Tracking app installs from a short link

Why the store breaks the chain

A short link can carry any parameters you like right up to the moment the operating system opens the app store. From there, your parameters are in someone else's hands. The store page is a native surface, the install happens through the store, and the newly installed app launches with no knowledge of the link that started the journey. Nothing about your query string automatically survives that boundary.

Everything in mobile attribution is a technique for bridging that gap. Understanding which bridge you are using, and what its failure modes are, matters more than the specific tool, because the numbers each bridge produces are not comparable with one another.

The install referrer on Android

Android provides an install referrer mechanism: a referrer string attached to the store link is preserved through the install and can be read by the app on first launch. This is the most reliable bridge available on the platform, because it is deterministic rather than probabilistic. The value you attach is the value the app reads.

The practical requirements are that the store URL must carry the referrer parameter, that the value must be URL-encoded, and that the app must read it early in its first run and store it before the user does anything else. Keep the string short and structured, typically a set of key-value pairs mirroring your campaign parameters, and agree the exact format with whoever writes the app code before you publish anything.

Deferred deep links on iOS

iOS has no equivalent install referrer, which is why deferred deep linking on that platform is fundamentally a matching exercise. The link records some context at click time, the app reports some context at first launch, and a service decides whether they are the same person. That decision is probabilistic and its accuracy degrades as platform privacy controls tighten.

The consequence to internalise is that iOS install attribution has an error bar. Treat it as directional. Comparing two campaigns measured the same way is reasonable; treating an absolute install count as a fact is not, and reconciling it exactly with store console numbers is a task that will consume days and end in disagreement.

Send existing users straight into the app

A large share of the people who tap your link already have the app. Sending them to a store page they do not need is a wasted step and a measurable drop-off. The link should attempt to open the app directly using a universal link or app link, and fall back to the store only when the app is not installed.

This is where a short link earns its place: one printed or shared URL that resolves differently depending on what is on the device. Configure the association files the platforms require, and test the case that people forget, which is a device that has the app but has never opened it since install.

Test the four cases separately

App installed and previously opened, app installed but never opened, app not installed and store available, and the link opened inside an in-app browser such as a social feed. These behave differently, and the in-app browser case is where most implementations quietly fail, because those browsers handle app-opening intents inconsistently and sometimes refuse them.

Run each case on a real device of each platform. Simulators do not model the store, and a desktop browser tells you nothing at all. Write the four cases into a checklist and repeat it after every platform update, because this is behaviour that changes without notice.

Keep one naming scheme across web and app

The campaign parameters you use on web links and the values you pass through the install referrer should be the same vocabulary. If web uses one convention for a channel and the app uses another, nobody will be able to compare a web campaign with an app campaign without a manual mapping table that goes stale.

Decide the vocabulary once, document it, and make the link creation process emit both forms from the same source. The translation is mechanical; letting each side invent its own names is what makes it hard.

Expect the numbers to disagree, and decide which one you trust

Clicks on the short link, store page views, installs reported by the store, and installs attributed by your tooling will all be different numbers, and each measures a genuinely different thing. Clicks include people who never reached the store. Store installs include organic traffic that never touched your link. Attributed installs include matching error in both directions.

Pick one number as the one you plan against, write down why, and report the others as context rather than as competing truths. Most of the time spent arguing about mobile attribution is spent comparing figures that were never intended to match.