There is no right number of ads for a mobile app agency to make each month. There is a right number of new concepts, and it comes from three things about your account: how quickly your winning ads wear out, how often a new idea turns into a winner, and how many new ideas your budget can test properly. Count concepts first. Hooks, sizes and languages multiply each concept into files, so the file count follows from the concept count and should never be the target. In the hypothetical example below, an app spending £30,000 a month needs about nine new concepts a month. The same app with ads that wear out twice as fast needs eighteen and cannot afford to test them.
This guide gives the formula for the number you need, shows how it changes when one input moves, explains what to do when you need more than you can test, and ends with the questions that reveal whether an agency's monthly output is tied to your account or to its own production line. It is part of our guide to choosing a mobile app marketing agency.
Key takeaways
- Count three different things separately: new concepts, iterations on winners, and asset files. Most confusion about creative volume comes from mixing them up.
- The number of new concepts you need is set by winner lifespan, hit rate and how much spend one winner can carry.
- The number you can use is capped by your testing budget divided by the cost of reading one concept.
- When need is bigger than capacity, more production does not help. Longer-lived winners, better ideas or a cheaper read do.
- Judge an agency on concepts tested and winners found, not files delivered.
Why there is no universal number
Search for this question and you will find answers anywhere from five to a hundred and more ads a month. They can all be right for someone, because they are counting different things for different accounts. A gaming app whose ads fade inside a fortnight, a meditation app whose winners run for two months and a language app running the same ad in six languages will need very different numbers, and none of them is wrong.
The first step is to agree what is being counted. These three units behave differently and cost different amounts to judge.
| Unit | What it is | Example | How quickly it can be judged |
|---|---|---|---|
| Concept | A distinct reason to download: a different problem, motivation or emotional angle | Learning a language to talk to your partner's family, rather than for a holiday | Slowly. Needs a full testing budget and several days of delivery |
| Iteration | A new hook, opening, creator or edit on a concept that has already won | Five new first lines on last month's winning video | Faster and more cheaply, because the concept is already proven |
| Asset file | Each exported file: every size, length and language of every version | One video cut in 9:16 and 4:5, in English and German, is four files | Not judged on its own. It is packaging |
Meta draws the same line. Its help page on creative fatigue recommendations says the fix for a fatigued ad is to create a new ad with a new image or video that is materially different from the original creative. A new crop of the same video is not that. Why the delivery system treats near-identical ideas as one ad is explained in our guide to creative diversity versus volume. The rest of this guide counts concepts first and treats iterations and files as what each concept turns into.
How many new concepts you need: the replacement formula
Your account needs enough winners live at any time to carry the spend you want to run. Each winner wears out after a while. So each month you must find enough new winners to replace the ones fading, and because most concepts lose, you must test several times that number.
Live winners needed = spend on proven ads ÷ spend one winner can carry before its cost per result rises
New winners needed a month = live winners × (4.33 ÷ winner lifespan in weeks)
New concepts to test a month = new winners needed ÷ hit rate
4.33 is the number of weeks in an average month. Every input can be measured from your own account history, and should be, because none of them has a reliable public benchmark.
- Winner. Define it before you count. A workable definition is a concept that held a meaningful share of spend for at least two weeks at or under your target cost per trial.
- Lifespan. The weeks from a winner's launch until its cost per result rises well above its own early level and stays there. Our guide to creative fatigue for mobile apps covers how to tell fatigue from other causes. On Advantage+ app campaigns you have to measure this yourself: Meta's help page says its fatigue status is not available for them.
- Carry. How much a month one winner absorbs before cost per result climbs. It grows with audience size and the number of markets the ad can run in.
- Hit rate. The share of properly tested new concepts that became winners over the last three to six months. We have not found a public benchmark measured the same way across apps, so the 1 in 5 used below is an assumption, not a norm.
How many concepts you can test: the capacity ceiling
The other side of the calculation is what your media budget can judge. A concept is only worth making if it gets enough delivery to tell a winner from noise, and that delivery is paid for in ad spend, not in the creative fee.
Concepts you can read a month = testing budget ÷ cost to read one concept
Cost to read one concept = results needed to judge it × cost per result
Our working rule, set out in our guide to Meta Ads creative testing budgets, is to hold about 15% of spend for testing and give each new concept enough budget for twenty to thirty conversions before judging it. Our performance creative pricing guide turns that into a table of readable concepts at five spend levels, so we will not repeat it here. This guide adds the demand side, because capacity alone tells you the most you could use, not what you need.
Worked example: an app spending £30,000 a month
Every figure here is hypothetical. It is an illustration, not client data and not a benchmark. The app spends £30,000 a month on Meta. It keeps 15%, or £4,500, for testing, so £25,500 runs on proven ads. One winner can carry about £7,500 a month, so the account needs about 3.4 winners live at any time. Winners last about eight weeks. One in five tested concepts becomes a winner. It judges concepts on 25 trials at £20 each, so reading one concept costs £500.
- New winners needed a month: 3.4 × (4.33 ÷ 8) = 1.84
- New concepts to test a month: 1.84 ÷ 0.2 = 9.2
- Concepts the testing budget can read: £4,500 ÷ £500 = 9
Need and capacity roughly match, so the answer for this app is about nine new concepts a month. Now change one input at a time.
| Scenario | What changes | Concepts needed a month | Concepts it can read | Result |
|---|---|---|---|---|
| Baseline | Winners last 8 weeks, 1 in 5 concepts wins, a winner carries £7,500 a month, £500 to read a concept | 9.2 | 9 | Roughly balanced |
| Faster fatigue | Winners last 4 weeks instead of 8 | 18.4 | 9 | Can only test half of what it needs |
| Weaker ideas | 1 in 10 concepts wins instead of 1 in 5 | 18.4 | 9 | Same gap, different cause |
| Iterating on winners | New hooks stretch a winner's life from 8 to 12 weeks | 6.1 | 9 | Spare capacity for bolder tests |
| Bigger audience | A winner can carry £15,000 a month instead of £7,500 | 4.6 | 9 | Fewer, larger winners |
| Deeper read event | Concepts judged on first payments at £60 each instead of trials at £20 | 9.2 | 3 | Reads a third of what it needs |
Three things stand out. First, spend is not in any of the sensitive rows. If all the ratios held, an app spending twice as much would need twice as many concepts and could read twice as many, so the balance would not change. What decides whether a creative programme keeps up is fatigue, idea quality and the cost of a read, not the size of the budget.
Second, two very different problems produce the same gap. An app whose winners fade in four weeks and an app whose ideas rarely win both need about eighteen concepts a month. The first needs a plan for making winners last. The second needs better ideas, and producing more of the same ideas will not fix it.
Third, the deeper the event you judge on, the fewer concepts you can read. The choice of event is covered in our guide to optimising for installs, trials or purchases. For creative testing, it is often sensible to judge a first cut on cost per trial and confirm the survivors on paid conversion, as long as you check that cheap trials are not coming from people who never pay. Our guide to why a lower CPI can make an app less profitable shows how that goes wrong.
When you need more concepts than you can test
This is the common case for apps in fast-fatiguing categories and for any account whose recent ideas have stopped winning. Asking the agency to make more ads is the instinctive answer and usually the wrong one, because the extra concepts cannot be read. These levers close the gap instead.
- Make winners last longer. New hooks, openings and creators on a proven concept can be judged quickly and cheaply, and in the example stretching lifespan from eight to twelve weeks cut the need from 9.2 to 6.1 concepts. Our guide to hook patterns for mobile app ads is the starting point.
- Raise the hit rate. Better research and a wider spread of angles do more than more production. We map concepts across psychological dimensions with the TSO Creative Framework, so new tests land in angles the account has not tried rather than in the one it has worn out. Check, too, that the ad and the paywall sell the same thing: ad-to-paywall message match explains why a mismatched concept can lose for reasons that have nothing to do with the ad.
- Make each read cheaper. Judge first on an earlier event, then confirm on revenue, as above.
- Raise the testing share for a period. Moving from 15% to 20 or 25% for a month or two is reasonable when the account is rebuilding after its winners have collapsed. It is not a permanent setting.
- Give winners more room. A broader audience or an extra market can raise how much one winner carries. Our guide to scaling Meta Ads for a subscription app covers the trade-offs.
The opposite case also happens. If your winners last a long time and your budget can read more concepts than you need, you do not need to buy more. Spend the spare testing capacity on bolder angles that might find a new audience, or keep the money in proven ads.
From concepts to a month of files
Once the concept number is set, the rest of the month's output follows from it. Continuing the hypothetical £30,000 app:
- New concepts: 9, each made in about three hooks and two sizes, so roughly 54 files.
- Iterations on winners: three or four new hooks or edits for each of the three or four live winners, so another 10 to 16 versions, each in two sizes.
- Markets: every extra language multiplies the files but not the concepts. A concept that wins in one language still has to be checked in another, because the angle may not travel.
- Formats: making a static and a video of the same idea is useful, but it is still one concept. A new format alone does not make an idea materially different.
So the same month can honestly be described as "9 concepts" or "about 80 ads". Both are true. Only the first tells you whether the account is getting enough new ideas.
Our AI UGC page describes one rebuild for Tonaly in exactly these terms: 40 AI variants briefed on three distinct customer pain points pulled from App Store reviews. That is three concepts and forty files, and it is the three that did the work.
How often to launch new creative
Volume also has a delivery cost. Meta's help page on significant edits lists adding a new ad to an ad set as a significant edit that can send it back into the learning phase. Its page on the learning phase says ad sets usually exit learning after about 50 results in the week after the last significant edit. Dripping a few new ads in every day keeps resetting that clock.
The practical answer is a fixed launch day, weekly or fortnightly, when a batch of concepts and iterations goes live together, with losers paused so the account does not fill with dead ads. Meta's developer documentation on ad volume notes that ads running or in review count against a per-Page ad limit, so an ever-growing library has a ceiling as well as a cost. How a fixed launch rhythm turns into more test rounds a quarter is worked through in our guide to whether the same agency should run your Meta Ads and your creative, and the Meta Ads learning phase guide covers the rest.
How to judge the monthly number an agency offers
If you are comparing agencies, the output number in a proposal is where the formula above becomes practical. These are the claims we see most often and the question that tests each one.
| What the proposal says | What to ask | A good answer includes |
|---|---|---|
| "We deliver 50 ads a month" | How many distinct concepts is that, and how do you define one? | A concept count, with the files per concept explained |
| "Unlimited creative" or a large fixed quota | How many new concepts can my testing budget read each month? | A number tied to your spend and your cost per result |
| "We refresh creative every week" | Refresh with new concepts, or new hooks on old ones? | Both, with a stated split that changes as winners age |
| The same output at every price tier | Why would an app spending £15,000 and one spending £75,000 need the same number? | Output that scales with spend, fatigue and markets |
| A monthly report listing files delivered | How many concepts did we read this month, and how many won? | Concepts tested, winners found and the hit rate over time |
A fixed quota is not automatically bad. It can be a fair way to price production. The problem is a quota that does not change when your fatigue rate, hit rate or spend does. Paying for concepts your budget cannot read buys files that age on a hard drive, a point our guide to ad spend before hiring an agency makes from the budget side.
How we set the number
Our pricing page says we do not sell a fixed number of ads, and that "the output volume follows the gaps, not a quota". In practice that means the formula above: we look at which angles the account has covered, which winners are fading and what the testing budget can read, then brief the next round from what won. That is what our performance creative work is built around.
Our LingoLooper case study is a fair example of the approach: a focused creative strategy, a simpler account structure that concentrated spend on what performed, and constant testing, with revenue up 49% and trials up 116% in two months. Nothing in it depended on producing a large number of files.
For a quick estimate before you do the full calculation, our creative refresh calculator turns your spend, vertical and audience size into a monthly concept number. If your app is already spending on Meta and the numbers above suggest you need more ideas than you are getting, you can apply to work with us.
Frequently asked questions
How many new ad creatives does a mobile app need each month on Meta?
There is no universal number. Work out how many new winning concepts you need a month to replace the ones wearing out, divide by the share of tested concepts that become winners, and compare the result with how many concepts your testing budget can read. In our hypothetical worked example, an app spending £30,000 a month needs about nine new concepts a month, delivered as several hooks and sizes each, plus new hooks on the current winners.
Is 50 ads a month a good output from an agency?
The number on its own tells you nothing. Fifty files can be five concepts with ten versions each or twenty-five concepts with two versions each, and those are very different months. Ask how many distinct concepts the output contains, how many are new hooks on proven winners, and how many of the new concepts your spend can actually test properly.
What counts as a new concept rather than a variation?
A new concept gives a different person a different reason to care: a new problem, a new motivation or a new emotional angle. A new creator, setting, edit, hook line or aspect ratio on the same idea is a variation. Meta's own advice for a fatigued ad is to create a new ad with a new image or video that is materially different from the original creative.
How often should new creatives be launched?
In regular batches, not a few at a time every day. Meta lists adding a new ad to an ad set as a significant edit that can send the ad set back into the learning phase, and says ad sets usually exit learning after about 50 results in the week after the last significant edit. A fixed weekly or fortnightly launch day keeps learning resets predictable and makes results easier to compare.
Does a smaller budget need fewer creatives?
It can read fewer, which is the binding limit. At low spend the testing budget may only be able to judge one to three new concepts a month, so every one has to be a considered bet on a genuinely different angle. Producing more than your budget can read adds files, not learning.
The short version
Ask how many new concepts a month, not how many ads. You need enough new winners to replace the ones wearing out, which depends on how long winners last, how much spend each can carry and how often a new idea wins. You can only use as many concepts as your testing budget can read. When the first number is bigger than the second, the answer is longer-lived winners, better ideas or a cheaper read, not a bigger production order. Hooks, sizes and languages then turn each concept into files, and the file count is the least informative number in the report.