To evaluate a marketing agency case study, ignore the headline percentage until you know eight things: the starting point, the period, what happened to ad spend, which system measured the result, what else changed, which markets and stores it covers, whether it lasted, and whether you can check it with the client. A percentage with none of that context is not false. It is just not evidence yet. Most of the context is rarely on the public page, so the real test is whether the agency will give it to you when you ask.
We are an agency, and we publish case studies with big percentages on them, so we have an obvious interest here. That is why this guide ends by running the same test on our own pages and saying plainly what they do not tell you. It is the companion to our list of mobile app marketing agency red flags, which covers what agencies do. This one covers what they publish.
Key takeaways
- The size of a percentage mostly tells you how small the starting point was. Ask for the base before you are impressed.
- The same "+150% revenue" headline can describe an account that got much better, one that scaled well, or one that got worse. Only the spend figure tells you which.
- For a subscription app, installs and trials are not the result. Paying subscribers from the new cohorts are.
- Missing detail on a public page is normal. Refusing to supply it on a call is the signal.
Why a case study percentage tells you so little
A case study is a piece of marketing. It shows the agency's best results, chosen by the agency, described in the agency's preferred metric. None of that is dishonest. It does mean you are looking at the top of a distribution with the denominators removed, and your job is to put them back.
Start with the base. These are hypothetical figures, not client data.
App B: £200,000 to £280,000 a month = +40%, which is £80,000 of new monthly revenue
App A makes the better headline. App B is the harder piece of work, because finding £80,000 of extra monthly revenue in an account that is already large means beating creative and audiences that have had real money behind them. If your app looks like App B, a wall of four-figure percentages from apps that looked like App A tells you the agency can launch. It does not tell you it can scale you.
The eight questions that turn a case study into evidence
| Question | Why it matters | What a good answer contains |
|---|---|---|
| 1. What was the starting point? | Percentages are biggest when the base is smallest. | An absolute before and after, or at least a range such as "low five figures a month". |
| 2. Over what period, compared with what? | A best month against a worst month is not a trend, and many app categories are seasonal. | Named months, and a like-for-like comparison period. |
| 3. What happened to ad spend? | Revenue can rise simply because more was spent, at worse efficiency. | Spend before and after, or the ratio of revenue to spend at both points. |
| 4. Which system measured it, gross or net? | Meta, an MMP and the app stores report different revenue for the same month. | The source, the attribution window, and whether store commission and tax are deducted. |
| 5. What else changed? | Price, paywall, onboarding, new features, press and other channels all move revenue. | A plain list of what changed alongside the ads, and who did it. |
| 6. Which markets and stores? | A CPI or ROAS from one country or store does not transfer to another. | Countries and platforms named, and whether the mix shifted during the period. |
| 7. Did it last? | A launch spike and a result held for a year are different achievements. | What the numbers did after the period shown, and whether the client is still a client. |
| 8. Can it be checked? | An unnamed result cannot be verified by anyone. | A named client, a named person, and a reference call if you ask for one. |
You will almost never get all eight from the page. Count how many you can answer from it, then take the rest to the call. The first three matter most, because between them they explain the majority of impressive numbers.
One headline, three different accounts
Spend is the context most often left out, and it changes the meaning of a revenue figure completely. Here is one headline, "+150% revenue", with monthly revenue going from £30,000 to £75,000 in each case. The numbers are hypothetical.
| Story | Monthly spend | Monthly revenue | Revenue per £1 spent | What it means |
|---|---|---|---|---|
| A. Spend held flat | £20,000 to £20,000 | £30,000 to £75,000 | 1.50 to 3.75 | The same budget now earns far more. Creative or funnel got much better. |
| B. Spend rose in step | £20,000 to £50,000 | £30,000 to £75,000 | 1.50 to 1.50 | Scaled two and a half times with efficiency held. Hard to do, and a real result. |
| C. Spend rose faster | £20,000 to £80,000 | £30,000 to £75,000 | 1.50 to 0.94 | Growth was bought at a loss on this measure. The headline is true and the account is worse. |
All three agencies can truthfully write "+150% revenue". Story B is worth dwelling on, because buyers often dismiss it as "they just spent more". Holding efficiency while spend rises two and a half times is difficult, and it is usually the result an app that already spends actually wants. The point is not that more spend makes a result worthless. The point is that you cannot tell A, B and C apart without the spend figure.
Even the ratio needs care. Monthly revenue divided by monthly spend is a blended figure, and for a subscription app it includes renewals from people acquired long before the period. Our guide to cohort ROAS versus blended ROAS shows how a blended number can look healthy while every new cohort is getting worse.
What each common headline metric can hide
| Headline | What it can hide | What to ask for |
|---|---|---|
| "+X% revenue" | A tiny base, a bigger budget, a price rise, or renewals from subscribers acquired before the agency arrived. | Absolute before and after, spend at both points, and revenue from new cohorts only. |
| "+X% trials" | A lower trial-to-paid rate, a changed trial offer, or cheaper traffic that starts trials and cancels. | Paying subscribers and cost per paying subscriber over the same period. |
| "-X% CPI" | A shift towards Android or cheaper countries, or lower-intent installs that convert worse. | CPI by country and store, and what happened to cost per trial and per payer. |
| "Nx ROAS" | The window (day 7 or a year), the source (Meta or your own revenue data), gross or net, and organic revenue mixed in. | Window, source, gross or net, and whether it is a cohort figure or a blended one. |
| "-X% CPA" | Which action. A cheaper registration is not a cheaper customer. | The exact event the A stands for, and the cost of the next event down the funnel. |
| "Record month" | Seasonality, a product launch, or simply record spend. | The same month a year earlier, and spend in both. |
Trials are not subscribers
Trial growth is the most common subscription app headline, because trials arrive within days and paid conversions take weeks. Another hypothetical shows why it needs a second number beside it.
After: £40,000 spend, 2,000 trials, 12% convert = 240 payers at £167 each
Trials +300%. Cost per trial unchanged at £20. Paying subscribers +37%. Cost per payer almost tripled.
Nothing in that case study would be untrue, and the cost per trial would look perfectly stable in the ad account. This is the pattern we cover in why a lower CPI can make a subscription app less profitable: broader, cheaper traffic that takes the free step and not the paid one. When a case study leads with trials, ask what trial-to-paid did over the same period, and compare it with the trial-to-paid benchmarks for the category.
A ROAS without a window is not a number
"3x ROAS" could be day-seven revenue reported by Meta, or a year of net proceeds from your own billing data. Those can differ by a multiple for the same cohort. In our worked example on why Meta, RevenueCat and your MMP show different ROAS, one month of spend read 1.20 in Meta and 0.80 as net proceeds at day 60, after refunds, VAT and store commission. Any ROAS in a case study needs three labels: the window, the source, and gross or net. Our subscription app ROAS benchmarks show how far the figure moves between day 7, day 30 and day 60.
A short period on a subscription app is an early read
A result measured over six or eight weeks is real, but for a subscription app it is mostly a result about trials and first payments. In the model in our subscription app funnel benchmarks, a monthly-plan cohort has delivered only about 27% of its first-year net revenue by day 35. Whether the new subscribers renew is not yet known when a short case study is written. That is not a reason to discount it. It is a reason to ask what the same cohorts did three and six months later.
What else changed at the same time?
Revenue has many parents. In the period a case study covers, the app may have raised its price, moved its paywall, shortened onboarding, launched in a new country, been featured by a store or entered its seasonal peak. A fitness app in January and a fitness app in November are close to being different businesses.
The strongest evidence that ads caused a result is a holdout test, where a share of the audience is deliberately not shown the ads and the difference is measured. Few case studies have one behind them, because the tests need volume: Meta's guidance for a Conversion Lift study suggests, as a guide, at least USD 5,000 of spend and 500 conversions. Our agency audit guide covers when a test like that is realistic. Without one, the honest version of a case study says what else changed and lets you judge. A page that credits the product team for part of the result is more believable than one where the ads did everything.
The case studies you are not being shown
Every case study page is a selection. If an agency has worked with forty apps and shows six, the six tell you what its best outcomes look like and nothing about the typical one. Two questions fix this better than any amount of reading: "What does a normal first six months look like for an app at my spend level?" and "Tell me about a client where it did not work, and why." An agency that has an answer to the second question, with a cause attached, has usually learned something from it. One that says it has never happened is either very new or not being straight with you.
Check the match as well as the size. A result from a hard-paywall utility app on iOS in the United States says little about a freemium wellness app on Android in Germany. Look for the case study closest to your category, store, price and spend, and weight that one above the largest number on the page.
Three grades of case study
- Grade A, evidence. Named client, a base or a range, a period, spend context, the measurement source, and a reference you can speak to. Rare on a public page, and worth a lot when you find it.
- Grade B, checkable. Named client and a description of the work, with percentages but not the full context. Most reputable agency case studies sit here. They become grade A or fall apart on the call.
- Grade C, decoration. No client name, no period, a percentage and a logo wall. Treat it as a statement that the agency would like your business.
Anonymity alone does not make a case study grade C. Plenty of app companies will not let an agency publish their name or their figures, and percentages are often used precisely because the client will not allow absolutes. What matters is whether something checkable replaces the missing detail, usually a reference call.
You are entitled to ask for the evidence
Buyers sometimes feel that asking an agency to back up its case studies is rude. It is not. In the UK, the advertising rules assume the evidence exists. Rule 3.7 of the CAP Code says marketers must hold documentary evidence to prove objective claims before they publish them. The Code's scope covers marketing claims on a company's own website, and it defines a consumer as anyone likely to see the communication, whether in the course of business or not. "Revenue up 400%" on an agency's site is an objective claim. This is not legal advice and other countries have their own rules, but the principle travels: a number on a sales page should have a file behind it.
These are the questions we would take to the call.
- What was monthly revenue and monthly spend at the start and at the end?
- Which months does this cover, and why those months?
- Where does the revenue figure come from, and is it before or after store commission?
- What did trial-to-paid and cost per paying subscriber do over the same period?
- What changed in the product, price or paywall while you were running the account?
- What has happened since, and is this client still with you?
- Can I speak to someone there for fifteen minutes?
You will not get a perfect answer to all seven. You are listening for whether the person knows the account well enough to answer without looking it up, and whether "I cannot share that" comes with a reason. For the wider set of pre-contract questions, see our questions to ask a Meta Ads agency.
The same test, run on our own case studies
It would be easy to write this guide and exempt ourselves. Here is what five of our published case studies state on the page today, and what they leave out.
| Case study | What the page tells you | What it does not tell you |
|---|---|---|
| Steps & Beasts | Named client and founder. Revenue +1112%, active subscriptions +118%. Credits an onboarding overhaul alongside creative testing and AI UGC. | The period, the starting revenue, the change in spend, the markets, and the system that measured it. |
| Oli Help | Named client. Trials +1,750%, revenue +394%. Angles were tested against the ideal customer, then the winners were scaled. | The period, the base, how much spend rose when the winners were scaled, and paid conversion from those trials. |
| LingoLooper | Named client. Revenue +49% and trials +116% in two months, with a simplified account structure. | The base, the change in spend, and whether paid conversions kept pace with trials. |
| Tonaly | Named client with a quote from the founder. Revenue up 175%, 3x ROAS sustained at scale, on a paid app. | The ROAS window and source, the period, and the spend level behind "at scale". |
| Preggers | Named client with a quote from the CEO. Lower CPI and the US opened as a profitable market. No percentage is claimed. | Any figure. Honest, but there is nothing on the page to compare against. |
By our own grading these are grade B. The clients are named, each page links to the app, two carry a quote from a named founder or CEO, and each page says what the work was. But the figures are percentages, most have no stated period, and none gives the base or the spend. A reader applying this guide to us would be right to say so.
Two details are worth pointing out because they cut against our own headline. The Steps & Beasts case study credits an onboarding overhaul as one of three workstreams, so the revenue figure belongs to the creative and the product change together, not to ads alone. And the Oli Help page leads with trial growth, which by the logic above should make you ask about paying subscribers. Both are fair questions, and you should ask us for the context behind any figure on those pages before you rely on it.
If you are comparing agencies now
Read each agency's case studies with the eight questions beside you, pick the one closest to your own app, and take the gaps to the call. Our guide to choosing a mobile app marketing agency covers the rest of the evaluation, and our comparison of the best Meta Ads agencies for mobile apps ranks them partly on verifiable app proof.
If you want to put these questions to us, our Meta Ads management reports cohort-level ROAS from your own revenue data rather than the platform's figure, so the numbers we would show you are built the way this guide recommends. If your app already spends at least £3,000 a month on Meta, you can apply to work with us and bring the list.
Frequently asked questions
How do you evaluate a marketing agency case study?
Ask eight questions of every headline number: what was the starting point in absolute terms, over what period and compared with what, what happened to ad spend, which system measured the result and whether revenue is gross or net, what else changed in the product or price, which markets and platforms it covers, whether it lasted, and whether you can check it with a named client. A case study that answers most of them is evidence. One that answers none is a claim of capability, which is a different thing.
Is a large percentage increase in an agency case study meaningful?
Not on its own. Percentages are largest when the starting point is smallest. Revenue going from £2,000 to £24,000 a month is a 1,100% increase and £22,000 of new monthly revenue. Revenue going from £200,000 to £280,000 is a 40% increase and £80,000 of new monthly revenue. The second result is smaller on the page and bigger in the bank. Always ask for the base, the period and what happened to spend.
Should I trust anonymous case studies from an app marketing agency?
Anonymity is not proof of anything bad. Many app companies do not allow agencies to publish their name or figures. But an anonymous case study cannot be checked, so it should carry less weight than a named one, and the agency should be able to offer a reference call with a real client instead. If every case study is anonymous and no reference is available, you have no evidence, only assertions.
What should a good mobile app marketing case study include?
The client's name or a stated reason it is withheld, the starting point, the time period, what happened to ad spend, the metric's definition and the system that measured it, which markets and stores are covered, what the agency actually did, what else changed at the same time, and what happened after the period shown. For a subscription app it should report paying subscribers or revenue from the new cohorts, not only installs or trials.
Can I ask an agency for evidence behind its case study numbers?
Yes, and a good agency expects it. Public pages often leave figures out for client confidentiality, but the agency should be able to talk you through the base, period, spend and measurement on a call, and arrange a reference with the client. In the UK, the CAP Code also requires marketers to hold documentary evidence for objective claims before they publish them, and that applies to claims on a company's own website.
The short version
A case study number becomes evidence when you know the base, the period, the spend, the measurement, what else changed, where it happened, whether it lasted and who can confirm it. Big percentages usually mean small starting points. Revenue growth means nothing until you know what was spent to get it. Trials are not subscribers. And a public page that leaves detail out is ordinary, while an agency that will not fill it in when asked has told you what its numbers are worth.