Every subscription app team we meet is measuring something on the day of install and something else thirty days later, and making decisions with neither. Day 7 ROAS sits in between, and it is the only one of the three you can actually run an account on.
Day 0 tells you almost nothing, because a subscription app collects most of its early revenue after a trial ends rather than at the moment someone downloads it. Day 30 tells you a great deal, and it tells you four weeks after the creative decision you should have made.
Here is why Day 7 is the operating metric, what good looks like in 2026, how creative quality shows up there before it shows up in cost per install, and how to read the number honestly when iOS attribution is lagging behind you.
Key takeaways
- Day 7 ROAS is the revenue a cohort generates in its first week divided by what you spent acquiring that cohort. Cohort is the word that matters, not blended weekly spend against weekly revenue.
- Day 0 understates subscription apps structurally, because trial periods push the money past the install date. Day 30 is accurate and arrives too late to act on.
- Across our health and wellness client portfolio we work to a Day 7 ROAS of 100 to 120 percent. Gaming benchmarks are an order of magnitude lower by design and should never be compared directly.
- Creative quality shows up in Day 7 before it shows up in CPI, because a bad angle buys cheap installs from people who were never going to subscribe.
- On iOS, read Day 7 with a deliberate lag built in. SKAdNetwork postbacks land days after the window closes, so a fresh cohort always looks worse than it is.
What day 7 ROAS is, precisely
Day 7 ROAS is the total revenue generated by a cohort of users during the first seven days after they installed, divided by the media spend that acquired that cohort. Expressed as a percentage, 100 percent means the first week paid for itself.
The definition is simple. The implementation is where accounts go wrong, and almost always in the same place: people divide this week's revenue by this week's spend and call it Day 7. That is blended ROAS. It mixes revenue from users acquired months ago with spend on users who have not converted yet, and it will happily show a healthy trend while every new cohort you buy is unprofitable.
You need the number cohorted by install date. If your analytics cannot do that, fixing it is a higher priority than any creative work you were planning this quarter, because without it you are optimising against a figure that does not describe your acquisition.
Why day 0 lies and day 30 arrives too late
Subscription apps have a monetisation shape that makes install day revenue close to meaningless. Most users enter a trial, and the money arrives when that trial converts. Adapty's 2026 data puts install to trial conversion at 14.5 percent in North America against 11.2 percent globally, which is the first real filter in the funnel and it happens before a penny moves.
Paywall model changes the shape again. Adapty found hard paywalls convert at roughly 5.5 times the rate of freemium, with about double the lifetime value, at the cost of a 1.7 times higher refund rate. Two apps with identical CPI and identical install volume can produce completely different Day 7 numbers purely because of where the paywall sits, which is worth knowing before you benchmark yourself against anyone else.
Day 30 captures all of this properly, and it means a concept launched on the first of the month gets its verdict on the last day of the month, by which point you have spent four weeks of budget on it and it has probably started fatiguing anyway. In fast fatiguing categories that is a full creative cycle spent waiting for information.
Day 7 is the compromise that works. It captures the trial conversion signal for weekly plans and the early cancellation signal for monthlies, it is stable enough to compare cohorts against each other, and it arrives while the creative that produced it is still running.
What good looks like in 2026
Across our own client portfolio in health and wellness, we work to a Day 7 return of 100 to 120 percent. That means the first week of revenue covers acquisition with a modest margin, and everything from day 8 onwards is the return. It is a demanding target and it is the right one for a category where renewal rates are strong enough to make early payback the binding constraint.
Gaming works completely differently and the numbers look alarming if you do not know that going in. Liftoff data puts casual games at a Day 7 ROAS of 7.6 to 7.8 percent and midcore at 4.3 percent globally, because those apps monetise through in app purchases accumulating over months rather than a subscription charge in week one. A casual game recouping 7 percent by day 7 can be perfectly healthy. A subscription app doing the same is dying.
For context on where the revenue ceiling sits, RevenueCat's State of Subscription Apps 2025 reports Health and Fitness leading all subscription categories on revenue per install at $1.21, with a median first year lifetime value of $27.21 across subscription apps and $49.30 for hard paywall apps. Those are different populations from your account, so read them as orientation rather than a target.
Whatever your own number is, sense check it against the acquisition side. FoxData's 2026 analysis puts the minimum viable ratio of lifetime value to cost per install at 1.5 times, and global average CPI now sits at $2.24 on iOS and $1.12 on Android according to Searchlab. If your target implies an LTV that does not clear that ratio against your real CPI, the target is wrong rather than the performance. Our mobile app CPI benchmarks for 2026 breaks down where those install costs land by category and geography, and the health and fitness CPI benchmarks go deeper on the subscription categories specifically.
Creative quality shows up in day 7 before CPI
This is the part most teams miss, and it is the reason Day 7 ROAS belongs on a creative dashboard rather than only a finance one.
A weak creative concept does not usually announce itself with an expensive install. It announces itself with a cheap one. An overpromising hook, a benefit the app does not really deliver, an angle that appeals to people outside your actual use case: all of these produce installs at a perfectly acceptable CPI. Those people open the app once, do not start a trial, and never appear in your revenue.
So the sequence you see in a real account is this. Day 7 on the new concept comes in below the account average while its CPI looks fine. A week or two later the CPI starts drifting up as the algorithm reads the weak downstream signal. By the time cost per install is visibly bad, you have spent a month funding it.
Reading the metric at concept level rather than only account level catches that first move. It also changes what you do about it. A concept with strong Day 7 and rising CPI is fatiguing and wants a refresh, a new execution of the same angle. A concept with weak Day 7 from the start is not fatiguing, it is simply wrong, and refreshing it just produces a second version of the same mistake. That distinction between refresh and retire is one we cover in full in our guide to Meta ads creative fatigue for mobile apps.
Using day 7 decay as a fatigue signal
Track Day 7 by weekly cohort for a single creative concept and you get a decay curve, and that curve is one of the cleaner fatigue reads available to a subscription app.
A healthy concept holds its Day 7 roughly flat across cohorts while spend scales. A fatiguing one produces a steady week over week decline in Day 7 while CPI is still stable, because the audience most likely to genuinely convert on that angle has already been reached and the system is now buying the next best available people.
Two practical rules we use on that curve:
- Three consecutive declining cohorts is a signal, one is noise. Weekly cohorts on a mid sized account are small enough to bounce around. Look for direction across three, not a single bad week.
- Judge the concept, not the ad. If every execution of one angle is decaying together, the angle is exhausted and you need a different psychological position rather than another cut. If one execution is decaying while its siblings hold, that is ordinary asset level wear.
Reading the number under SKAdNetwork lag
On iOS the number you are looking at is not finished. SKAdNetwork postbacks are deliberately delayed, with the first postback arriving at least 24 to 48 hours after its measurement window closes, and later postbacks landing considerably further out. A Day 7 cohort read on day 7 is systematically incomplete, and it always reads worse than the truth rather than better.
The failure mode is predictable and expensive. A team reads a fresh cohort, sees a Day 7 figure well below target, kills the concept, and the postbacks that would have shown it clearing target arrive three days after the ad set was paused. We have watched perfectly good creative retired this way more than once.
The fix is to build the lag into the reporting rather than correcting for it in your head. Hold a fixed window, read Day 7 ROAS only for cohorts whose attribution has had time to land, and compare like for like across cohorts at the same maturity. The mechanics of those windows, and how much distortion each one introduces, are covered in our guide to SKAdNetwork postback delays.
Android gives you the number faster and more completely, which is a good reason to use Android cohorts as your early creative read and iOS as the confirmation, provided you remember the two platforms buy different people at different prices.
Frequently asked questions
What is a good day 7 ROAS for a subscription app?
It depends entirely on your monetisation model and your payback target, so there is no universal number. Across our own health and wellness client portfolio we work to a Day 7 ROAS of 100 to 120 percent, meaning the first week of revenue covers the acquisition cost with a little headroom. Gaming operates on a completely different shape: Liftoff data puts casual games at 7.6 to 7.8 percent and midcore at 4.3 percent globally, because those apps recoup over months of in app purchases rather than in the first week.
How do you calculate day 7 ROAS?
Take the revenue generated by a cohort of users in the first seven days after install, then divide it by what you spent acquiring that cohort. The critical word is cohort. You are measuring the same group of users across time, not dividing this week's revenue by this week's spend, which is a blended figure that tells you almost nothing about whether your acquisition is working.
Is day 7 ROAS better than day 30 ROAS?
They answer different questions. Day 30 ROAS is the more accurate picture of whether a cohort is profitable, because it captures trial conversions and early renewals that day 7 misses. Day 7 ROAS is the better operating metric, because it arrives while you can still act on it. Use Day 7 to make weekly creative and budget decisions, and Day 30 to check that your Day 7 target is still calibrated correctly.
Why is my day 7 ROAS dropping while CPI stays flat?
That pattern almost always means your creative has started attracting a worse quality of install rather than a more expensive one. A hook that overpromises, or an angle that pulls in people outside your actual use case, will keep buying cheap installs while the revenue behind them collapses. It is one of the earliest and most reliable signs that a creative concept needs retiring rather than refreshing.
Does SKAdNetwork affect day 7 ROAS reporting?
Significantly, on iOS. Postbacks arrive on a delay of at least 24 to 48 hours after their measurement window closes, so a Day 7 figure read on day 7 is systematically incomplete and will look worse than reality. Read Day 7 ROAS on a rolling basis with a lag of several days built in, and never make a kill decision on a cohort whose postbacks have not landed yet.
Want this run for you?
Getting Day 7 ROAS reported correctly is a measurement job and most teams can do it in a fortnight. Moving the number is a creative job, because once the funnel and the paywall are sensible, the remaining variable is which psychological position your ads occupy and how many of them you can cover at once.
That is the work we do as a performance creative agency for mobile apps: finding the angles an account is not covering, producing against them at a cadence that matches how fast the category fatigues, and reading the results at concept level rather than asset level. If you would rather have it run for you, apply to work with us. We take a small number of mobile app clients per quarter.