How iOS 14 Changed Facebook Advertising Forever

The iOS 14 update did not break Facebook advertising. It changed what you can measure, which changed how you have to think about it. Advertisers who adapted their reporting and optimisation approach came through fine. Advertisers who kept expecting pre-iOS 14 data fidelity spent the next two years confused about why their numbers looked wrong.

This is what actually happened, what it changed permanently, and how to run campaigns properly in a post-ATT world.

What ATT Actually Did to Attribution

Apple's App Tracking Transparency framework, rolled out with iOS 14.5 in April 2021, required every app to explicitly ask users for permission to track them across other companies' apps and websites. The permission prompt is binary: allow tracking or ask app not to track.

On Facebook's app, a large portion of iOS users chose to opt out. The exact rate has never been officially confirmed but independent research consistently estimated 60 to 80% opt-out rates in early rollout markets, including Australia.

What this meant for attribution: when an iOS user who had opted out of tracking clicked a Facebook ad and visited a restaurant website, Facebook's browser pixel could not reliably connect that website event back to the Facebook user profile. The cross-site tracking that the pixel relied on was unavailable for opted-out users.

Attribution requires connecting an ad exposure or click to a downstream conversion. ATT removed the primary mechanism for making that connection for a significant portion of the iOS audience.

The 28-Day to 7-Day Window Change

At the same time as ATT rolled out, Meta was forced to change attribution windows for iOS traffic. Before iOS 14.5, advertisers could use 28-day click, 28-day view, and 1-day view attribution windows. After ATT, Meta restricted iOS attribution to a 7-day click, 1-day view window.

This matters more than most advertisers realised at the time.

A 28-day click window means that if someone clicks a Facebook ad today and books a table 18 days later, that conversion attributes to the campaign. Under a 7-day window, that same conversion would not be attributed. It falls outside the window.

For businesses with longer consideration cycles, corporate catering enquiries, private events, high-value retail, the window change alone caused a significant drop in attributed conversions, even with no change in actual business performance. The bookings were still happening. The attribution window was no longer capturing them.

Reported ROAS dropped. Reported CPA climbed. Business did not necessarily change at all.

Why Reported ROAS Dropped Even When Real ROAS Held

This is the part that caused the most confusion in agency-client relationships post-iOS 14.

Consider a restaurant that had a 4x ROAS on Meta in February 2021. By June 2021, the same restaurant showed a 2.2x ROAS in reporting. The temptation is to conclude that Meta stopped working. The actual cause was usually a combination of three things:

Attribution window compression from 28-day to 7-day meant late-converting customers no longer appeared in the data. Opt-out rates for iOS users meant a portion of conversions were occurring from users that the pixel could not track. The shift in data caused the algorithm to receive fewer conversion signals, which temporarily affected bidding optimisation.

Meanwhile, the restaurant might have been generating similar or better actual revenue. The revenue did not show up in the Meta reporting because the path from ad exposure to conversion had become less visible, not less common.

The advertisers who panicked at this point pulled budget from Meta and attributed their decision to "Meta not working anymore." The advertisers who understood the attribution shift correctly recognised that their real-world metrics, actual covers filled, actual revenue, actual bookings received, were the ground truth and the Meta reporting was now an undercount.

Modelled Conversions

Meta's response to the data gap was to introduce modelled conversions. When actual event data cannot be received for opted-out iOS users, Meta uses statistical modelling to estimate what conversion activity likely occurred based on patterns from consented users with similar characteristics.

Modelled conversions appear in the reporting data alongside directly observed conversions. Meta does not separate them in the standard reporting interface. This means some of the conversions you see in your Ads Manager are modelled estimates, not directly tracked events.

This is not fraud. It is a statistically reasonable response to a data gap. But it does mean that campaign performance numbers in Meta Ads Manager include some estimation, and the quality of that estimation depends on the quality of the signal Meta has from consented users. Good CAPI implementation, strong first-party data, and maintaining ad delivery to non-iOS audiences all improve the modelling accuracy.

The practical implication: Meta reporting is directionally useful but is not the single source of truth. Combine it with your real-world business metrics.

How We Adapted Our Reporting and Optimisation

The adaptation has two parts: reporting and optimisation.

On reporting, we moved to a blended measurement approach. For restaurant clients, the primary performance question is not "what does Meta Ads Manager show as ROAS." It is "how many covers or bookings did we generate this month compared to last month, and what did we spend on marketing." Media Efficiency Ratio, which is total revenue divided by total ad spend across all channels, is a more reliable number than channel-attributed ROAS in a post-ATT environment.

For campaigns where we need more precision than blended MER provides, we use conversion lift testing, which runs a geographic or audience holdout and measures incremental bookings attributable to the campaign. This is slower and requires volume to produce statistically significant results, but it gives genuine incrementality data rather than last-click attribution.

On optimisation, we shifted toward signals that Meta can still measure reliably. Landing page views, rather than conversions, as the optimisation event for awareness-phase campaigns. Video views and Reel completion rates as creative performance indicators. On-Meta lead forms, which capture data directly in Facebook without the cross-site tracking dependency, for certain lead generation objectives.

The account structure also changed. Post-ATT, broad audience targeting with strong creative performs better than heavily segmented, heavily interest-targeted campaigns. The reason is that narrower audiences are more affected by data loss because the modelling has a smaller sample to work from. Broader audiences with strong creative let Meta's algorithm find its own signal within a larger pool.

What Has Not Changed

The fundamentals of good Meta advertising have not changed at all. Strong creative still drives performance. Relevant offers to the right geographic audience still convert. Understanding the customer and showing them something specific and compelling still outperforms generic reach at any budget.

The measurement infrastructure around those fundamentals needed updating. The strategy underneath them did not.

Restaurants running well-targeted local campaigns with high-quality Reel creative, direct booking links, and proper exclusion logic are still generating strong returns on Meta. The numbers in the dashboard look different from 2020. The real-world results, when you look at actual reservations, are real.

iOS 14 changed the measurement game. It did not end it. If you want to see how we structure campaigns and measurement for hospitality clients specifically, see our Meta Ads services or get in touch.