The three numbers that tell you if your app can scale.

Most founders look at the wrong number. They look at downloads, or at a single return-on-ad-spend figure on day one, panic at it, and either keep spending on hope or quit too early. So let me give you the three numbers I actually look at, what they mean, and roughly where the line is. If you only take one thing from this, take this: scaling is a math question, and the math is knowable in about thirty days.

Before the numbers, one idea that makes them make sense. You do not judge an app on a single point in time. You judge it across a funnel, install to activation to retention to money, and you find out where it breaks. Most apps do not break everywhere. They break in one place. The job is to find that one place, decide if it is fixable, and only then decide whether to pour fuel on it. The three numbers below map to the three places it tends to break.

Number one. Cost per install, against your real LTV.

This is the first gate, and it answers one question: can we get attention at a price that pays back? CPI on its own is meaningless. A two dollar install is great for a subscription app at forty dollars of lifetime value and a disaster for an app that makes you four. So the number you care about is CPI relative to what a user is actually worth to you over time.

Roughly, for the apps I work with, a viable CPI lands somewhere between eighty cents and two dollars, depending on category and how you monetize. But the band is not the point. The point is the relationship. If your best honest estimate of lifetime value is fifteen dollars and you cannot acquire installs under, say, four, the market is telling you something, and no amount of creative tweaking will close a gap that size.

Here is the part people get wrong. They run one ad, see a bad CPI, and conclude the app is the problem. In 2026 that is a creative volume problem far more often than a product problem. The algorithm learns from the creative now, the cookie is gone, and one hero ad does not give it enough to find your buyers. You need many persona-tailored variations before you can trust the CPI signal at all. So the rule at this gate is: after two or three real creative iterations, if the market still will not respond at a price near your LTV, that is a positioning answer, not an optimization answer. You stop. You do not spend your way through it.

Number two. Activation rate.

Say the CPI works. Attention is affordable. The second gate asks: do the people who install actually reach the moment that matters? The core action that makes your app worth opening again. The first journal entry, the first completed workout, the first real result. Activation is the share of installs that get there.

I look for something like twenty five to forty percent, again depending on the app. Below that, you have what I call expensive vanity. You are paying for installs that never become users. The brutal thing about this gate is that it is the one most makers never even measure. They watch installs go up, feel good, and never notice that three out of four of those installs opened the app once and never activated.

And here is the good news hiding in the bad news. When an app breaks at activation, the fix is usually not in the ad and not in the price. It is in the onboarding. It is the gap between what the ad promised and what the first ninety seconds in the app delivered. That is fixable, fast, and it is often where the single biggest unlock lives. I have seen activation move ten or fifteen points on an onboarding change that took a few days. That swing alone can be the difference between an app that cannot scale and one that obviously can.

Number three. Day-7 ROAS, measured across time, not at a point.

The third gate is the money gate: does the math actually close? This is where the single-number mistake does the most damage. Someone looks at day-one return on ad spend, sees ten or twenty percent, and concludes the app loses money. Of course it does on day one. Almost everything does. You are paying the full acquisition cost up front and collecting revenue over weeks and months.

So you read ROAS as a curve. Day one tells you there is early willingness to pay. Day seven, somewhere in the thirty to fifty percent range, tells you engagement and early conversion are holding. Day thirty, climbing toward payback, tells you the curve is going to close. The single most common way good apps get killed is someone judging a thirty day payback business on a one day number and panicking. Do not do that. Watch the shape.

Now put the three together, because the sequence is the method.

CPI tells you the market responds. Activation tells you the users stick. Day-7 ROAS tells you the money comes back. Run them in order, gate by gate, and the app tells you exactly where it stands. If it passes all three, the math closes and scaling is a fuel question, not a faith question. If it breaks at one, you know precisely which one, and whether it is fixable. If it breaks in a way that will not fix, you have your answer in thirty days instead of finding out in six months of invoices.

This is the discipline mobile games publishers have run for years. Test cheap, find where it breaks, fix what is fixable, kill what isn't, scale what works. The reason it feels unfair is that they ran it two thousand times a year and indie makers got to run it zero times. The numbers above are not secret. What was missing was someone whose job is to run them on your app, with money behind the answer.

One more thing, and it is the thing that makes this honest rather than a sales funnel. Before any of this runs, you write down the kill criteria. The exact thresholds below which you stop. You sign them, I sign them, before a single dollar is spent. That one document is what turns ad spend from a gamble into a diagnostic, because it means neither of us can move the goalposts when the numbers come in. The most valuable thing a publisher does is not scale the winners. It is say no, cleanly, on criteria everyone agreed to in advance.

So if you have an app with real revenue and you want to stop guessing, here is the practical next step. Score yourself on these three. I built a free diagnostic that walks you through it with your own data and gives you a verdict, scale-ready, fixable, or kill. And if the diagnostic says you are scale-ready and you want to see the real thing run, with my money on the line for the test, that is what the Launch Test is for.