3,340 people saw Kleos Music. I don't know if any of them wanted it.

The short version: between 2 and 15 October — the first two weeks Kleos Music was on Google Play — 3,340 people saw it in the store, 63 installed it, 23 opened it, and 3 were still there a week later. I have been looking at these numbers for days and I still cannot tell whether the problem is the store, the app, or me. This is me trying to figure it out in public, which is probably a mistake.

The five numbers

I am putting them at the top because I am not going to dress them up. These are the first two weeks on Google Play, 2 to 15 October, straight from the store — not Reddit, not any pre-release push.

• Device impressions: 3,340. Up more than 999%.
• Device acquisitions: 63. Up 7%.
• First-time device opens: 23. Down 30%.
• Monthly active devices: 50. Up 6%.
• Devices retained at 7 days: 3. Down 50%.

That last one is the one that keeps me up. Three. Out of 3,340 people, three were still around a week later, if that number is even measuring what I think it is.

3,340 impressions, 63 installs

63 installs out of 3,340 impressions is 1.89 per 100. Two people out of every hundred did something. The other ninety-eight kept scrolling.

I could call that our conversion rate. I am not going to, because it is an impressions-to-acquisitions ratio and the two metrics may not count the same events in the same window, and I am not going to inflate a number I am about to complain about.

But I do not need a precise definition to know the shape of it. 3,340 people saw Kleos Music and did not install it. That is the fact. The question is why, and there are two answers I cannot pick between.

Either the listing is bad. The icon, the screenshots, the description, the way it sits next to everything else on the store. One of those, or all of them, fails to win someone in the two seconds they give a search result.

Or 3,340 people who were never going to install a local music player saw it anyway. Impressions are not intent. Being seen is not the same as being wanted.

I would love it to be the first one, because the first one is fixable. I do not get to choose.

63 installs, 23 opens

This is the number that actually scares me.

63 devices installed the app. 23 were opened for the first time. Read it plainly and 40 people installed Kleos Music and never opened it. That is a great story for a blog post. It is a terrible way to read two different metrics.

Acquisitions and first-time opens are not the same measurement. They may use different definitions, count different events, or sit in differently aligned windows. A phone that installs, uninstalls and reinstalls. A device two people share. The gap between the store recording an install and the app recording an open.

So I am not going to write that 40 people abandoned us, because I cannot prove they ever arrived. What I will say is that a gap between 63 and 23 is big enough that I would investigate it before I investigated anything else on this list.

Three devices came back

Three. Down 50%.

I could make a whole tragedy out of that, and I refuse to, because three is three and small numbers lie. Two devices entering or leaving can produce a percentage that looks like a cliff and means nothing.

But if it is real, it says one specific thing: the problem is after the install. Someone who installs an app and does not come back is describing the first few minutes, not the marketing.

I do not know which it is yet. I know I have to check what that 50% is even measured against before I let it ruin my week.

I think I know what is wrong, and I might be wrong

Here is my hypothesis. I am labelling it, because I have been confident and wrong before.

Kleos Music asks you for something before it gives you anything. Before your music plays, there is a setup step. You point the app at your library. You wait for it to find the files already on your phone. On paper, simple. In practice, that is the exact moment where an app stops being a download and becomes a chore.

If that step is too long, too confusing, or asks for too much before it has earned the right to ask, then someone can install Kleos Music, hit setup, and never reach the part of the app that would have convinced them to stay. That one failure would explain a low number of first opens, low seven-day retention and a healthy number of installs, all at once. It is the cleanest explanation I have.

It might also be completely wrong. The app could just not be good enough. I have to say that out loud, because the version of me that gets defensive here is the version that never ships a fix.

The next update automates and simplifies setup. The goal is to get people to their music faster, with fewer decisions before the payoff.

I am not going to tell you that fixes retention. Nobody knows that until it is measured, and I have learned not to trust a fix I have not seen numbers for. What I know is that setup is the only part of this whole funnel I can actually change right now, so it is the part I am changing.

The 50

Fifty monthly active devices, up 6%.

I am putting this here on purpose and not drawing a line to anything else. Monthly active devices is a looser metric with a wider window, not comparable to seven-day retention, and not 50 people. One person can carry two phones. One phone can pass through three hands.

What it tells me is that there is some use underneath all of this, which is not a defence of the numbers above and not a consolation. Just another signal with its own caveats.

So what is actually broken

The real question is not whether these numbers are good. They are not. The real question is where the leak is.

Is it the store, failing to turn a glance into an install? Is it setup, failing to turn an install into a first real use? Or is it the app itself, failing to give anyone a reason to come back?

The numbers point at all three and prove none. What they do suggest is that the mess is in the middle — the space between installing the app and using it — and that is the only part I can touch in the next release.

What I will check, and what would tell me I was wrong

I am not going to announce a win until I see it in these numbers, watched long enough to be more than noise:

• Impressions and acquisitions, to see whether the ratio between them moves.
• First-time opens against acquisitions, to find out whether the 63-to-23 gap was real and whether closing setup changes it.
• Devices retained at 7 days, with the cohort definition checked first, so I actually know what I am looking at.
• Monthly active devices, as a slow, separate read on whether anyone is still using it.

If setup was the problem, the first two should move before the fourth. If they do not, then I was wrong, and I will have learned something more useful than a nicer onboarding screen: that the problem is somewhere I have not been looking.

I do not know yet whether Kleos Music is failing to be found, failing to be used, or failing to be wanted.

I am going to find out, and then write about it either way, because the version of this blog that only gets published when the numbers go up is worth nothing.