Hire a creative agency when your account can learn and your users convert after the install, but you cannot produce enough different ideas to replace the ads that are tiring. Hire a media buyer when good ads already exist but the account is structured so Meta cannot learn from them: ad sets starved of optimisation events, constant edits, or the wrong event being optimised. And hire neither if the real leak is after the install or in your tracking, because no agency of either kind fixes an onboarding problem or a broken revenue number. Most apps that ask "creative or media?" have one clear answer, and finding it takes an afternoon in the ad account rather than a round of agency pitches.
This guide gives you that afternoon: what each type of agency actually controls, the four problems that look alike from the dashboard, the order to check them in, and a worked example of an app that needed both, in a particular sequence.
Key takeaways
- There are four bottlenecks, not two: measurement, product, media buying and creative. Two of them are not agency problems at all.
- Check measurement first and product second. A creative or media diagnosis built on wrong numbers or a leaky funnel is wasted.
- The most useful single split is cost per trial or subscriber by concept. A wide spread points to creative; a narrow, uniformly poor one points to structure first.
- Media fixes are cheap and fast. Creative fixes take weeks. When both are broken, fix the structure first so new concepts can be read.
- Creative is not always the answer. It is the answer when the account can learn, the funnel holds, and the ideas have run out.
What a media buyer controls and what a creative agency controls
The two jobs used to overlap less than they do now. A media buyer ran targeting, placements, bids and budgets. A creative agency made the ads. On Meta, much of the first job has been automated. Meta's engineering team describes Advantage+ as automating budget allocation, audience targeting and placement, and its Andromeda retrieval system narrows tens of millions of candidate ads to a few thousand for each person before ranking even starts. The practical effect, which our explainer on Meta Andromeda for mobile apps covers in more depth, is that the ad itself now does much of the work targeting used to do.
That does not make media buying obsolete. It changes what it is. Here is how the levers divide up on a typical subscription app account.
| Lever | Who owns it | How fast a fix shows |
|---|---|---|
| Optimisation event (install, trial, purchase) | Media buyer | Days to set, two to four weeks to judge |
| Campaign and ad set structure, budget split | Media buyer | Days |
| When to scale, and by how much | Media buyer | Weeks |
| Audience targeting and placements | Largely automated by Advantage+ | Little left to tune |
| Angles, hooks and the promise the ad makes | Creative partner | Weeks, because concepts must be made, launched and read |
| Concept supply: how many genuinely different ideas reach the account | Creative partner | Weeks to months |
| Onboarding, paywall, offer and price | Your product team (neither agency) | Weeks, plus release cycles |
| Event tracking and the revenue source of truth | Your data or engineering team, with the media buyer | Days to weeks |
Two rows in that table belong to neither agency. That is the point most comparisons of the two skip, and it is where a lot of agency spend gets wasted.
Four problems that look the same from the dashboard
Rising cost per subscriber is a symptom, not a diagnosis. The same line on a chart can come from four different causes, and each needs a different person to fix it.
| Problem | What you see | The test that confirms it | Who fixes it |
|---|---|---|---|
| Measurement | Meta's trial or purchase counts disagree badly with RevenueCat or your MMP. ROAS looks good in Ads Manager and bad in finance, or the reverse. | Reconcile one matured week of trials and first payments, Meta against your revenue system, by campaign. | Your data team and the media buyer. Not a creative hire. |
| Product | Every ad concept converts poorly after the install. Install-to-trial or trial-to-paid sits well below category benchmarks whichever ad brought the user in. | Split install-to-trial and trial-to-paid by concept. If the best and worst concepts both convert poorly, the leak is after the click. | Onboarding, paywall or offer work. Neither agency type. |
| Media buying | Good concepts exist but sit in ad sets that never collect enough optimisation events. Frequent edits, many small ad sets, a mismatched optimisation event. | Count optimisation events per ad set per week and significant edits per month. Check the event matches what you pay for. | A media buyer, in house or agency. |
| Creative | One or two old ads carry most of the spend. New concepts rarely beat them. Cost per result on the leaders creeps up while CPMs hold. | Rank concepts by matured cost per trial or subscriber. Count genuinely new concepts launched in the last 30 days. | A creative partner, or more creative capacity in house. |
Check them in this order
The order matters because each check depends on the one before it. A creative diagnosis read off broken tracking is wrong, and a media diagnosis made while the paywall leaks will blame the wrong ads.
1. Is it a measurement problem?
Take one week of trials that is old enough to have settled and count it in two places: Meta, and whatever you treat as the revenue source of truth, usually RevenueCat or your MMP. Some gap is normal. Meta attributes within its own windows, and on iOS much of the signal arrives aggregated and late, which our guide to SKAdNetwork postback delays explains. A gap that is large, unstable from week to week, or different in direction between campaigns means every number below it is suspect. Fix that first. Neither a creative agency nor a media buyer can make good decisions on numbers that do not describe your revenue.
2. Is it a product problem?
Next, look at what happens after the install, split by ad concept. RevenueCat's State of Subscription Apps 2026 puts the Health & Fitness median download-to-trial rate, measured within 30 days of install, at 6.9%, and the category's median trial-to-paid rate at 37.7%. Our benchmark guides for install-to-trial and trial-to-paid conversion give the other categories and explain why the two figures cannot be multiplied together. Those benchmarks blend paid and organic users, so treat them as a range to sit near rather than a pass mark.
The useful split is by concept. If one concept's users start trials at twice the rate of another's, the ad is shaping who arrives, and creative can move the number. If the best and worst concepts both convert poorly, the leak is in onboarding, the paywall or the offer, and buying more users into it only makes the leak more expensive.
3. Is it creative or media?
Only now is the question in the title worth asking, and one split answers most of it.
The concept spread test
Rank every concept that has spent meaningfully in the last 60 days by its cost per trial, or cost per subscriber if you have the volume, using cohorts old enough to have matured. Keep the comparison inside one campaign and one optimisation event so you are comparing ads, not setups. Then look at the shape.
- A wide spread with old leaders. One or two concepts are much cheaper than the rest, they are weeks old, they take most of the spend, and nothing launched recently gets near them. The account knows how to find good ads. It is not being given enough new ones. That is a creative supply problem.
- A narrow spread, uniformly poor. Every concept performs about the same, and none performs well. This can mean the concepts are all too similar, but check the structure before concluding that. Ads that sit in ad sets which never stabilise all read badly, good ones included.
- A wide spread with fresh leaders. New concepts regularly take over from old ones. The creative engine is working. If costs are still too high, look at structure, the optimisation event, or how hard you are pushing spend into a saturating audience, which our guide to scaling Meta Ads for a subscription app separates from fatigue.
The narrow-spread case is where most misdiagnosis happens, so these signals help separate the two sides.
| Signal in the account | What it usually means | Points to |
|---|---|---|
| Ad sets regularly below about 50 optimisation events a week | Meta says ad sets usually leave the learning phase after about 50 results in the week after the last significant edit. Below that, results stay unstable and every test reads noisily. | Media |
| Several significant edits a week across the account | Each one can send an ad set back into learning. A busy account is not the same as a well-run one. | Media |
| Optimising for installs while judging the account on subscribers | Meta finds people likely to install, not necessarily people likely to pay. The event and the goal disagree. | Media |
| One concept takes most of the spend for weeks, and new ones lose | The account has a winner and nothing ready to replace it. That is a supply problem, not a buying one. | Creative |
| Leader's cost per result rising while CPMs stay flat | The audience is tiring of the ad rather than the auction getting dearer. Meta flags fatigue at twice the ad's past cost per result, but not on Advantage+ app campaigns, so most app teams must spot it themselves. | Creative |
| New concepts are variations of the winner with a different first line | Plenty of assets, very few distinct ideas. The account is testing iterations, not angles. | Creative |
The 50-event figure comes from Meta's learning phase guidance, which our learning phase guide unpacks. The fatigue threshold is from Meta's creative fatigue help page, which also says the status is not available on Advantage+ app campaigns. The last row is the one we see most. An account can launch twenty new ads a month and still be short of ideas, which is the difference our piece on creative diversity vs volume is about.
Worked example: an app with two problems and one right order
The numbers in this example are hypothetical. They describe a fitness app spending £20,000 a month on Meta, about £4,600 a week, with a cost per trial around £24 against a £20 target. The founder is choosing between a creative agency and a media buyer.
| Check | Reading | Verdict |
|---|---|---|
| Measurement | Meta trials within about 8% of RevenueCat trials for the matured week | Healthy |
| Install-to-trial | 7.4% (Health & Fitness median 6.9% within 30 days) | Healthy |
| Trial-to-paid | 36% (Health & Fitness median 37.7%) | Close to median |
| Structure | Six ad sets at about £770 a week each, about 32 trials each | Below the ~50 a week Meta describes |
| Concept spread | Best concept £16 per trial, nine weeks old, 64% of spend. Two new concepts this month, both above £30 | Creative supply is thin |
Measurement and product pass, so this is an agency problem. But it is both kinds. The structure is starving every ad set: £4,600 a week split six ways is about £770 each, which at £24 a trial buys about 32 trials per ad set per week. And the creative supply is thin: one nine-week-old concept is carrying the account, and the two new ones have not had a fair test.
Notice why those two new concepts may not have had a fair test. They launched into ad sets that never stabilise, so their £30-plus cost per trial is partly a structural reading, not purely a creative one. Hire a creative agency first and it will ship concepts into the same six starved ad sets, where good ideas and bad ones read much the same. You pay for production you cannot learn from.
So the order is media first, then creative. Consolidating six ad sets into two puts about £2,300 a week behind each, roughly 96 trials a week at the current cost, comfortably above the level Meta describes. That is a change of days, not weeks, and it costs no production budget. Consolidation does not promise a lower cost per trial by itself. What it buys is readable tests, so that the new concepts which follow are judged on what they are rather than on where they landed.
Once the structure can learn, the remaining constraint is plainly creative, and the right next spend is on concept supply. Our creative refresh calculator gives a starting estimate for how many new concepts an account at this spend needs each month.
Real accounts are rarely this tidy, and the answer is often "both, in sequence". Our Steps & Beasts case study is an example of a fix that needed more than one lever: heavy creative testing alongside onboarding changes that lifted install-to-subscriber conversion, which together took revenue up 1112% and active subscriptions up 118%. Neither change alone tells the whole story.
What it costs to hire the wrong one
Each wrong hire fails in its own recognisable way, which is useful because you can spot it within a couple of months rather than a year.
- A creative agency hired for a media problem. Lots of new ads, very few clear winners, and a growing belief that "nothing works". The concepts were never read properly because the account could not learn.
- A media buyer hired for a creative problem. A burst of improvement from cleaner structure, then the same ceiling as before, followed by restructure after restructure. An account rebuilt every month is usually being asked to solve a creative problem with settings.
- Either hired for a product problem. Installs rise, subscribers do not, and the agency's reporting shifts quietly towards installs and click-through rate. You are paying to pour more users into the same leak.
- Either hired for a measurement problem. Platform ROAS improves while finance sees no change, because the agency is optimising towards a number that does not describe revenue.
If you already have an agency and recognise one of these patterns, our guide to auditing a mobile app marketing agency turns them into a scorecard.
When the answer is neither
Three situations call for neither agency. The first is the product case above: fix onboarding, the paywall or the offer, and come back when users convert after the install. The second is the measurement case: no agency should be scaling spend against numbers you cannot trust. The third is budget. If your account cannot afford to read a handful of concepts a month, a full agency fee buys opinions rather than evidence, and our guide to how much ad spend an app needs before hiring an agency sets out the cheaper alternatives.
When it is both: who owns the read
Plenty of apps need both disciplines, and the choice then becomes one partner or two. RocketShip HQ makes a fair point in its own comparison of the two agency types: the thing that matters most is who owns the read on whether a creative worked. Two good partners with nobody reading results across both will underperform one average partner who does.
If you go with two, write down who owns each seam before the work starts: the ad's promise against the paywall, platform numbers against revenue, and fatigue against saturation. Our comparison of specialist UA agencies and full-service app agencies lays those seams out, and our guide to performance creative agencies vs UGC agencies covers the creative side of the same decision. For the shortlist itself, see our rankings of user acquisition agencies and performance creative agencies for mobile apps, and our guide to hiring a performance creative agency for how to run the selection.
Where we fit
We run both sides for mobile apps: the Meta Ads management and the performance creative that goes into it, read by the same team. We built it that way because in our experience the handover between the two is where accounts lose most. Our Preggers case study is a qualitative example: a more diverse UGC creative structure brought CPI down and made the US a profitable market. The page does not publish the underlying figures, so we do not quote any here.
We are the wrong buy if you already have a strong in-house media buyer and only want concepts, because our published tiers bundle Meta management with creative. We are also the wrong buy for the product and measurement cases above. Our pricing starts at £1,500 a month for accounts spending up to £20,000 a month on ads, with a £3,000 minimum monthly ad spend. If you have run the checks above and the answer is creative, media or both, you can apply to work with us and we will tell you on the first call which one we think it is.
Frequently asked questions
Should a mobile app hire a creative agency or a media buying agency?
Hire for the constraint you can prove. If your ads convert well after the install, your tracking agrees with your revenue system and your account structure lets Meta learn, but a few ageing ads carry most of the spend and new concepts rarely beat them, you need creative. If good concepts exist but sit in ad sets that never collect enough optimisation events, budgets are split too thin or the optimisation event is wrong, you need media buying. If neither describes you, check the product funnel and measurement before hiring either.
How do I know if my Meta Ads problem is creative or media buying?
Compare cost per trial or cost per subscriber across ad concepts inside the same campaign and optimisation event, over a period long enough for the results to mature. A wide spread, where one or two old concepts carry the account and new ones lose, points to creative supply. A narrow spread where everything performs similarly badly, combined with ad sets that cannot reach about 50 optimisation events a week, points to account structure first.
Does media buying still matter with Advantage+ app campaigns?
Yes, but the job has changed. Meta's engineering team describes Advantage+ as automating budget allocation, audience targeting and placement, so the manual targeting work that used to define media buying matters less. What remains is choosing the optimisation event, keeping the structure consolidated enough to learn, avoiding unnecessary significant edits, deciding when to scale, and reading results against cohort revenue rather than platform numbers.
Can an agency fix low install-to-trial or trial-to-paid rates?
Only partly. Creative can lift install-to-trial when the ad attracts people who want what the onboarding sells, and a mismatch between the ad's promise and the paywall can be fixed from the ad side. But if every concept converts poorly after the install, the constraint is usually onboarding, the paywall, the offer or the product, and the work belongs to your product team or a monetisation specialist rather than a creative or media agency.
Is it better to hire one agency for both creative and media buying?
It depends on whether you already have a strong media buyer. One partner for both closes the loop between what gets made and what gets spent, which matters most when nobody in house reads creative results. If you already employ a capable media buyer, a creative-only partner can be the better buy, provided someone clearly owns the brief and the read on each test.
The short version
Check your numbers, then your funnel, then the spread of cost per trial across concepts. Old winners carrying the account with nothing to replace them means creative. Starved, constantly edited ad sets mean media, and that fix comes first because it is cheap and it makes every later creative test readable. If both checks pass and costs are still high, the problem probably is not the ads.