would make me change the plan. If ticket review shows the complaints are dominated by one specific invoice line, this stops being a page-design study and becomes a copy problem that can be fixed in a day. I'd say so and hand back the two weeks.
Schedule.
Days What 1–2 Ticket review, analytics pull, narrow the question 3 Screener out, discussion guide written 4–8 Six sessions, analysing as I go 9–10 Synthesis, readout written 11 Readout, with recommendations at three levels of effort What I'd deliver. One page of findings, a recommendation split into "fix this week / fix this quarter / needs a bigger decision", and the session clips that make the case. Not a sixty-slide deck.
The three sections candidates leave out
Reviewers look for these first, precisely because they're the ones that get skipped.
What would make you abandon the plan. A plan that can't fail isn't a plan, it's a schedule. The paragraph above — "if it's one invoice line, I hand back the two weeks" — is the single most senior thing in the document. It says you're optimising for the decision rather than for having run a study.
Recruitment as a real constraint. Junior plans say "recruit six participants" as though that were a calendar event. Experienced ones name the screening ratio, the group that will be hard to reach, and what they'd trade when it slips. Every reviewer has had a study slip on recruitment and they're scanning for whether you know it.
The thing you're not doing, and why. Naming the survey you ruled out is worth more than naming the sessions you're running. It shows the method was chosen against an alternative rather than reached for by habit.
The hidden grading criterion is whether your plan survives contact with a stakeholder. Notice how much of the document above is written to be read by someone who is not a researcher: a decision named at the top, a recommendation split by effort, a schedule with dates. A plan that only makes sense to another researcher signals someone who will need translating, and translating is most of the job.
Two failure modes
The methods essay
Three pages comparing contextual inquiry, diary studies, usability testing and card sorting, with a paragraph on the epistemology of each.
It reads as coursework. The reviewer wanted to know what you'd do on Monday, and you've told them what you know. Nothing in the document could be executed without a follow-up conversation.
The confident schedule
“Week 1: research. Week 2: synthesis and readout. Method: 6 user interviews.”
Executable and empty. No narrowing of the question, no recruitment reality, nothing that could go wrong. It's the answer of someone who hasn't yet had a study fall apart on them.
If they ask you to present it
Many take-homes are followed by a thirty-minute conversation about the document, and that conversation is the real assessment. Expect three pushes:
"Why not a survey?" — you already answered it in the plan, so say the same thing out loud, calmly. Consistency between the document and the conversation is itself being checked.
"What if you only get two participants from the second group?" — the honest answer is that the finding gets weaker in a specific way, and you should name which way: "Then I can't claim anything about the silent group as a group. I'd report it as a signal to follow, not a finding, and I'd say that in the readout rather than let it get quoted as if it were solid."
"Leadership wants a number. What do you give them?" — the trap is inventing one. "From this study, the analytics number — how many people visit billing repeatedly without resolving. That's real and it's the whole population. I wouldn't put a percentage on 'confused' from six sessions, and if that's what they need I'd scope a survey as a second step with a week of its own."
For the wider loop — the questions about stakeholder pushback, running the session itself, and what your data can't support — the list is here: UX Researcher Interview Questions. And when the take-home turns into presenting a past project instead, the shape of that is covered in presenting a UX research case study.
Writing yours
Two pages. Decision at the top, narrowing second, method with its rejected alternative third, recruitment reality fourth, and a paragraph naming what would make you stop.
Then read it back and ask one question: could someone who isn't a researcher act on this? If the answer is no, the plan is a methods essay wearing a schedule.
Practical target: write the "what would make me abandon this plan" paragraph first, before the method section. If you can't fill it in, you haven't understood the brief well enough to choose a method yet — and that's a much cheaper thing to discover on your own desk than in the follow-up conversation.
