For non-gaming apps actively growing paid acquisition, Pantheon recommends starting with 25-30% of monthly ad spend for creative testing. With the assumptions below, a $15,000 monthly ad budget funds initial tests of 30 creatives and deeper tests of three promising ads.
For indie developers and app studio teams, the goal is to turn that budget into a repeatable process: produce different ideas, learn which attract valuable users, and give the strongest ads more investment.
Here is how to calculate the volume you can support - and make room for more testing as you grow.
First, separate the two types of test
An initial creative test gives an ad a starting budget to assess attention, clicks, and early conversion signals.
A deeper performance test gives promising ads more budget to evaluate trials, paid subscriptions, retention, or revenue.
Both cost money. Testing 30 ads initially does not mean you have established the business performance of all 30. Plan for both stages, and keep creative production costs separate from ad spend.
What percentage of your budget should go to testing?
Published recommendations vary because teams define testing differently and work with different accounts.
| Source | Published recommendation | What we take from it |
|---|---|---|
| AdManage | 10–20% for ongoing testing | Reserve budget for regular learning |
| AdsX | 25–30% at $5K–$15K monthly spend; 10–15% above $150K | Consider actual dollars alongside percentages; this framework addresses ecommerce |
| Aden’s Lab | 40–60% for new accounts; 10–30% for stable/scaling accounts | The account’s stage changes its needs |
| The Social Outline | Starts around 15%, adjusted for replacement needs | Budget for finding the next effective ads |
These are practitioner recommendations, not a universal standard. Our starting point for non-gaming apps is:
- 15-20%: Stable acquisition with several proven ads.
- 25-30%: Active growth, new audiences, or more frequent creative replacement.
- 40%, reviewed regularly: Launching or rebuilding performance.
An account without established winners may need more discovery spending. Review the split after each testing cycle rather than applying the same percentage forever.
How many creatives can that budget support?
Start with the cost of an initial test. CPM means the cost of 1,000 ad impressions.
Initial test cost = Planned impressions ÷ 1,000 × CPM
At 5,000 impressions and a $15 CPM, the initial test costs $75 per creative. These are example inputs; replace them with costs and evidence requirements appropriate for your account.
Next, reserve money for deeper tests. Suppose 10% of creatives receive another $750 each:
Average testing budget per creative = $75 + (10% × $750) = $150
Monthly creative count ≈ Total testing budget ÷ $150
Using a 30% testing allocation:
| Monthly ad spend | Total testing budget | Initial creative tests | Ads funded for deeper testing |
|---|---|---|---|
| $5,000 | $1,500 | 10 | 1 |
| $15,000 | $4,500 | 30 | 3 |
| $50,000 | $15,000 | 100 | 10 |
| $150,000 | $45,000 | 300 | 30 |
For the $15,000 account, that is 30 × $75 + 3 × $750 = $4,500. Both stages are included.
The 10% selection rate is a planning assumption, not a winner rate. If 20% advance, the average budget becomes $225 per creative, and $4,500 funds 20 initial tests plus four deeper tests. Reserve additional control-ad spend when a formal comparison requires it.
Pantheon recommendation: use earlier funnel metrics for initial decisions
Use video engagement, click-through rate (CTR), and install signals to decide which creatives deserve deeper testing. These events usually occur more often and sooner than paid subscriptions, allowing faster feedback at lower cost.
For common low-rate events, a higher baseline rate generally means fewer observations are needed to detect the same relative improvement, holding confidence and statistical power constant. Funnel position alone does not determine sample size.
For example, our approximate two-sided calculation at 95% confidence and 80% power requires:
- 1% > 1.5%: About 7,750 observations per version.
- 0.1% > 0.15%: About 78,400 observations per version.
Both represent a 50% relative improvement. The calculation assumes independent observations and equal groups; compare rates using the same observation unit. Evan Miller’s calculator lets you explore these inputs.
Use early results to prioritize ads, then check whether those signals predict trials, payment, and retention. Five thousand impressions are an initial budget assumption, not automatic proof of a winner.
Follow promising ads through to revenue
For a subscription app, follow results beyond the install. Consider this hypothetical example:
| Ad | Cost per install | Install-to-payer rate | Cost per paying user |
|---|---|---|---|
| A | $2 | 2% | $100 |
| B | $3 | 6% | $50 |
Cost per paying user = Cost per install ÷ Install-to-payer rate.
Ad B brings more expensive installs but acquires paying users at half the cost. Check retention and revenue before increasing its budget.
A $750 deeper test could buy approximately 50 trials at $15 per trial - or only 12-13 paying users at $60 each. Neither count automatically proves a winner.
Compare users with equal time to convert. Someone three days into a seven-day trial cannot yet answer the trial-to-paid question. On iOS, also allow for reporting delays: Apple’s SKAdNetwork documentation describes conversion windows followed by delayed reports.
Pantheon recommendation: increase meaningful creative testing
Make increasing the number of properly funded, meaningfully different creative tests a growth priority. More tests give you more opportunities to find an effective message.
At a stable win rate, the expected number of winners increases directly with the number of evaluated creatives:
Expected winners = Creatives evaluated × Win rate
At an illustrative 5% win rate, 20 evaluated creatives yield one expected winner; 60 yield three. This is an average, not a guarantee. Producing more near-identical ads or spreading the same budget too thinly does not preserve that relationship.
There is evidence that substantial creative variety matters in non-gaming app marketing. AppsFlyer’s 2025 research found that high-spending non-gaming apps averaged 2,365 variations per quarter, while the top 2% of non-gaming creatives received 43% of spend. The broader sample included apps running at least 200 variations each, so it is not representative of every indie app. These are production and spend patterns, not proof that volume alone causes growth. AppsFlyer’s report.
Our takeaway: keep producing fresh concepts and useful variations while current ads are working. Build the next batch before declining performance makes it urgent.
Pantheon recommendation: use GEO proxy markets to lower initial test costs
A GEO proxy market is a country used for early testing because advertising costs are lower and its audience is relevant enough to inform your target market.
We recommend considering GEO proxy testing when it lets you evaluate more ideas affordably, followed by deeper testing in the market where you intend to grow. AdManage also describes testing creatives in lower-cost markets as a way to stretch the testing budget.
The arithmetic is straightforward. At a hypothetical $5 CPM, 5,000 impressions cost $25. At $15 CPM, they cost $75. A $750 initial-testing budget could therefore fund 30 creatives instead of ten. This triples initial testing capacity; the target-market testing budget still needs funding.
Choose the proxy by language, audience needs, platform, and device mix - not price alone. Keep country results separate. Compare several ads in both markets, including established ads, to check whether their relative performance carries over.
Use the proxy to explore hooks and messages. Confirm paid conversion and revenue performance in the target market: lower acquisition costs elsewhere do not establish local willingness to pay.
Turn the budget into a production plan
For accounts with proven messages, we recommend starting with 30% new concept executions and 70% variations by creative count. Give new concepts more room when you are still discovering what works.
A concept is the central idea. For a fitness app, “work out in ten minutes” and “follow a plan without deciding what to do” are different concepts. New openings for the same demonstration are variations.
For a productivity app, three concepts with two opening hooks and two demonstrations each create 12 executions. Motion designers can build these from reusable compositions with editable footage, voiceover, subtitles, and end cards. Simple resizing adds files, not new ideas.
Product teams check that the promise matches the app. Performance marketers set budgets and decision rules. Data teams connect each ad to activation, payment, and retention. Label every creative so results can inform the next brief.
For an indie developer, these may all be one person’s responsibilities. Start with one market, one channel, and a batch you can fund. A $450 testing budget could cover three $75 initial tests and another $225 for one promising ad, with limited evidence about paid subscriptions.
Increase production as the testing budget grows. A 100-creative month can mean four weekly batches of 25, with tests continuing long enough to collect useful results.
Build your next creative batch with Pantheon
A testing plan only works when the ads are ready. Research, concept development, variations, localization, and campaign setup all compete for your team’s time.
Pantheon brings those tasks into one workflow for app teams, including Meta campaign setup. Automate repeatable production work and spend more time deciding what to test and what the results mean.
Start free with Pantheon. Add your App Store link and build your first batch of creative concepts and variations for your next test.
Frequently asked questions
How much should I budget for creative testing on Meta ads?
For non-gaming apps in active growth, Pantheon recommends starting with 25–30% of monthly ad spend. Stable accounts can start at 15–20%; launches or performance rebuilds may need 40%, reviewed regularly. These are Pantheon planning recommendations, not Meta requirements. Keep creative production costs separate from media spend.
How many ad creatives should I test per month?
Test as many creatives as you can fund through both initial screening and deeper evaluation. In this guide’s example, a $15,000 monthly ad budget with 30% allocated to testing provides $4,500: enough for 30 initial tests at $75 each and three deeper tests at $750 each. The 10% advancement rate is an assumption, not a guaranteed winner rate.
How much does it cost to test one ad creative?
Initial media cost equals planned impressions ÷ 1,000 × CPM. At 5,000 impressions and a $15 CPM, that is $75 per creative. If 10% of creatives receive a further $750 test, the average testing budget is $150 per initial creative. These are illustrative inputs; replace them with your account’s costs and evidence requirements.
How long should I run a creative test before choosing a winner?
Run the test until it has enough observations for your decision and users have had time to convert. There is no universal number of days. A user three days into a seven-day trial cannot yet establish trial-to-paid performance. Allow for attribution delays and compare cohorts with equal time to convert; 5,000 impressions alone do not prove a winner.
Which metrics matter most when testing app ad creatives?
Use video engagement, CTR and install signals to shortlist creatives, then evaluate trials, paid subscriptions, retention and revenue. In this guide’s example, a $2 CPI with a 2% install-to-payer rate costs $100 per payer, while a $3 CPI with a 6% rate costs $50. The cheaper install is not necessarily the better acquisition.