What IDFA is
IDFA (Identifier for Advertisers) is a 128-bit random identifier that iOS assigns to each device. It
looks like an ordinary UUID: A1B2C3D4-E5F6-7890-ABCD-EF1234567890. The defining property is that
IDFA is not tied to an Apple ID — it is a device identifier for advertising purposes, and in
principle the person can reset it.
Until 2021, IDFA was available to any app by default unless the person explicitly opted out in
settings. That let the advertising ecosystem build detailed cross-app profiles.
The turning point: iOS 14.5 and ATT
In April 2021, Apple shipped iOS 14.5 with a mandatory App Tracking Transparency (ATT) mechanism.
Every app now has to request explicit consent before touching IDFA.
| Before iOS 14.5 | After iOS 14.5 |
|---|---|
| IDFA available by default | Opt-in required through the ATT dialogue |
| Around 90%+ availability | Around 30–40% availability across the market |
| Precise cross-app attribution | Aggregated attribution through SKAdNetwork |
| Retargeting by device ID | Retargeting on first-party data |
The market reaction was that people grant consent far less often than anyone expected. In e-commerce
apps, where the value of personalization is easier to explain, opt-in rates are higher — but they
rarely pass 50%.
What it means for e-commerce
The loss of IDFA hit several scenarios:
Install attribution: it became impossible to say precisely which ad on Facebook or Instagram led
to an install and the purchase that followed. Marketing ROAS calculations lost accuracy.
Retargeting: the classic audience of people who added an item to the cart but did not buy became
partial — only those who consented to IDFA.
Lookalike audiences: the quality of scaled similar audiences fell because the seed data shrank.
In-app analytics: untouched — it runs on the app’s own identifiers.
Important: content and recommendation personalization inside a single app does not depend on
IDFA. It is a cross-app identifier, not a tool for an app’s internal logic.
Where the market stands now
The advertising ecosystem adapted along three lines:
- SKAdNetwork — Apple’s aggregated attribution, available without IDFA; it returns a limited set
of parameters and does so with a delay - Media mix modelling (MMM) — statistical modelling of channel contribution without identifying
individual people - First-party strategies — email addresses and phone numbers as deterministic identifiers inside
the ad platforms
For e-commerce the conclusion is unambiguous: the more registered customers in the database, and the
tighter the integration with ad platforms through first-party data, the smaller the dependence on
IDFA.