mbfinotti/advertising-skills / skills/advertising-hiring
advertising-hiring
VerifiedBuild the complete hiring packet for a paid-advertising role - the rung decision (media buyer vs performance marketer vs growth marketer; creative strategist is its own role), an outcome scorecard, a 2-3 round interview loop with a judgment-first question bank, a time-boxed in-interview work sample, and a 30-60-90 ramp with staged spend authority. For the marketing leader or founder recruiting. Use for "hire a media buyer", "interview questions for a performance marketer", "paid media work sample", "scorecard for our first ad hire", or "30-60-90 for a new paid media hire". Does not source candidates, post jobs, run applicant tracking, or write offers, and does not cover the candidate side - that is the sibling mbfinotti/advertising-skills@advertising-career.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 25.1 KB |
| evals/evals.json | config | 38.4 KB |
| references/question-bank-and-scorecard.md | markdown | 8.6 KB |
| references/ramp-and-spend-authority.md | markdown | 5.1 KB |
| references/role-ladder-and-leveling.md | markdown | 13.9 KB |
| references/work-sample-design.md | markdown | 8.3 KB |
| references/worked-examples.md | markdown | 3.7 KB |
references/work-sample-design.md
View on GitHub# Work Sample Design
The richest first-hand source is the PPCChat "Hiring & Getting Hired in PPC" roundup (April 19, 2022), which devoted a full question to test projects. Its verdict is unusually one-sided.
## Unpaid take-homes actively damage the funnel
Anu Adegbola's summary of the community position: "test projects during the interviews are mostly hated by the community."
- Sarah Stemen: "At this point in my career I have a NO HOMEWORK policy."
- Julia Vyse (senior media buyer): "I hate this. It's straight-up time-wasting, unpaid labor... it should not have a role in hiring."
- Melissa Mackey: "doing free work for someone who may or may not hire me? No thanks. And frankly, they're not great from the employer side either, too hard to proctor."
- Robert Brady (Righteous Marketing): "The 'findings' from a project should be discoverable in an interview (or two). If not, you need to improve the interview process."
- Julie Bacchini (Neptune Moon): "If you are asking someone to do work on an account... you should compensate them for their time."
This matters commercially, not just ethically: senior practitioners decline unpaid exercises, so issuing one silently shrinks the funnel to candidates with uncommitted hours.
## The formats, ranked
Pick by signal returned per candidate-hour, net of who declines rather than comply: `live screen-share audit > data-table case > own-work presentation > paid trial > job shadow`. The multi-hour unpaid take-home is not on this list at any position - it is the one format whose funnel cost exceeds its signal.
| Format | What it returns | What it costs |
| --------------------------------------------------------------------------- | ----------------------------------------------------------------------------- | ---------------------------------------------------------------------- |
| Live screen-share account audit | Method exposed in real time, immune to audit tooling and generated output | An hour of the interview plus a panelist able to grade an audit live |
| Data-table case (Marc Barraza's variant) | Reasoning about metric relationships, gradeable by anyone with the answer key | An hour, and a table built in advance |
| Own-work presentation (Tim Jensen) | Depth on work they actually did, no scheduling friction, near-zero drop-off | Their framing, their choice of case - you see the wins, not the misses |
| Paid trial or paid consulting engagement (Andrew McGarry) | The strongest signal available: real work, real constraints | A week of calendar, a budget approval, pay terms, and account access |
| Job shadow day, "no teacher/student dynamic" (Mike Ryan, Smarter Ecommerce) | Working style and how they ask questions | A day of a team member's time for a diffuse read |
- value (how much of their actual method you see): `paid trial > live screen-share audit > data-table case > own-work presentation > job shadow`
- effort (scheduling, panel hours, prep, budget approval, calendar drag): `paid trial > job shadow > live screen-share audit == data-table case > own-work presentation`
- candidate drop-off (seniors who decline rather than comply): `unpaid take-home > paid trial > job shadow > live screen-share audit == data-table case > own-work presentation`
- compliance cost (pay terms and worker classification for real work; access to live accounts and customer data): `paid trial > job shadow > live screen-share audit > own-work presentation > data-table case`
The two ties are structural, not undecided: the audit and the data-table case both consume one hour of an already-scheduled interview plus one prepared thing, and neither asks for an hour outside the loop, which is the only thing that drives drop-off. They separate on compliance, so that axis breaks them - an own-work presentation invites a former employer's account data into your room, which a synthetic table never does.
What this order starves is the paid trial: first on value and first on effort, so a ratio ranks it last forever while the in-interview audit wins on being cheap to run. Promote it on purpose for a senior hands-on seat, for a creative strategist whose portfolio cannot be verified as their own, or when a candidate already contracting elsewhere makes a week of their time cheap to buy.
This ordering is a default, not a law, and it moves with who is running the loop.
Delete rather than demote what your constraints rule out, and name the deletion:
- No panelist who can grade an audit live removes the top row.
- Accounts that cannot be exposed to an outsider remove it too.
- No budget or approval for paid work removes the paid trial and the job shadow outright.
Promotions work the same way in reverse - an existing anonymized dataset from a previous search collapses the data-table case's prep to near-zero.
Whichever row wins, keep it time-boxed and inside the interview:
- Duane Brown: "Testing within the 60-minute interview is fine. We do it. I wouldn't ask someone to take stuff home or do a 4-hour media plan for me."
- DigitalSamIAm (agency owner): "We have a test exercise, takes 90 minutes. I'm happy to pay for someone's dinner or whatever during it... Gives us a chance to evaluate how candidates think + handle challenges within a fixed time period."
- Ekta: "The sweet spot is to have a small test project with more of scenario-type problems instead of 'run 5 ads for me' type projects."
- Jon Kagan: take-homes "only for mid to senior level roles."
## The live screen-share account audit
Jason Parks (The Media Captain; published June 28, 2020, updated November 2023) describes the highest-signal exercise found: "I recommend asking the person you're interviewing to do a screen share with you where they audit your [ad] account... It's easy to send off a generic audit generated by third-party software. It's a vote of confidence to see someone confidently dissecting your [ad] account."
The live format is the whole point: it defeats both third-party audit tooling and AI-generated output. This is also why Foxwell Digital recommends paid real-world assignment tests for creative strategists, since AI-generated portfolios have made portfolio screening unreliable for judging creative-strategy skill.
Parks pairs the audit with a diagnostic prompt: "A month into our campaign, what results would raise a red flag?" - and expects a specific number (for example, ecommerce ROAS below 2:1), not a vague answer.
## Ambiguity detection is the signal, not a nuisance
Steve Gibson (PPCChat), from the candidate side of an ambiguous data case: "I... pointed out why the data wasn't conclusive enough to answer their 'What would you do...?' questions. I passed with flying colours. (They hadn't realized this, nor had any of the other candidates.)" Design cases with genuine ambiguity or insufficient data on purpose, and score naming it as a top answer. A candidate who confidently answers an underpowered question is showing the exact failure mode the job punishes.
A lighter-weight variant of the same idea is Marc Barraza's data-table exercise: present a realistic performance data table and have the candidate talk through optimization recommendations, scoring reasoning about metric relationships over rote definitions.
## Design rules
- Use synthetic or anonymized data. A take-home on a real client account leaks data and asks for free work; if real account work is unavoidable, pay for it.
- Write the rubric before the first submission, weighted toward commercial reasoning over technical recall.
- The debrief carries more signal than the artifact - reasoning walked aloud exposes method a polished deliverable hides. Budget debrief time explicitly.
- Match the format to the rung: live audit or data-table case for buyers and performance marketers; a paid concept-brief assignment plus a debrief on past creative wins and losses for creative strategists.
- Any numeric point values or time caps beyond the 60-90 minute consensus are this skill's own convention rather than a published standard - say so in the packet.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |