Most first calls run backwards. The agency asks about your spend, your targets and your timeline, you answer everything honestly, and forty minutes later you have learned almost nothing about how they work. The questions to ask a marketing agency are not complicated, but you have to actually ask them, and you have to know what a real answer sounds like.
We sit on both sides of this. We pitch, and we also inherit accounts from agencies that were hired on a good deck and a warm call. The pattern in the ones that went wrong is almost never that the agency lied. It is that nobody asked the question whose answer would have been awkward.
Below are the nine we would ask if we were buying. Each one comes with what a good answer sounds like and what a bad one sounds like, because the value is not in the question, it is in knowing whether what you just heard was a system or a sentiment.
Key takeaways
- Good answers are specific and falsifiable. Bad answers are general and flattering, and they all sound fine in the moment.
- Creative volume and the decision rules behind it are the real variables in a Meta ads engagement. Account management is a few hours a week and almost never the difference.
- Ask what they do when a winner dies. How an agency handles fatigue tells you more about the next twelve months than any case study will.
- Ownership of the ad account, the pixel and the raw creative files should be settled in writing before you start, not at the point you want to leave.
- An agency that cannot name a situation in which it would tell you to stop spending, or hand the account back, is selling retention rather than performance.
What you are actually buying
Before the questions, the frame. When you hire a Meta ads agency in 2026 you are mostly buying creative: how much of it gets made, how different each piece is from the last, and how quickly it gets replaced once the audience has seen it. The account work matters, but bid strategies and budget splits are a solved problem compared to the volume and variety of ideas going into the auction.
That is why almost every question below is really a question about creative in disguise. If you only take one thing from this list, take the habit of translating every answer into: does this tell me how much thinking gets produced each month, and by whom. Our full buyer's guide to hiring a performance creative agency works through the evaluation criteria in more depth.
The nine questions to ask a marketing agency
1. How many creative concepts will you produce each month, and who writes them?
This is the single most useful question on the list because it is the hardest to answer vaguely without it showing. Concepts, not variations. Ten resizes of one video is one concept. You are asking for a number, and then you are asking whose brain that number comes out of.
A good answer: a specific monthly volume, justified by how fast creative fatigues in your category, with a named strategist who writes the briefs. In our experience mobile gaming creative fatigues in one to two weeks at scale, health and fitness in three to four, and finance or SaaS in five to eight, so the right number is different by vertical and a good agency will say so unprompted.
A bad answer: "as many as it takes", "we scale it based on performance", or any version of creative being something you supply and they merely traffic.
2. How do you decide what to test next?
You are looking for a system with a shape to it. Every agency tests. The question is whether the next test is chosen because it fills a known gap or because somebody saw a competitor ad on Tuesday.
A good answer: they describe a map of some kind, then say which parts of it your account is not covering. We use psychological zones, the emotional position an ad occupies rather than the format it uses, and the useful property of any such map is that it tells you what you have never tried. The reasoning behind that approach is in our guide to Meta ads creative fatigue for mobile apps.
A bad answer: "we double down on what's working." Doubling down on winners is how an account ends up with fourteen versions of the same idea and a rising cost per install nobody can explain.
3. What do you do when an ad that was working stops working?
Every account you will ever run hits this. It is the most common single event in performance marketing and the answers separate operators from managers faster than anything else you can ask.
A good answer: they distinguish between the asset being tired and the idea being exhausted, because the responses are completely different. They mention the diagnostics they read first, something like hook rate declining before cost per acquisition moves, or frequency climbing past the 3.0 mark in a seven day window that is the standard cold audience threshold. Then they tell you whether that means refresh, pivot or retire.
A bad answer: "we turn it off and shift budget to the next best ad", or duplicating the ad set to reset learning. Both are motion rather than diagnosis, and both work exactly once.
4. Who owns the ad account, the pixel and the raw creative files if we stop working together?
An unglamorous question that decides how expensive it is to change your mind later. Ask it early and in a friendly tone, because the reaction is as informative as the answer.
A good answer: you own the Business Manager, the ad account, the pixel and the historical data, they are granted partner access, and you receive the raw project files and creator footage with usage rights specified in the contract. They will have said all of this before you finished asking.
A bad answer: the ads run inside the agency's Business Manager, or the raw files and creator usage rights are described as something to sort out at the time. Learning data and creative libraries are assets, and rebuilding either costs months.
5. Which single number are you accountable for, and how often will I see it?
Agencies that report on everything are accountable for nothing. A forty metric dashboard is not transparency, it is cover, because there is always one line trending up.
A good answer: one business relevant number, named without hesitation, with a cadence attached. Day 7 return on ad spend for a subscription app, cost per qualified claimant for a legal campaign, blended cost per acquisition against a payback window. A strong answer also volunteers where the number is unreliable, for example that iOS reads lag because of delayed SKAdNetwork postbacks.
A bad answer: a headline built on click through rate, cost per mille or engagement. Those are diagnostics, not outcomes, and an agency that leads with them is choosing metrics it can always make look good.
6. How exactly do you charge, and what happens to your fee if my spend doubles?
Every pricing model creates an incentive, and none of them are neutral. You are not looking for the model with no downside, you are looking for an agency that can articulate the downside of its own.
A good answer: the model is named plainly, the behaviour at scale is spelled out, and they tell you unprompted what their structure biases them towards. A percentage of spend agency admitting the model rewards them for recommending more budget is a good sign, not a bad one. We break down all five common structures in what Meta ads agencies actually charge in 2026.
A bad answer: "we're flexible, let's work something out." Flexibility before scoping means the price is being set by what they think you will pay rather than by the work.
7. What have you run in my vertical, and what transferred?
Note the second half. Anybody can produce a logo wall. The valuable part is whether they learned something structural about your category or simply had a client in it once.
A good answer: they talk about the constraints of your vertical rather than its successes. Compliance limits on health claims, the way subscription apps hide revenue behind trials, seasonality, review cycles. Then they tell you what did not transfer from an adjacent category and why.
A bad answer: a list of brands with no specifics, or the claim that the playbook is identical everywhere. Vertical experience is real and so is its limit, and both are worth hearing about.
8. Who is actually doing the work day to day?
The oldest problem in agency buying. The people in the pitch are frequently not the people in the account, and nobody tells you when the handover happens.
A good answer: named people, their actual roles, roughly how many other accounts they carry, and ideally at least one of them on this call. A small senior team saying "us, and here is what we outsource" is a better answer than a large org chart.
A bad answer: "you'll have a dedicated account manager." Dedicated to what, and doing which part? If the person you have been impressed by for an hour cannot say whether they will be in the account next month, assume they will not be.
9. What would make you tell me to stop spending, or hand the account back?
The last question and the most revealing, because there is no prepared answer for it. You are asking whether they have a threshold at which honesty beats revenue.
A good answer: a real scenario. Retention too weak for paid acquisition to ever pay back. A price point the category will not support. A product that needs six months of work before media makes sense. We tell people this regularly, and it is usually cheaper for them than a year of optimising something that was never going to work.
A bad answer: "there is always something we can improve." True, and irrelevant. An agency with no exit condition has an incentive to keep the account alive long after it should have told you the truth.
What the answers add up to
Score the call on one axis: specificity. Good agencies answer with numbers, names, decision rules and at least one thing they cannot do. Weak ones answer with adjectives, and the adjectives are always the same three, so you can spot the pattern by the third call.
Two smaller signals worth watching for. The first is whether they asked to see the account before quoting, because anybody who can price the work without looking at your fatigue rate, creative library or historical spend is pricing a package rather than a problem. The second is whether they disagreed with you about anything during the call. An agency that agrees with every assumption you brought is either lucky or selling, and after twelve months of that you will be paying for a service that never once told you something you did not want to hear.
If you want the wider evaluation framework around these questions, the pricing models, the red flags and what a competent first ninety days looks like, that is all in how to hire a performance creative agency.
Frequently asked questions
What questions should I ask a marketing agency before hiring them?
Ask about creative volume and who writes it, how they choose what to test next, what they do when a winning ad dies, who owns the account and the files, which single number they are accountable for, how the fee behaves as your spend grows, what they have run in your vertical, who does the work day to day, and what would make them tell you to stop spending. The pattern that matters is whether the answers are specific and falsifiable or general and flattering.
How do I know if a Meta ads agency is any good?
Good agencies describe a system and bad ones describe an attitude. If the answers to your questions contain numbers, named people, decision rules and at least one honest limitation, you are talking to an operator. If they contain the words passionate, data driven and bespoke without a single specific behind them, you are talking to a sales process.
Should I ask a marketing agency for case studies?
Yes, but ask a better version of the question. A case study is a curated best result, so ask instead what they tried in that account that did not work and what they changed as a result. An agency that can talk fluently about its own failed tests is describing real work, while an agency that has only ever succeeded is describing a deck.
How long should I give a new agency before judging them?
Around ninety days, with checkpoints rather than one verdict at the end. The first month is account and creative groundwork, the second is when a meaningful volume of new concepts is in market, and the third is when the winners from that testing are scaling. On iOS, delayed SKAdNetwork postbacks mean any read shorter than a full week is systematically incomplete.
Is it a red flag if an agency will not guarantee results?
No, the opposite. Nobody controls the auction, your product, your onboarding or your pricing, so a guaranteed cost per acquisition is either heavily caveated or being funded by the agency choosing safe, low volume tactics. What you should expect instead is a guaranteed process: a stated volume of creative, a stated testing cadence and a stated reporting rhythm.
Want this run for you?
If you are going to ask us these nine questions, we would rather you asked them properly. We will tell you the creative volume, who writes it, what we do when a winner dies, and the conditions under which we would tell you not to spend at all.
We are a performance creative agency for mobile apps, and the work is mapping the psychological zones an account is not covering, then writing and producing against the gaps. If that sounds like what your account is missing, apply to work with us. We take a small number of mobile app clients per quarter.