BlogAnalytics

Deep link analytics: how to measure taps, opens and installs per channel

· 8 min read · By the Strait team

What deep link analytics can and can't measure. How taps, app opens, installs and purchases are counted and joined, how a channel like WhatsApp is worked out without UTM tags, and why your install numbers will never match the store's.

Every deep-linking service shows a dashboard. The numbers look simple: taps, installs, maybe revenue. The hard part is knowing what each number means, because a tap, an app open and an install happen in different places (a chat app, a browser, the store, your app), and no single system sees all of them.

This post explains deep link analytics from the ground up: what can be measured, how the steps are joined, how a channel like WhatsApp is worked out, and the one thing nobody can measure. The definitions are the ones Strait uses, but the questions apply to any provider.

The funnel, and why it needs joins

The deep link funnelFour steps joined by one tap id: tap, app opened, install, purchase. Tap App opened Install Purchase one tap id carried through every step
Each step is recorded by a different part of the system, so each one carries the id of the tap that started it.

A useful funnel has four steps: taps, app opens, installs and purchases. Each is recorded somewhere different:

  • A tap is seen by the link server when the browser loads the link, or by your app when a verified link opens it directly without any browser.
  • An app open is seen only by your app, through the SDK.
  • An install is only visible when the new app opens for the first time.
  • A purchase happens in your app or on your server.

To follow one person from tap to purchase, each later step has to carry an identifier from the first. In Strait, every tap gets a tap id. It travels to the app in the hand-off URL (as strait_click), inside the Google Play install referrer, or in the server's answer when it matches an open. The SDK remembers the most recent one for seven days and attaches it to purchase events. That's what lets revenue land on the exact tap's channel and campaign instead of a guess.

When you evaluate any provider, ask how its steps are joined. Taps and installs counted separately and divided make a ratio, not a funnel.

Taps: what counts, what doesn't

A tap is a person tapping your link. Three details change the count:

  • Preview robots aren't people. WhatsApp, Telegram, LinkedIn and others fetch your link to draw a preview card. Those fetches must never count as taps. Strait serves them the preview and files them as previews, never billed.
  • Verified links can skip the server. When Android App Links or iOS Universal Links are verified, the phone opens your app directly from WhatsApp or Gmail, and the link server never sees the tap. It's counted when your app reports the open.
  • Browser hand-offs aren't two taps. If a browser loads the link and then hands it to your app, that's still one tap.

Where we sent them is the next field: app or store, app or website, store, website, the iPhone "Get the app" page, or the in-app-browser page. On Android the phone makes the final choice, so this is the server's decision, not the outcome. Keeping the two apart is what lets you debug a link that "opened the browser".

Channels: WhatsApp without UTM tags

Indian apps live on WhatsApp shares, and most of those links carry no UTM tags. A channel can still be worked out, in this order:

  1. A utm_source on the tap URL. Your own label always wins.
  2. A utm_source in the link's destination.
  3. The in-app browser the tap was opened in, when that browser names its app. Our engine looks for WhatsApp, Instagram, Facebook, Telegram, LinkedIn, Snapchat and X in the user agent, and for Pinterest and Paytm on Android. An app that hands the link to the phone's own browser, or opens it in a browser view that doesn't name the app, can't be recognised this way.
  4. The referring Android app or website, when the phone reports one, for example com.whatsapp or reddit.com.

If none of these is present, the tap is direct. Some web views don't name their app, such as Pinterest and Paytm on iPhone, so tag those links yourself. Our open your app from any app pages show, app by app, how a tap is labelled.

The practical rule: tag campaigns, let the rest be worked out. A broadcast called utm_source=whatsapp&utm_campaign=diwali-sale gets its own row; a link a customer forwarded to a friend is still filed under WhatsApp from the browser it was opened in. The UTM builder builds tagged links, including the Google Play referrer form.

App opens

An app open is your app telling the server "a link just opened me". The useful breakdowns:

  • Open rate: taps that opened the app, directly or through the browser, divided by all taps. This is the single number that tells you whether your links work.
  • New vs existing users: the first launch after installing, versus an app that was already there.
  • App state: closed (the link launched the app), background (it brought it back) or foreground.
  • Unresolved opens: a link opened the app but couldn't be resolved (expired, deleted, wrong workspace), so the person didn't land on the right screen. These are worth watching after a migration.

Each open is reported once. If the phone is offline, the SDK keeps the report and retries for up to seven days, and duplicates are ignored. Opens store the host and path of the URL, never the query string.

Installs

Installs are where analytics gets honest or doesn't.

  • Installs from links: first launches of a new install tied to one of your links.
  • Joined to the exact tap: those matched to the precise tap through the Play install referrer or a browser hand-off, not just to the link.
  • Unattributed installs: first launches no link could be found for: store search, word of mouth, or a match that failed.
  • Match rate: installs reconnected to a tap, divided by all attributable installs.
  • Time to install: the median time between tap and install, for installs matched to their tap.

On Android, the join is exact: the link travels through Google Play in the install referrer, and the app reads it on first open. On iPhone, nothing passes through the App Store, so Strait uses a short-lived match, which is opt-in.

The one thing nobody can measure

An install that is never opened. Until your app launches, no linking product can see it. Only Google Play Console and App Store Connect see raw installs. So expect a link service's install count to be lower than the store's, and be suspicious of any dashboard that claims to match the store exactly.

Store-bound taps where the app never opened and no install followed are therefore estimates. Strait labels them "probably reached the store" for that reason.

Purchases and revenue

Send purchases from your app (trackEvent) or your server (POST /v1/event) with a value and currency. With the tap id attached, revenue appears at the bottom of the funnel on the tap's channel, campaign and A/B variant. Two habits keep the numbers right:

  • Send each purchase from one place. App and server both sending the same order counts it twice.
  • Keep the 7-day window in mind. A purchase more than seven days after the last link open, or after the app was opened without a link, shows as direct.

A worked example

Say a food-delivery app sends a weekend offer as a WhatsApp broadcast and an SMS. The WhatsApp link is tagged utm_source=whatsapp&utm_campaign=weekend, the SMS link utm_source=sms&utm_campaign=weekend. People also forward the WhatsApp message to friends.

In Analytics, the campaign weekend splits into two channel rows. Forwards with the original tag stay on the WhatsApp row. Open rate per row shows whether the links worked: a low rate on SMS often means App Links aren't verified yet, so Android opens the browser. Installs joined to the exact tap show how many new users the offer brought, and purchases with the tap id show the revenue per channel. When a customer says "the link didn't open the app", the tap's trace shows the device, the in-app browser, the decision and where it was sent.

Questions to ask any provider

  1. Is each later step joined to the tap, or are steps counted separately?
  2. Are preview robots and bots excluded from taps and from billing?
  3. How are channels worked out without UTM tags?
  4. Does it count opens from verified links that skip the server?
  5. Does it say which numbers are estimates?
  6. Can you export all your raw data, with no date cap?

Try it

You can see how a single tap is decided without signing up: the Will it open? simulator runs the link server's own code on a channel and phone you pick. When you're ready, the analytics docs list every definition used here.

Sources and how we checked

  • Every definition in this post is from Strait's Analytics: how we count page and the link engine's code (the channel order is the engine's channel rules; the tap id fields are the engine's analytics contract).
  • The Android exact join was verified on a real Android phone with a Google Play internal testing build in October 2026. iPhone install matching is opt-in (off until you turn it on in Settings → iPhone installs) and proven on iPhone simulators; the real-iPhone run waits on our App Store listing.
  • The worked example is an illustration, not customer data.