How to tell if your app is working: three numbers that matter more than downloads

A download counter goes up and almost never goes down. That is why it feels good and tells you so little. Someone who installs your app, opens it once and deletes it two minutes later is one download, forever.

Here is a week from my own apps. Across five of them, 90 installs and 90 deletions in the same seven days. The install chart for that week looked healthy. The deletion column said the apps had, on net, gained nobody.

Three numbers fix most of this: the ratio of installs to uninstalls, what people do during a session, and where your installs come from. None of them need a paid tool. App Store Connect and Play Console already hold the data, with some limits worth knowing before you trust a single figure.

1. The install to uninstall ratio

Divide the installs in a period by the deletions in the same period. A ratio of 3 means three installs for each deletion. A ratio under 1 means the app lost more copies than it gained.

What Apple counts matters here. Installations include redownloads on the same device and Family Sharing installs. Deletions cover removing the app from the Home Screen or through Manage Storage, and do not cover someone erasing the whole phone. For installs I count first-time downloads and redownloads, and leave updates out, since an update is not a new person choosing your app.

Both numbers come only from users who agreed to share analytics with developers. Apple also hides rows built from fewer than five devices. So the figures in App Store Connect are a sample. They are good for comparing one week with the next and one app with another. They are not a store total.

Over six weeks, the five apps that had data landed between 0.8 and 3.0 on this ratio, week to week. One week was 28 installs against 33 deletions. Another was 54 against 18. A single week means close to nothing at this size. The trend over four to six weeks is the thing to read.

Two cautions.

First, deletions in a week do not come from that week's installs. Someone deleting today may have installed in August. The weekly ratio is a flow, not a cohort, and it will look worse right after a burst of installs ends, because the deletions arrive late.

Second, a deletion is not always a verdict on the app. A person who installed on a Friday evening, looked, and removed it on Saturday is telling you about your product page, not your code. A person who removed it after three months is telling you something else. If you can split the ratio by device or by app version, do it. A ratio that collapses on iPad only, or on one release, points straight at a cause.

On Google Play the words change. The console counts users and devices separately. "User loss" includes people who stopped using every device with your app for more than 30 days, which is not the same as an uninstall. Uninstall events only count when the app was opened within 180 days of install. Do not set an Android ratio next to an iOS one and conclude anything.

2. Session data, and what you will not get for free

Apple defines a session as the app being used for at least two seconds, and a return from the background counts as a new session. Analytics gives you sessions and active devices. It does not give you session duration.

So the free number is sessions per active device. If your active devices open the app four times a week, that is a habit. If it is once, they came, looked, and left the app on the phone.

Duration is something you measure yourself. Record a timestamp when the app comes to the foreground and another when it goes to the background, and store the difference. A dozen lines of code. Use the median, not the average, because a few phones left open overnight will drag the average up.

Then read it against what the app is for. A timer or a shopping list is working well when people spend 20 seconds in it and come back daily. A language course or a quiz is working well when a session lasts several minutes. A long session in a utility can mean it is slow or confusing. There is no good duration in general, only one that fits the job.

Retention sits next to this. Analytics has a retention view that shows how many people come back after a first launch. If sessions per device are low, retention will say the same thing in a different shape. Look at both, and fix onboarding before anything else, since the first launch is where most people leave.

3. Where your installs come from

App Store Connect splits downloads by source type. Apple's list:

• App Store search: people who found the app by searching, including from search ads.
• App Store browse: people who found it in Today, Games or Apps, or in charts.
• App referrer: people who tapped a link in another app, including Apple apps other than Safari.
• Web referrer: people who tapped a link on a website. Safari taps land here. Taps from other browsers show up as app referrers.
• Others: App Clips, institutional purchases, and an "unavailable" bucket that includes promo codes, gift cards and device management.

Across 256 first-time downloads over the last several weeks in my apps, 72% came from App Store search, 15% from app referrers, 11% from web referrers and 2% from browse. One app got 60 of its 78 downloads from search. Another got a quarter of its downloads from a website link, which I would not have guessed.

This matters for a reason that has nothing to do with vanity. Search installs mean someone was already looking for what you do. Your growth then depends on ranking, and ranking depends on your keywords and your ratings. That is a narrow road. Installs from links mean someone told another person about the app, or put it somewhere a reader could tap. That is a different kind of audience, and it grows with effort that has nothing to do with keywords.

Cross this with section 1. Take the ratio for search installs and the ratio for link installs separately. If link visitors delete the app twice as fast, the page or the post that sent them promised something the app does not do. That is a fixable problem, and you can only see it with the split.

Some limits. A manual redownload resets the source, so a user who first came through a link and later redownloaded from search is counted under search from then on. Referrer sources need a minimum volume before they appear, so a small app can have real link traffic that the console hides.

A ten-minute weekly check

1. Open the install and deletion figures for the last four to six weeks. Write the ratio down. Compare it with the weeks before, not with a number from an article.
2. Check sessions per active device. Did it move after the last release?
3. If you record session length, look at the median for the same weeks.
4. Look at the source split. Did a new link or post change it?
5. Pick one thing to change, and look again in two weeks.

Stop watching the daily download count. It moves for reasons you cannot act on.

Where I can help

I have built native iOS apps since 2008. If you want a second pair of eyes on your numbers, or help adding session tracking without slowing the app, book a Monday slot.

App success metrics FAQ

What is a good install to uninstall ratio?

There is no published figure worth trusting, because it depends on the category, on how many installs come from ads, and on how many users share analytics. A ratio above 1 means the app is gaining more copies than it loses. Compare yourself with your own last six weeks.

Does App Store Connect show session duration?

No. It shows sessions and active devices. Session duration needs a small piece of code in your own app.

Why are my numbers lower than my real downloads?

Installations, deletions and usage in Analytics come only from users who agreed to share analytics with developers. Treat them as a sample.

Where can I see where my downloads come from?

In App Store Connect, open Analytics and the Sources view. It breaks downloads down by search, browse, app referrer, web referrer and campaigns.