Apple Is Blocking Identity Resolution. Deterministic Tracking Just Lost Another Chunk

Safari 27 Closes the Email Workaround

This month the WebKit team announced that Safari and WebKit will block network access to identity resolution infrastructure in the iOS, iPadOS, macOS and visionOS 27 releases (Blocking Identity Resolution Infrastructure, WebKit). It is a short post, and it deserves a longer reaction than it has had.

Here is the mechanism WebKit is describing. A user logs in, signs up to a newsletter or fills in a form. A third party, not the website owner, quietly collects the email address and sends it to its own servers. It creates an opaque tracking ID permanently tied to that address and hands the ID back to the site to store. Do that across enough sites and the third party builds an identity graph: one profile, stitched from every login and form it has ever seen. Logging out or unsubscribing does not remove the ID or the stored address.

WebKit's answer is to block access to that infrastructure at the domain level, which means providers cannot get round it by restructuring their requests. First-party personalisation is explicitly left alone. The target is the central resolver that can match people across sites and parties.

Why this is bigger than another browser tweak

WebKit says Safari fully blocked cross-site cookies six years ago. The industry's answer was not to give up on identifying people. It was to swap the cookie for something that looked sturdier: the email address. Hashed email became the deterministic backbone of the post-cookie stack because, unlike a cookie, it follows the person rather than the browser.

That was the patch.

Apple has now gone after the patch.

I would resist the temptation to treat this as a Safari problem you can footnote. Third-party cookies went first. Then came browser limits on first-party cookie lifetimes, then consent banners and reject-all models, and now the identity layer that was supposed to fill the gap

Each step looked survivable on its own. Stack them up and the deterministic share of your measurement is a good deal smaller than your dashboards suggest. The rest has not disappeared. It has been modelled.

Modelled is not measured

We have just put a number on what that costs. When RS Group PLC moved to a strict reject-all consent model, conventional analytics lost most of the raw observation it depends on and fell back on modelling. Over a four-month test across RS Group's properties worldwide, Corvidae tracked £193m more revenue than the existing analytics and covered over 95% of the active customer base. Customer journeys turned out to be 46% longer in the UK and 68% longer in the EU than the platforms had reported.

The gap matters beyond reporting, because bidding algorithms learn from whatever the measurement layer hands them. Feed them a modelled guess and they optimise against a guess. No amount of reporting corrects a bid that was placed on a guess. In the Italy pilot, correcting the data feeding the bidding delivered 6% more Paid Search revenue on 5% less ad spend, and improved ROAS by 12%.

Apple's announcement pushes in the same direction. Every Safari user whose identity can no longer be resolved moves from the observed column to the modelled column, and your bidding inherits the difference.

The wrong response is to find the next identifier

Expect a wave of announcements from vendors explaining why this does not affect them, or how they have already moved on. Ask the question underneath: what is the replacement? Every identifier, whether it is a cookie, a device ID or a hashed email, is something a browser vendor can see, classify and block. WebKit's decision to block at domain level shows how little room that leaves. A strategy that depends on out-running the browser vendors has a shelf life, and it keeps getting shorter.

The alternative is to stop depending on identifying the person at all. Corvidae reconstructs customer journeys from raw first-party eventstream data using predictive AI, with no cookies and no persistent identifiers, and is compliant with GDPR and CCPA by architecture. It does not rely on the class of signal that browsers keep closing, which is why we see it as the measurement layer that belongs underneath your bidding, not another identity graph on top of it.

What we do not know yet

It is worth being precise about the limits of the announcement. WebKit does not name the providers affected, does not say how identity resolution infrastructure is identified or how the list of domains is maintained, and gives no timeline beyond the iOS, iPadOS, macOS and visionOS 27 releases. The size of the hit for any one business depends on two things nobody outside that business can see: how much of its traffic comes through Safari, and how much of its journey matching depends on this kind of identity call. WebKit also points users to Hide My Email, which creates unique, randomised addresses for each site and makes any surviving email-based matching less useful still.

Four things worth doing before iOS 27 reaches your customers

Size your Safari exposure. Find out what share of your sessions and revenue comes through Safari and iOS, and how many of your attributed conversions rely on a matched identity rather than an event you observed on your own site.
Ask every vendor in the stack the same question, in writing. Does your tag call an identity resolution endpoint from the browser, and what happens to the data when that call is blocked? If the answer is that it gets modelled, ask how the model is validated.
Ask what proportion of your sessions you actually observe. If nobody can answer that, and say what the remainder is made of, that is the finding.
Check what is feeding your bidding. If automated bidding is trained on modelled conversions for Safari users, your bids inherit the gap. Hold back a comparable campaign or region and judge the result on incremental revenue, not platform-reported ROAS.

As more of the stack becomes automated and platform-controlled, the advantage moves to whoever can still independently verify what actually happened. Apple has just made that harder for everyone who was relying on identity. It has not made it harder for anyone who never needed it.

Ready to try attribution the Corvidae way?

Work with your dedicated Success Manager

Getting started on the road to accurate attribution doesn’t have to be complicated. Our team are on-hand to make the move to a new attribution tool as easy as possible.