Skip to main content
FHAVI

Research · Tracking

What still works in iOS and Safari, and what quietly stopped

Verified November 2026. Which signals survive on Apple devices, what the platform-side workarounds actually compensate for, and which measurement habits no longer hold.

2 min read

In brief

  • Cookie-based cross-site tracking on Safari is not reliable enough to be the primary basis of measurement — this has been true for years and has not reversed.
  • Server-side events and first-party identifiers are the practical response; neither restores what was lost, both keep the count believable.
  • Any article on this subject without a verification date should be treated as stale, this one included.

What actually survived

  • First-party cookies and server-set identifiers on your own domain still function within their lifetime limits.
  • Server-side events still reach platforms, because they do not depend on the browser allowing a third-party request.
  • Modelled conversions still fill some of the gap on the platform side, and are still estimates, not observations.
  • Aggregated reporting still exists, and still cannot answer product-level questions at low volume.

Verified on

November 2026. Apple changes these behaviours at the OS and browser level without notice to advertisers. Re-check before relying on any specific mechanism described here.

The practical configuration

  1. 1

    Send purchase and checkout server-side with shared event ids

    This is the part that survives browser restrictions, because the order happened on your server.

  2. 2

    Capture consent properly and pass it with the order

    Consent state is what the server needs to decide whether an event may be sent at all.

  3. 3

    Use your own data for decisions where possible

    Store revenue, blended ROAS and margin do not depend on which browser the customer used.

  4. 4

    Accept a modelled gap and document it

    Write down the size of the gap between platform counts and store counts. A documented gap is a measurement system; an undocumented one is a recurring argument.

Habits that no longer hold

  • Treating platform purchase counts as the order count. They are a matched subset, with a modelled remainder.
  • Judging creative on platform-reported revenue alone at low volume.
  • Assuming a tracking gap is a bug to be fixed. Some of it is structural and permanent.
  • Expecting a single source of truth across four platforms. There is one source of truth — the order record — and everything else is an estimate of how it was found.

Apple did not break your tracking; it removed a mechanism you were relying on. The replacement is server-side events with shared ids, consent captured at source, and the discipline of deciding from your own order record.

Questions

Asked about this directly

Can I get the old attribution back?
No. The mechanisms that provided it are gone and are not returning. What you can do is keep the count believable with server events and decide from your own order data.
Are modelled conversions reliable enough to optimise on?
They are better than optimising on an incomplete count and worse than observing the event. Use them, and check them against store revenue monthly.
Does this affect Android too?
Similar pressures exist across platforms, with different mechanics and timing. The response — server-side events, first-party identifiers, decisions from order data — is the same.

Keep reading

Other pieces in the library

All research

Numbers first, promises second

Send your store and we will tell you which of these failure modes is already happening — before you pay for a fix.

Prefer email? Write to md@fhavi.com