In the first 90 days, a mobile app marketing agency should spend roughly the first month making sure your measurement can be trusted, learning who your customers are and getting a first wave of genuinely different ad concepts live by about week three. The second month should be structured testing against written hypotheses, read on trial starts rather than installs. The third should be iterating the winners and raising budget in steps. What you should hold at day 90 is a working process, an honest record of wins and losses, and a first read of the cohorts that are old enough to judge. For a subscription app that is only about a month of installs, because paid conversions keep arriving for weeks after the install.
Our guide to hiring a performance creative agency sketches a healthy first quarter from the creative side. This guide is the full operating plan for an app on Meta: what the agency should do in each phase, what you have to provide, why the subscription cohort clock makes day 90 earlier than it looks, a scorecard for the 30, 60 and 90 day checkpoints, and the things that would worry us if we were the one buying.
Key takeaways
- The first deliverable is a measurement audit, not an ad. An agency that launches before checking the tracking cannot read its own tests.
- Judge the quarter at three checkpoints with different questions, not with one verdict on day 90.
- About one in five paid conversions arrives six weeks or more after install, so on day 90 only cohorts acquired before about day 50 to 55 are close to complete.
- Half the success of the quarter depends on the client: fast access, a named owner and a paywall that is not changing every fortnight.
- No one can promise a result in 90 days. An agency can promise a process, and you should hold it to one.
The 90-day plan at a glance
The three phases overlap in practice, and the boundaries move with your budget. An account that can read only a few concepts a month moves through them more slowly than one that can read a dozen. But the order should not change.
| Phase | What the agency should do | What you need to provide | The checkpoint question |
|---|---|---|---|
| Days 1 to 30: audit, research, baseline | Measurement audit, funnel baseline, customer research, testing plan, first wave of concepts live by about week three. | Every access request granted in week one. A named owner. Answers to research questions within days, not weeks. | Is the tracking trustworthy, and are genuinely new ideas in market? |
| Days 31 to 60: testing and learning | Structured tests on written hypotheses, weekly cohort reporting, second wave briefed from what the first wave showed. | Hold the paywall, pricing and onboarding steady, or log every change with a date. | Can the agency say which ideas are winning on trial starts and why? |
| Days 61 to 90: iterate and scale | Variations of proven concepts, new angles still entering, budget raised in steps, first read of matured cohorts, next-quarter plan. | Agree the scaling rules and the budget ceiling in advance. | Do the earliest cohorts pay back on the path your target CAC needs? |
Why day 90 comes early for a subscription app
Most agency onboarding guides assume a business where a click turns into a sale the same day. A subscription app does not work like that, and it changes what a fair 90-day review can conclude.
Trial starts are fast. Adapty's State of In-App Subscriptions 2026 reports that 89.4% of trial starts happen on the day of install, so a test read on trial starts can produce an answer within days. Paid conversions are slow. RevenueCat's State of Subscription Apps 2026 finds that 50.6% of paid conversions happen on day 0 but 19.2% arrive in week six or later. Both of those are shares of eventual paid conversions, not of installs. In plain terms, a cohort is only about four-fifths converted once it is five weeks old.
On iOS there is a second clock. Apple's documentation on postbacks in multiple conversion windows sets three windows, days 0 to 2, 3 to 7 and 8 to 35 after first launch, and sends each postback after a random delay of 24 to 48 hours for the first and 24 to 144 hours for the second and third. So the last postback for an install can arrive as late as day 41. Apple also notes that apps in the lowest data tier receive only the first postback, which is why your own revenue data matters more than the ad platform's view for anything past the first week. We cover those delays in more detail in our guide to SKAdNetwork postback delays.
Put the two clocks against a 90-day engagement. The table assumes the first concepts go live on day 21, which is a realistic pace after a proper audit.
| Cohort installed on | About 81% of paid conversions in by | Last iOS postback by | Status at day 90 |
|---|---|---|---|
| Day 21 (first concepts live) | Day 56 | Day 62 | Readable |
| Day 35 | Day 70 | Day 76 | Readable |
| Day 49 | Day 84 | Day 90 | Only just readable |
| Day 60 | Day 95 | Day 101 | Incomplete |
| Day 75 | Day 110 | Day 116 | Incomplete |
The paid column adds 35 days to the install date, using RevenueCat's overall timing. Your own category will differ: RevenueCat reports that Productivity apps convert 71.9% on day 0 and Education only 28.5%. The postback column adds Apple's maximum of 41 days. Either way the result is the same shape.
A 90-day engagement produces roughly four to five weeks of subscription cohorts that are mature enough to judge on revenue. Everything acquired after about day 50 is still converting when the quarter ends.
This is not an excuse for an agency to avoid accountability. It is a reason to judge each checkpoint on the evidence that exists at that point: tracking and process at day 30, trial starts and early paid conversion at day 60, and matured cohorts at day 90. An agency that claims a revenue verdict on the whole quarter at day 90 is reading incomplete data, whether the verdict flatters it or not.
Days 1 to 30: audit, research and a baseline
Week one: access, and nothing launched yet
The first week is access. A competent agency will send one list on day one and chase it. Expect to grant:
- Partner access to your Meta Business portfolio, the ad account and the app or dataset your events flow into, without transferring ownership of any of them.
- Read access to your mobile measurement partner, if you use one.
- Read access to your subscription revenue data, for example RevenueCat or your own backend reporting.
- Read access to App Store Connect and Google Play Console analytics.
- Your existing ads, raw files, past performance exports and any brand rules.
Ownership matters from the start. Your ad account, dataset and creative files should stay yours, with the agency added as a partner. That makes the relationship easy to end later, which is exactly why a good agency will insist on it.
Weeks one to two: the measurement audit
This is the most valuable and least glamorous work of the quarter. The agency should confirm that installs, trial starts and purchases reach Meta once each, with the right values and without duplicates from two sources sending the same event. It should check that iOS conversion values are mapped to events that predict revenue, and that the numbers in Meta, your measurement partner and your revenue tool are close enough to reconcile. It should also find the event your campaigns can realistically optimise for at your budget, given that Meta's Business Help Centre says an ad set usually leaves the learning phase after about 50 results in the week after its last significant edit.
The output should be written down: what was checked, what was broken and what was fixed. If tracking needs real engineering work, that becomes the first deliverable by agreement, not something the agency discovers in week five. We explain why the platforms will never agree exactly in our guide to CPI, CPA and CAC for subscription apps, which also sets out which number to act on.
Weeks one to three: research and the baseline
Research for an app is concrete. Read the App Store and Google Play reviews, the support inbox and the cancellation reasons. Walk through onboarding and the paywall as a new user would. Look at what competitors are running in the Meta Ad Library. The point is to find the problems your customers describe in their own words, which become the raw material for concepts, and to map where your current ads already sit so the first wave covers new ground. We map that coverage with the TSO Creative Framework.
In parallel the agency should freeze a baseline: cost per install, install-to-trial, trial-to-paid, cost per trial and early revenue per install, split by platform and country, for the eight to twelve weeks before it arrived. Without a baseline, nobody can say at day 90 what changed. Our install-to-trial benchmarks are a useful outside reference for where your funnel stands.
Week three: the first wave goes live
By about week three the first concepts should be in market. The word that matters is concepts. Five ads making five different arguments to five different kinds of customer teach you far more than twenty variations of the ad you already run, a point we make at length in creative diversity versus volume. Each concept should come with a sentence saying what it is meant to prove.
Days 31 to 60: testing and learning
The second month is the core loop: run the tests, read them properly, and brief the next round from what they showed. How many concepts the account can read depends on the budget and the cost of the event you read on, which we work through in our guide to how much ad spend an app needs before hiring an agency. Three habits separate a good second month from a busy one.
Patience with new tests. A concept in an untested angle often looks weak for its first few days while delivery settles, and on iOS the most recent two or three days are always the least complete. Our rules for when to turn off a Meta ad set minimum spend and time before any verdict.
Reading on the right event. Installs are where cheap, low-intent traffic hides. For a subscription app the minimum useful read is cost per trial start, and the better read is trial-to-paid by cohort as soon as the cohorts are old enough.
A test log, not a highlight reel. By day 60 you should be able to see every test that ran, the hypothesis behind it and the result, including the ones that lost. Losers are information. An agency that only shows you winners is either not testing or not telling you.
This is also the month to watch your own side of the bargain. If the paywall, the price or the onboarding changes in the middle of a test, no one can say whether the result came from the ad or the product. You do not have to freeze the product for a quarter, but every change should be logged with a date so the agency can separate the effects.
Days 61 to 90: iterate, scale and read the first cohorts
By the third month, a few concepts should have earned the right to more money. The work shifts to variations of those winners, new hooks and formats on proven arguments, while genuinely new angles keep entering at a smaller share of budget so the account is not relying on one idea when it fatigues.
Budget should rise in steps, with cost per trial watched after each one, rather than in a single jump because the money was available. Before raising it, agree the ceiling: the cost per subscriber at which growth stops paying, which our guide to setting a target CAC shows how to work out.
Then comes the first honest revenue read: the cohorts acquired in roughly weeks three to seven, which by now are old enough to judge. Compare their revenue per install at matched ages with the baseline and with the path your payback target needs, rather than reading a blended ROAS figure that mixes old organic subscribers with new paid ones. Our subscription app ROAS benchmarks explain how to read D7, D30 and D60 against each other.
The quarter should end with a plan for the next one, built on named angles that won and named ones that did not, plus a view on whether the main constraint is still acquisition. Sometimes the honest finding is that it is not. Our Steps & Beasts case study credits the result to heavy creative testing and to onboarding changes that lifted install-to-subscriber conversion. The app needed both stronger creative and a better path from install to subscription, and an agency should be willing to say so.
A 30/60/90 scorecard
Use this at each checkpoint. One concerning sign is a conversation. Several at the same checkpoint is a pattern.
| Checkpoint | Good signs | Concerning signs |
|---|---|---|
| Day 30 | A written measurement audit with fixes made. A frozen baseline by platform and country. Research you recognise as your customers. Five or more distinct concepts live. | Ads live in week one with no audit. A plan with no hypotheses. Every concept a variation of your current best ad. Reporting installs and CPI only. |
| Day 60 | A test log listing what ran, what it was meant to prove and what happened, including the losers. Reads on trial starts or a deeper event, by cohort. A second wave visibly briefed from the first. | Blended ROAS as the headline. Tests killed after a day or two. Repeated restructures. No explanation of why a winner won. |
| Day 90 | A first read of cohorts old enough to judge. Budget raised in steps with cost per trial holding. A next-quarter plan built on named winning angles. | A verdict on revenue from cohorts that are still maturing. Budget raised because spend was available, not because tests supported it. A plan that is the first plan again. |
What would concern us if we were the buyer
We sit on the agency side of this table, so here is the list we would use if we were hiring one. None of these is proof of a bad agency on its own, but each deserves a direct question.
- Ads before an audit. Launching in week one feels fast and usually means tests that cannot be read.
- A restructure on day two. Rebuilding the account before the baseline is frozen makes every later result impossible to attribute.
- Install-only reporting. A falling CPI with no trial or cohort data next to it is the easiest number in the account to improve and the least useful.
- Asking you for nothing. An agency that never asks for reviews, support themes or revenue data is not doing research.
- A guaranteed number. A promised ROAS or cost per subscriber is either heavily caveated or bought by running only safe, low-volume tactics.
- No account of the losers. If nobody can tell you which test is running, why, and what the last three taught you, there is no process to judge.
Several of these overlap with the questions to ask a Meta Ads agency before you sign, and asking those questions up front is the cheapest way to avoid finding the answers at day 60.
What 90 days cannot tell you
A quarter can show whether the process works, whether the tracking is sound and whether new concepts are finding customers at a cost your economics can bear. It cannot show long-term subscriber value, because a monthly subscriber acquired in week six has had at most one renewal by day 90 and an annual subscriber has had none. It also cannot separate a good agency from a good season with full confidence. Treat day 90 as the end of the setup and the first read, not the final verdict. If the agency was hired before the funnel was ready, our guide to when a mobile app should hire a Meta Ads agency is the place to check that first.
How we run the first quarter
Our engagements start with a creative and account audit, and our pricing page sets out the rest: most accounts see a meaningful signal within 30 to 60 days and a compounding effect from month three, the first cycle is research, coverage and a clean baseline, and we ask for an initial three-month commitment because that is the shortest honest window for research, launch and a proper read. After that, clients can leave with 30 days' notice. Our Meta Ads service combines the creative and the campaign management in one team, so the person reading the test is the person who briefed it.
If your app is spending enough to read tests and you want a first quarter that runs like the one above, see how we work with consumer and subscription apps, read our case studies, and apply to work with us. For where paid acquisition sits in the wider funnel, start with our subscription app marketing strategy guide.
Frequently asked questions
How long should you give a new app marketing agency before judging results?
Around 90 days, judged at three checkpoints rather than once at the end. At day 30 judge the groundwork: tracking, research and a first wave of concepts in market. At day 60 judge the testing process and the early trial-start data. At day 90 judge the first fully matured cohorts. For a subscription app, cohorts acquired in the last five or six weeks of the quarter are still incomplete on day 90, so a final verdict on revenue usually needs a few more weeks.
What should an app marketing agency deliver in the first 30 days?
A written measurement audit confirming that installs, trial starts and purchases reach Meta and your revenue tool correctly, a baseline of your current funnel by platform and country, customer research drawn from reviews, support messages and your onboarding, a testing plan with stated hypotheses, and a first wave of genuinely different ad concepts live by about week three.
Can a mobile app marketing agency guarantee results in 90 days?
No honest agency can. Nobody controls the auction, your product, your paywall or your pricing. What an agency can commit to is a process: a stated volume of new concepts, a testing method, reporting on cohorts rather than blended totals, and a clear record of what won and what lost. Be wary of a guaranteed cost per acquisition or ROAS.
What access does an app marketing agency need on day one?
Partner access to your Meta Business portfolio, ad account and app or dataset; read access to your mobile measurement partner if you use one; read access to your subscription revenue data, such as RevenueCat; read access to App Store Connect and Google Play Console analytics; and your existing creative files with past performance. Delayed access is one of the most common reasons a first month slips.
Should a new agency restructure our Meta account straight away?
Usually not in the first week. Restructuring before the measurement audit and the baseline are done makes it impossible to tell later whether results changed because of the new structure, the new creative or broken tracking. Fix tracking first, record the baseline, then change structure deliberately and say in writing why.
The short version
Month one is access, a written measurement audit, customer research, a frozen baseline and a first wave of different concepts by week three. Month two is structured testing read on trial starts, with a log of every test including the losers. Month three is iterating winners, raising budget in steps and reading the first cohorts that are old enough to judge. For a subscription app that means roughly the installs from weeks three to seven, so hold the agency to a process at day 90 and to revenue a few weeks after it.