A take-home assignment without evaluation criteria is not a test.
It is a haunted Google Doc.
The company says, “This should only take two hours,” which is adorable because nobody who has ever said that has personally timed it while anxious, unemployed, and trying not to look like a person who owns sweatpants.
Then the assignment lands:
“Analyze our churn problem and propose a strategy.”
Due Monday. No rubric. No data dictionary. No definition of success. No promise of next steps. Just a little unpaid consulting grenade rolled under your weekend.
This is how good candidates lose leverage before the work even starts. Not because they lack skill. Because the candidate screening process quietly turns into a free consulting interview task with vibes for boundaries.
So before you open the spreadsheet, deck template, Figma file, code repo, support queue, or “quick prompt,” make them define done.
Here’s the tactical system.
The rule: no scorecard, no serious work
A real evaluation can answer four questions:
- What skill is being tested?
- How much time is expected?
- What will be judged?
- What happens after submission?
If they cannot answer those, the assignment is not “ambiguous like real work.” It is ambiguous like a raccoon in a conference room.
You are allowed to ask for the frame before donating your brain.
Not aggressively. Not dramatically. Not with a 900-word manifesto called “My Boundaries, My Throne.”
Just clearly.
Step 1: classify the assignment before you touch it
Do not start by estimating the work. Start by naming what kind of work they are asking for.
Use this quick classification:
| Assignment type | What it looks like | Risk level | Your move |
|---|---|---|---|
| Skill sample | Generic prompt, fake company, constrained deliverable | Low | Do it if timebox is sane |
| Work simulation | Realistic scenario, but not their live business problem | Medium | Ask for evaluation criteria |
| Live business audit | Their actual funnel, customers, roadmap, churn, pricing, backlog | High | Cap scope or request paid work trial |
| Production-ready output | They could use it Monday morning | Extremely spicy | Redline, replace, charge, or walk |
A product marketer asked to “draft messaging for our upcoming launch” is not doing a harmless writing sample. That is launch work wearing interview cologne.
A data analyst asked to “find insights in our real retention export” is not just showing SQL. They are being pointed at a business problem the team has not solved yet.
A customer success candidate asked to “create a renewal save plan for three at-risk accounts” is not doing a cute exercise. That is revenue triage with a bow on it.
Decision point
Ask yourself:
- Could they use this work without hiring me?
- Does it require company-specific strategy, not just role skill?
- Is the prompt vague enough that doing it well requires extra discovery?
- Is the requested output more valuable than the evaluation itself?
If two or more are yes, you need boundaries before effort.
Step 2: send the Define Done note
The goal is not to sound difficult. The goal is to make the hidden interview scorecard visible before it eats your weekend and burps out “stronger culture fit.”
Use this template:
Hi [Name],
Thanks for sending this over. Before I start, I want to make sure I’m aiming at the right target and keeping the exercise appropriately scoped.
Could you confirm:
1. The expected timebox for the assignment
2. The top 2–3 criteria the team will use to evaluate it
3. Whether you prefer depth in one area or a broader high-level pass
4. What the next step will be after submission, assuming the work meets the bar
I’m happy to complete a focused version that demonstrates my approach without turning it into unpaid project work.
Best,
[Your Name]
That last sentence is doing real work. It says: I am serious, useful, and not available as a free department.
If they respond with clear criteria, great. You now have a target.
If they respond with “just show us how you think,” congratulations, you have discovered the hiring equivalent of a fog machine with benefits.
Step 3: convert vague criteria into proof blocks
Once you have the criteria, turn each one into a proof block.
A proof block is a compact chunk of evidence that says:
- Here is what I noticed.
- Here is the judgment I used.
- Here is the action I recommend.
- Here is the tradeoff.
- Here is what I would measure next.
This works for take-homes, live working sessions, behavioral interview answers, and even a one-way video interview where the bot interview questions are dressed up as “tell us about your process.”
Example: the prompt says:
“Review our onboarding flow and identify improvements.”
Bad candidate trap:
19-slide teardown of every button, tooltip, and existential wound in the product.
Better proof block:
Observation: The first-session flow asks for setup before showing value.
Judgment: This likely increases drop-off because users have not yet seen the payoff.
Recommendation: Move one success moment earlier: let users complete a lightweight action before full configuration.
Tradeoff: This may reduce setup completeness at first login, so I’d monitor activation quality, not just completion rate.
Next metric: Compare first-session activation, day-7 retention, and setup completion across cohorts.
That is enough to show product judgment without handing over a full roadmap on a decorative platter.
Step 4: choose your scope cap
Now decide how much work you are willing to do.
Use this scope menu:
Option A: the 90-minute sample
Use when the assignment is reasonable but underdefined.
Deliver:
- 1 page
- 3 findings max
- 1 recommendation per finding
- clear assumptions
- “what I would do with more data” section
This is the sweet spot for most unpaid take-home assignment requests. It shows skill without turning your Sunday into a shareholder contribution.
Option B: the annotated outline
Use when the prompt is live company work.
Deliver:
- framework
- sample analysis
- decision logic
- questions you would ask
- no full implementation plan
This says: I can think. I am not uploading the entire goose.
Option C: the paid work trial request
Use when they ask for production-ready work, deep research, code, strategy, or a deliverable they can directly use.
Template:
This looks closer to a real work product than a screening exercise. I’m open to doing it, but I’d suggest treating it as a paid work trial with a defined timebox and evaluation criteria.
If unpaid, I can provide a shorter sample: a 1-page approach, 2–3 prioritized recommendations, and the assumptions behind them.
This gives them a clean choice: pay for depth or accept a sample.
Option D: the replacement live session
Use when the exercise is too broad or likely to trigger revisions.
Template:
Given the scope, would the team be open to a 45-minute live working session instead? I can walk through how I’d diagnose the problem, make tradeoffs, and prioritize next steps in real time.
That may give you stronger signal while keeping the process fair and contained.
Hiring teams love saying they need “signal.” Fine. Offer signal. Not a donated strategy deck.
Step 5: protect against revision creep
The first submission is rarely where the theft happens.
The theft happens when they say:
- “Could you just add one more scenario?”
- “Can you revise this based on feedback from the VP?”
- “Can you make it more specific to our enterprise segment?”
- “Can you mock up the implementation plan?”
That is how a two-hour exercise becomes a three-round unpaid sprint, and somehow you’re the one feeling weird for noticing.
Use this revision boundary before submitting:
I kept this to the agreed [timebox] and focused on [criteria]. I’m happy to discuss the work live and explain tradeoffs. If the team wants additional iterations or implementation-level detail, I’d suggest scoping that as a paid work trial.
Simple. Calm. Adult.
The system may be ridiculous, but you do not have to be its unpaid intern.
Step 6: submit with assumptions, not apologies
Candidates often undermine themselves at submission because the assignment was vague and they feel obligated to pre-apologize.
Don’t write:
“Sorry if this isn’t what you wanted.”
Write:
I made three assumptions to keep the work focused:
1. The goal is to evaluate my diagnostic approach, not receive a full implementation plan.
2. The intended timebox is approximately [X hours].
3. The highest-value signal is prioritization and tradeoff thinking.
Based on that, I focused on [scope]. In a real role, I would validate these recommendations with [data/stakeholders/customers] before implementation.
This is executive communication interview energy without the blazer cosplay.
You are showing judgment. You are also creating a paper trail in case someone later complains that you did not solve their whole business between dinner and laundry.
Step 7: ask for the decision path after submission
A take-home without a decision path is just content farming with nicer fonts.
After submitting, send:
Thanks again for the opportunity. Could you share when the team expects to review the assignment and what the next decision point is?
If helpful, I’m happy to walk through the reasoning in a short discussion rather than make additional unpaid revisions asynchronously.
This does two things:
- It makes ghosting slightly more expensive for them.
- It blocks the “quick revision” trap before it starts stretching.
If they go silent anyway, that tells you something. Not about your worth. About their process.
The red flags that mean “do not open the laptop”
Some assignments are not worth negotiating. They are warning labels with bullet points.
Walk carefully if you see:
- no timebox
- no evaluation criteria
- due over a weekend “because the team is moving fast”
- real company data with no context
- confidential work from your current employer requested as a sample
- more than one unpaid revision
- “we want to see how badly you want it” energy
- seven stakeholders and zero owner
- no salary range but a 12-slide deck request
- “strong culture fit” used as pressure to comply
That last one deserves a tiny gong.
Culture fit interview language gets weaponized when boundaries are treated as attitude problems. Wanting clarity before free labor does not mean you are hard to work with. It means you have encountered calendars before.
A compact decision tree
Use this before every take-home:
Is the work generic and timeboxed?
→ Yes: Do a focused sample.
→ No: Ask for criteria and timebox.
Did they provide criteria?
→ Yes: Build proof blocks against the criteria.
→ No: Offer a limited sample or live session.
Could they use the output directly?
→ Yes: Ask for paid work trial or reduce to an outline.
→ No: Proceed if the time cost fits your search priorities.
Do they request revisions?
→ One clarification: okay.
→ New scope: paid, live discussion, or decline.
Is the process producing decisions?
→ Yes: continue.
→ No: stop feeding the ritual.
If you are also dealing with an AI interview screen before or after the assignment, NoSweatKing can help decode the questions and shape answers in your own voice so your proof blocks survive the bot-readable answers portion of the circus.
Final quality-control pass before you send anything
Before submission, run this checklist.
Scope check
- Did I stay inside the agreed timebox?
- Did I state what I did not do?
- Did I avoid building production-ready work for free?
- Did I avoid solving the entire business problem?
Scorecard check
- Did I map the work to the stated evaluation criteria?
- If criteria were missing, did I state my assumptions?
- Did I show judgment, tradeoffs, and next steps?
- Did I include enough evidence without overbuilding?
Boundary check
- Did I make clear that revisions should happen live or paid?
- Did I ask for the next decision point?
- Did I keep a copy of the prompt, response, and deliverable?
- Would I feel okay if they used this without hiring me?
That last question is the truth serum.
If the answer is no, cut the deliverable down.
The point is not to be precious. The point is to be employed, not harvested.
There are good take-homes. A focused, fair work sample can help candidates who do not interview like TED Talk goblins. Some people are better at doing the job than narrating their greatness under fluorescent judgment. A sane assignment can reveal that.
But a sane assignment has edges.
It has a timebox. It has a rubric. It has a decision. It respects that you are a candidate, not a temporary department with no invoice button.
So the next time a company sends a vague “quick exercise,” do not immediately become a heroic unpaid consultant in pajama pants.
Make them define done.
Then decide how much of your talent they have earned access to.
