Replacing Your Agency? Test Your Next Meta Ads Operator
If you are replacing an agency, do not hire the next person because they can explain an impressive dashboard. Give them a small, realistic account problem. Ask what they would change, what they would leave alone, and what evidence could change their mind.
I am Timofei Tskhovrebov. My proposition is direct ownership of the work: media buying, creative direction, landing pages and tracking within the agreed scope. For a business already spending at least $15,000 a month on Meta, that ownership needs to survive contact with an actual decision. A polished pitch is not enough.
Give every candidate the same problem
Use a fictional account or a small anonymized extract you have permission to share. Do not hand over customer records, logins or a folder of confidential creative. This is an evaluation of judgment, not permission to operate your business. Agree a paid, bounded exercise if you want substantial bespoke work.
Here is a hypothetical SaaS brief: $30,000 monthly Meta spend, 600 trials and 30 paying customers from a cohort old enough to observe conversion. A second cohort has spent $10,000 and generated 250 trials, but it is only five days old. The product team reports onboarding failures on one integration. The founder wants to double the budget next week.
Give the candidate the cohort dates, what counts as paid, refund treatment and the source of each number. Leave one useful fact unresolved, such as whether the integration issue affects every trial or one segment. A strong operator should identify the missing fact and explain how it changes the decision.
Ask for three decisions, with reasons
First, ask for the acquisition calculation. The mature cohort has a $50 trial cost and $1,000 media cost per paying customer. The newer cohort has a $40 trial cost. That does not yet establish a cheaper paying customer. Management and creative costs are also absent from those media-only figures.
Second, ask what they would do with tomorrow's budget. A useful answer names the conditions for holding, reallocating or increasing spend. It separates an onboarding defect from a targeting hypothesis. It does not pretend an average account metric identifies the broken campaign.
Third, ask for one creative test. The response should connect a buyer problem to a truthful product demonstration and a measurable next step. Ask which existing assets they would reuse and what would have to be produced. Ten unrelated headlines tell you less than one well-reasoned test.
Score the work before you hear the pitch
This is my proposed rubric, not a validated predictor of hiring success. Score each area from zero to three: zero misses the issue, one names it, two proposes a workable action, and three adds the evidence and owner needed to close the decision.
| Area | What earns a strong score | What to challenge |
|---|---|---|
| Economics | Correct cohort math, costs and maturity | Calling trials customers |
| Diagnosis | Separates tracking, product and acquisition | Changing targeting before checking the failure |
| Creative | One specific hypothesis and truthful proof | Asset volume without a buying reason |
| Ownership | Names implementation, approval and review owners | Every action belongs to someone else |
The maximum is twelve points. I would use the score to structure discussion, not invent a universal pass mark. A serious measurement error deserves attention even if the candidate presents beautifully. Ask them to correct it. How they handle the correction is part of the work sample.
Make them explain the handoff
Ask who changes the page, who reviews the claims, who checks the events and who approves the spend. If a specialist is needed, ask what they will receive and what the operator will verify afterward. An independent operator can coordinate specialist work without pretending to be every specialist.
Then change one input: the integration defect is fixed, or the mature cohort loses five customers to refunds. Ask which decision changes. With 25 retained paying customers, the same $30,000 media spend is $1,200 per retained customer at that observation point. It is a different metric; it must be labeled. You want someone who can update the plan without defending yesterday's slide.
Compare the answer with actual proof and scope
My anonymous AI SaaS case gives you an original account export: December spend of $17,852.93 at $139.48 per purchase, then May spend of $148,907.37 at $131.08. Those are spend and purchase CPA. The fictional trial numbers above are a separate exercise.
Read what I can own instead of an agency engagement, then compare that scope with the work sample. My current offers define the starting commercial terms. If you want the person making the decisions directly on the account, bring the account to me. I will explain the work I would take on and the decisions that need evidence first.