The two red flags that justify walking away from a mobile app marketing agency on their own are an agency that owns your ad account or measurement set-up instead of you, and an agency that guarantees a ROAS or CPI number. The other ten are warning signs to challenge: reports built only from Ads Manager, cheaper installs celebrated without any trial or subscriber data, blended ROAS as the only view, an account rebuilt every few weeks, and nobody able to explain what the last three tests taught you. Every one of them is something you can see in a pitch deck, a contract, a monthly report or the ad account itself, without taking anyone's word for it.
That is the difference between this list and our guide to the questions to ask a Meta Ads agency, which is about what to ask before you sign. Answers can be rehearsed. The signs below are things an agency does, so they are harder to talk your way around. If you already suspect a problem with an agency you work with, our step-by-step agency audit turns these signs into a scored review.
Key takeaways
- Ownership and guarantees are the only two walk-away signs on sight. Everything else deserves a written challenge and a fixed period to fix it.
- For a subscription app, the most expensive red flags are in reporting, because a report that stops at installs can show progress while the cost of each paying subscriber rises.
- Every red flag has an innocent explanation worth ruling out first. The test is whether the agency fixes it once you raise it.
- Some things that feel like warning signs are good signs: no guarantees, requests for revenue data, and advice to spend less.
The 12 red flags at a glance
Severity here means what to do when you see it. "Walk away" means the problem cannot be fixed by a better conversation. "Challenge" means raise it in writing and expect a fix inside about a month. "Watch, then audit" means one instance proves little and a pattern over a quarter proves a lot.
| # | Red flag | Where you see it | Severity | Innocent explanation to rule out |
|---|---|---|---|---|
| 1 | Guarantees a ROAS, CPI or CPA figure | Pitch | Walk away | None worth accepting. A forecast with stated assumptions is fine; a promise is not. |
| 2 | Case study results with no base, period or spend | Pitch | Challenge | Client confidentiality. The agency should still be able to give context privately. |
| 3 | Pitch speaks in reach, CTR and engagement, not trials and payers | Pitch | Challenge | A junior presenter. Ask to speak to whoever will run the account. |
| 4 | The offer is a fixed number of ads a month | Pitch | Challenge | Shorthand for capacity. Ask how many distinct concepts are behind the number. |
| 5 | Your ad account, dataset, app or MMP sits in the agency's business | Onboarding | Walk away | A temporary set-up for a brand-new app. It should move to you, in writing, on a date. |
| 6 | You cannot see inside the ad account yourself | Onboarding | Walk away | Nobody has granted you access yet. That takes minutes to fix. |
| 7 | A long lock-in with no notice period and no stated reason | Contract | Challenge | A short initial term for research and launch is normal and defensible. |
| 8 | Every reported number comes from Ads Manager | Reporting | Challenge | Early weeks before your subscription data is connected. It should not last. |
| 9 | Cheaper installs are celebrated without trial or paid data | Reporting | Challenge | The report is weekly and paid conversions lag. Ask for the cohort read later. |
| 10 | Only blended, calendar-month ROAS is ever shown | Reporting | Challenge | Cohorts are too young to judge. Fine early, not as a permanent format. |
| 11 | The account is rebuilt every few weeks | Ad account | Watch, then audit | A one-off consolidation after onboarding, or a real change in strategy. |
| 12 | Nobody can explain the last three tests or what replaced a tired ad | Ad account | Challenge | A new account manager. The test log should survive staff changes. |
Red flags in the pitch
1. They guarantee a ROAS, CPI or CPA
A subscription app's return on ad spend depends on the ads, but also on the onboarding, the paywall, the price, the trial length, the market and the product itself. No agency controls all of that, so a guaranteed number means one of two things. Either the agency will hit it by reporting the most flattering version of the metric, such as platform-reported ROAS with a generous attribution window, or it does not understand which variables it does not control. Neither is someone you want running your money.
Be careful not to confuse a guarantee with a forecast. "If your install-to-trial rate holds at around this level and we can get CPI to here, the maths gives you roughly this cost per trial" is a forecast with its assumptions on show, and it is exactly the kind of thinking you want. Our own Meta Ads service page answers the guarantee question with a flat no, and commits to a process, a creative volume and a reporting cadence instead.
2. Case studies with no base, no period and no spend
"+400% revenue" means something very different if revenue went from £2,000 to £10,000 a month while spend tripled than if it went from £200,000 to £1 million on flat spend. A case study that gives a percentage with no starting point, no time period, no spend context and no market is not lying, but it cannot tell you much. Confidentiality is a real reason for leaving figures out of a public page. It is not a reason the agency cannot talk you through the context on a call. If it will not, treat the headline number as marketing.
3. The pitch never gets past the install
Listen for the vocabulary. An agency that talks about reach, click-through rate, cost per click and engagement, and goes quiet when you ask about install-to-trial, trial-to-paid or cost per paying subscriber, is pitching a generic paid social service to an app. The same goes for iOS measurement: anyone running app campaigns should be able to explain in plain terms why recent iOS results arrive late and incomplete, which we cover in our guide to SKAdNetwork postback delays. A presenter who is not the account lead is an innocent explanation, so ask to meet the person who will actually run the account before you judge.
4. The offer is a fixed number of ads a month
"Forty ads a month" sounds like value, and output does matter. But forty variations of one idea give Meta far less to learn from than eight genuinely different concepts, and a fixed quota tends to be filled with the cheapest variations to make. The right amount of creative depends on your spend, audience size, number of markets and how fast ads fatigue, which is why we argue in how many ad creatives a mobile app needs each month that there is no universal number. Ask how many distinct concepts sit behind the quota, and how the number would change if your spend doubled.
Red flags in the contract and onboarding
5. Your accounts live in the agency's business
This is the most serious item on the list, because it is the one that hurts most after you leave. On Meta, an ad account belongs to a single business portfolio, and Meta's own help pages warn that adding an existing ad account to a portfolio can move it there permanently. If the agency creates your ad account, dataset or pixel inside its own portfolio, or claims your app in its own business, then your campaign history, audiences and conversion data sit in an asset you do not control. The same applies to an MMP contract or a subscription analytics workspace opened in the agency's name.
The correct set-up is simple. You own every asset and grant the agency partner access to it. The only innocent version is a brand-new app where the agency set things up to save time, and even then the transfer to you should be agreed in writing with a date. Our guide to switching agencies without losing momentum lists every asset to check.
6. You cannot see inside your own ad account
If your view of the account is limited to a PDF or a dashboard the agency curates, you cannot check anything in this article. You should be able to open Ads Manager yourself, at any time, and see the campaigns, the spend, the ads that are getting delivery and the account's change history. Most founders never look, and that is fine. The point is that they could. An agency that resists giving you direct access is protecting something.
7. A long lock-in with no stated reason
A minimum initial term is not a red flag by itself. Research, tracking checks, the first creative round and enough delivery to read the first tests take time, and an agency that lets you leave after three weeks is being asked to judge work before it can be judged. We use a three-month initial commitment, set out on our pricing page, followed by 30 days' notice. The warning sign is a term far longer than the time needed to read the work, a contract with no notice period after it, or an agency that cannot tell you what you should be able to see by the end of the initial term.
Red flags in the reporting
For a subscription app these three are the most expensive, because they let a failing account look healthy for months.
8. Every number comes from Ads Manager
Meta's reported ROAS counts only the users Meta credits to its own ads, inside its attribution window, at whatever value your app sends it. Your subscription platform counts what users actually paid. The two will never match exactly, and we set out every reason in why Meta, RevenueCat and your MMP show different ROAS. The red flag is not the gap. It is a report that never acknowledges the gap, never shows a figure from your own revenue data and never explains which number is being used for which decision.
9. Cheaper installs are celebrated on their own
A falling CPI is the easiest win to report and one of the least reliable. New hooks and broader targeting can bring in cheaper installs from people less likely to start a trial or pay. Here is a hypothetical example of how a report can show progress while the business goes backwards:
This month: CPI £2.40 (20% lower), install-to-paid 3.6%, so cost per paying subscriber = £2.40 ÷ 0.036 = £66.67
The headline says CPI fell by a fifth. The number that pays the bills rose by about 11%. If a report leads with CPI and has no line for trials or paying subscribers by cohort, ask for one. Our breakdown of why a lower CPI can make a subscription app less profitable works through the break-even point in detail.
10. Only blended ROAS, only by calendar month
Blended ROAS divides this month's revenue, much of it renewals from users acquired months ago, by this month's spend. It can hold steady while every new cohort is worse than the one before, because old subscribers are still paying. A report that only ever shows the blended view hides the one trend that matters for spend decisions. We explain how to read both in cohort ROAS versus blended ROAS. Early on, when no cohort is old enough to judge, a blended view is a reasonable stopgap. Six months in, it is not.
Red flags in the ad account
11. The account is rebuilt every few weeks
Meta's guidance is that an ad set usually leaves the learning phase after about 50 results in the week following its last significant edit, and significant edits include new ad sets, large budget changes and changes to the optimisation event. An account that is restructured every two or three weeks keeps resetting that clock and never produces a clean read on anything. Constant rebuilding often looks like activity to a client, which is part of why it happens. Our guide to the Meta Ads learning phase covers what counts as a significant edit.
You can check this yourself in the account's activity history. One consolidation after onboarding is normal. A new campaign structure every month, each with a different explanation, is a pattern worth auditing.
12. Nobody can explain the last three tests
Ask three questions. What were the last three creative tests? What did each one teach you about our users? What replaced the last ad that stopped working, and why that? An agency with a real creative process answers from a test log in a couple of minutes. One without it gives you a description of activity: "we refreshed the creative", "we tried some new hooks". That usually means there is no plan for creative fatigue beyond waiting for performance to drop and then making more of the same. A new account manager is a fair excuse once, but the log should outlive any one person.
Five things that look like red flags but usually are not
Buyers burned by one bad agency sometimes over-correct and screen out the good ones. These five deserve a second look before you count them against anyone.
| Looks like a warning sign | What it usually means |
|---|---|
| They will not guarantee results | Honest. Outcomes depend on your product, onboarding, paywall and price as well as the ads. |
| They ask for access to your revenue data | Necessary. Without subscription or store data they can only optimise to Meta's view of the world. |
| They recommend spending less, or not yet | A good sign. An agency that argues against more spend is showing you it will tell you the truth later. |
| The first few weeks look quiet | Normal if research, tracking checks and the first concepts are being built. Ask for the plan and dates, not instant results. |
| They only run one or two channels | Focus, not a weakness. Breadth only matters if your next dollar of growth is genuinely on another channel. |
How to decide: walk away, challenge or watch
A single red flag rarely tells the whole story, so match the response to the type of problem.
- Ownership or guarantees: walk away. Before signing, end the conversation. With a current agency, get every asset moved into your own business first, then decide whether to stay.
- Reporting or process: challenge in writing. Name the gap, say what you want to see instead, and give a fixed period of around 30 days. A good agency fixes reporting faster than that. One that argues about whether you need to see subscriber numbers has answered the question.
- Account behaviour: watch, then audit. Restructures, quiet spells and uneven creative output can each have good reasons. Track them for a quarter, then run a structured review using our agency audit scorecard.
Three or more challenge-level flags at once is a different situation. At that point the issue is usually not one fixable habit but a mismatch between what you need and what the agency is built to do. Our comparison of creative agencies and media buying agencies can help you work out which kind of partner your bottleneck actually calls for.
Hold us to the same list
We wrote this knowing a prospective client could run it against us, and they should. We do not guarantee ROAS. Our Meta Ads management reports cohort-level ROAS rather than the platform's figure, with MMP and SKAdNetwork attribution set-up and QA in scope. Our fee is a flat monthly amount, quoted to the scope before you commit and never a percentage of your ad spend, and the initial term and notice period are on our pricing page. Our case studies are named clients, and you should ask us for the context behind any figure on them.
If you are spending at least £3,000 a month on Meta and some of the flags above sound familiar, you can apply to work with us and we will tell you honestly whether the problem is the agency, the funnel or something else. For the wider picture of what should sit behind paid growth, start with our subscription app marketing strategy guide.
Frequently asked questions
What are the biggest red flags in a mobile app marketing agency?
Two justify walking away on their own: the agency owning your ad account, dataset or measurement setup instead of you, and the agency guaranteeing a ROAS or CPI figure. After those, the most damaging signs are reporting that only ever shows platform-reported numbers, celebrating cheaper installs without showing what happened to trials and paid subscribers, rebuilding the account every few weeks, and having no clear record of what was tested and what was learned.
Is it a red flag if an agency will not guarantee results?
No, it is the opposite. Paid acquisition results depend on your product, onboarding, paywall, pricing and market as well as the agency's work, so no agency controls the outcome. An agency that guarantees a specific ROAS or CPI is either planning to report a flattering number or does not understand the variables. What you can reasonably expect is a committed process, a stated creative cadence and a clear reporting rhythm.
How do I know if my app marketing agency is reporting the right numbers?
Ask whether the report reconciles to your own subscription data. A healthy report shows cost per trial, cost per paying subscriber and cohort ROAS at a fixed age, taken from a subscription platform such as RevenueCat or from store data, alongside Meta's figures. If every number in the report comes from Ads Manager, and none has ever been checked against what users actually paid, the report cannot tell you whether spend is profitable.
Is a minimum contract term a red flag for an agency?
Not by itself. A short initial term, typically around three months, covers research, launch and enough time to read the first tests. It becomes a warning sign when the term is much longer than the time needed to judge the work, when there is no notice period after it ends, or when the agency cannot explain what you should expect to see by the end of it.
Should I fire my agency after one red flag?
Only for the serious ones. If the agency owns your ad account or measurement setup and will not transfer it, or guarantees outcomes, end the conversation. For reporting and process problems, raise the issue in writing and give the agency a fixed period, around 30 days, to fix it. For account-behaviour signs such as frequent restructures, watch for a quarter and then run a structured audit before deciding.
The short version
Walk away from an agency that owns your accounts or guarantees your numbers. Challenge one whose reports stop at installs, never touch your own revenue data or only show blended ROAS, and give it a month to fix that. Watch for an account that is rebuilt constantly or a team that cannot explain its last three tests, and audit after a quarter if the pattern holds. And do not count honesty against anyone: an agency that refuses to guarantee results, asks for your revenue data and tells you to spend less is usually showing you exactly what you want to see.