Media Buyer Test Assignment Guide: How to Scope and Answer It

A fair media buyer test assignment evaluates a bounded capability with enough context to reason, a stated time expectation, and clear review criteria. It should not require you to fund or launch a live campaign, transfer a complete unpaid strategy for a real client, expose previous employers' data, or use personal advertising accounts.

Your goal is not to guess the hiring manager's secret answer. It is to show how you define the question, validate data, make assumptions visible, control risk, and turn evidence into a next action.

Before accepting the task

Ask five questions in writing:

1. Is the brief fictional, historical, anonymized, or currently live?
2. How much time is the candidate expected to spend?
3. Which capability and level are being evaluated?
4. Who will review it, and can feedback be shared?
5. Will any output be used commercially?

Also confirm the format, deadline, presentation time, and whether external research or AI assistance is allowed. Follow the stated rules. If AI can assist, remain responsible for every fact, calculation, and recommendation and disclose use when requested.

When to narrow or decline

Propose a smaller version if the task asks for a full account audit, finished creative batch, detailed market launch, or weeks of historical analysis. Offer a representative slice: one channel, one funnel issue, three creative directions, or a 60-minute analysis followed by discussion.

Decline if the company asks for passwords, a live launch, personal spend, deceptive policy evasion, confidential data from another employer, or identity documents unrelated to a verified hiring process. These are not tests of media-buying ability. Review the broader job-offer red flags.

Build an assumption register

Most case briefs omit important information. Do not silently fill the gaps. Create a table:

Unknown	Working assumption	Decision impact	What you would request
Conversion definition	Approved first purchase	Changes allowable CPA	Event dictionary and refund rule
Attribution	Seven-day click for exercise	Affects channel credit	Business source of truth
Creative approval	All supplied assets approved	Affects launch timing	Compliance owner
Capacity	No operational cap stated	Limits scaling confidence	Daily fulfillment limit

Prioritize assumptions that could reverse your recommendation. You do not need a page of minor caveats.

A reliable answer structure

1. Executive decision

Begin with your recommendation, confidence, and the next check.

I would not scale the campaign yet. The apparent CPA improvement is based on recorded leads, while approval is delayed and mobile tracking changed mid-period. I would hold spend, reconcile the event status, and run one controlled landing-page test after the measurement path is stable.

This gives the reviewer a thesis to test as they read.

2. Objective and metric contract

State the intended business result, primary metric, guardrails, attribution assumption, data maturity, and cost boundary. Distinguish platform conversion, tracker event, approved action, and financial value.

If the assignment only supplies clicks and conversions, do not invent profit. Explain what can be decided from those metrics and what cannot.

3. Data quality checks

Before optimization, inspect:

• date range, time zone, and currency;
• missing or duplicated rows;
• spend and event reconciliation;
• definition changes inside the period;
• extreme outliers and small samples;
• cohort maturity and conversion delay;
• status differences such as recorded versus approved;
• whether totals equal the sum of segments.

Document any correction. Never quietly delete inconvenient data.

4. Funnel diagnosis

Move from business value upstream: approved value, conversion, landing behavior, click, impression, and delivery. Calculate only metrics that help locate the constraint. The media buying metrics guide provides formulas and interpretation.

For each finding, separate observation and hypothesis:

• Observation: mobile landing conversion fell after 12 July.
• Hypothesis: a form release added friction.
• Alternative: mobile audience composition changed.
• Evidence needed: field errors, page version, and segment mix.

5. Prioritized action plan

Rank actions by expected information or value, confidence, effort, and downside. A simple matrix is enough:

Action	Why now	Risk	Success evidence
Validate mobile events	Current decision may use broken data	Low	Test event reconciles
Restore or split-test form	Funnel break is localized	Medium	Qualified rate improves
Expand audience	Premature before diagnosis	High	Defer

Avoid a list of 25 “optimizations” with no sequence.

6. Test design

Define audience, fixed variables, changed variable, budget or sample logic, duration or maturity condition, primary outcome, guardrails, stop condition, and decision branches. If sample size cannot be estimated from the supplied data, say what baseline and minimum effect you would need.

Do not claim statistical significance because one variant has a higher percentage. Explain uncertainty in plain language.

7. Creative component

If asked for creative ideas, provide original briefs, not copied ads. Each brief should include insight, message, support, format, opening, CTA, required disclosure, and the variable tested.

Example:

• Question: Does visible setup proof reduce uncertainty?
• Opening: Show the first product step, not a generic claim.
• Evidence: On-screen demonstration from approved product flow.
• Fixed: audience, offer, landing page.
• Guardrail: no unsupported time-saving promise.

Three distinct hypotheses are stronger than twenty superficial slogans.

8. Measurement and QA plan

Show how you would verify the tagged click, destination, event, value, currency, identifier, deduplication, status update, and report. Name the technical or analytics owner you would involve. Do not pretend the buyer should independently fix unknown backend code.

9. Risks and constraints

Include commercial, measurement, creative, policy, operational, and access risks relevant to the case. State mitigations and escalation owners. This is especially important in regulated verticals.

10. Next review

End with owner, evidence, and timing: “After the approved-conversion cohort through [date] matures, compare mobile form variants and decide whether to restore budget.” A plan without a next decision is a set of activities.

Handling a campaign dataset

Keep a calculation appendix with raw columns and formulas. Use consistent denominators. For example:

• CTR = clicks / impressions;
• CPC = spend / clicks;
• conversion rate = conversions / relevant visits or clicks, clearly stated;
• CPA = spend / defined conversions;
• ROAS = attributed revenue / spend, only if valid revenue is supplied.

Check weighted totals. Averaging daily conversion rates without weighting can mislead. Keep “zero” distinct from “missing,” and do not divide by zero silently.

Segment with a reason. Device matters when it changes the action; a tiny city slice may be noise. State minimum data or maturity needed before acting.

Handling a no-data strategy brief

If no dataset exists, do not fabricate a forecast. Provide a measurement-first launch plan:

1. business and customer assumption;
2. funnel and event map;
3. channel rationale;
4. three hypotheses;
5. creative and landing requirements;
6. QA and compliance owners;
7. staged budget with loss limit;
8. decision scenarios;
9. missing inputs and how to obtain them.

Ranges should be labeled illustrative, not market benchmarks.

Presentation format

For a 20-minute discussion, use roughly:

• 2 minutes: decision and assumptions;
• 5 minutes: data quality and diagnosis;
• 6 minutes: prioritized plan and test;
• 3 minutes: risks and measurement;
• 4 minutes: questions and alternatives.

Put formulas and extra charts in an appendix. Speak to the decision, not every cell.

Questions reviewers may ask

“Why did you not scale the best row?” Explain maturity, quality, sample, and marginal risk.
“What would change your mind?” Name specific evidence.
“What if tracking cannot be fixed this week?” Offer a lower-risk proxy and capped plan, or pause if no responsible decision is possible.
“Which action comes first?” Choose one and explain opportunity cost.
“What did you assume?” Open the register instead of improvising.

Submission checklist

1. The brief and time box are documented.
2. Your direct answer appears first.
3. Assumptions are explicit.
4. Formulas use defined denominators.
5. Facts and hypotheses are separated.
6. Recommendations are prioritized.
7. The test has guardrails and branches.
8. Claims and creative are compliant.
9. No confidential or personal data appears.
10. Files open without private access.
11. Simulation is labeled.
12. The next decision is clear.

A good assignment lets the employer see your judgment and lets you see theirs. Notice whether reviewers discuss tradeoffs, respect boundaries, and provide feedback. The process itself is evidence about the team's operating culture.