iOS 14.5 and Facebook Ads: Measuring Results After App Tracking Transparency

Since iOS 14.5 rolled out at the end of April, several clients have asked why their Facebook conversion numbers look lower, even though actual sales or bookings have not dropped. The answer traces back to one change from Apple, but it means reworking the whole ad account setup to keep reading reliable results. Here is what I have been checking and fixing in client Facebook Ads accounts since this spring.

What exactly changed with iOS 14.5?

Apple added a mandatory consent screen called App Tracking Transparency, which asks every iPhone user for permission before an app can track their activity across other apps and websites. When someone opens the Facebook or Instagram app on an updated iPhone, that prompt appears with two choices: allow tracking or ask the app not to track. Since the prompt started showing up, a large share of people who see it choose to decline, which cuts into how much data the Facebook pixel can tie back to a specific person on those devices.

Why do your conversion numbers look lower since April?

Because the pixel on your site keeps working exactly as it did before, but Facebook can no longer attribute some of those conversions to a specific ad for users who declined tracking on iPhone or iPad. Those conversions do not disappear entirely, they become aggregated, modeled data handled separately, with a delay of a few days before they show up in Ads Manager. An account that used to compare conversion numbers in near real time now shows a report that keeps filling in for a day or two after the reporting period ends, which explains part of the gap if you check the dashboard too soon.

How do you verify your domain to keep measuring conversions?

The first step, before touching any campaigns, is verifying the site's domain inside Business Manager, under Brand Safety then Domains. Facebook offers three methods: add a meta tag to the site's header, upload an HTML file Facebook provides to the root of the site, or add a DNS TXT record through the domain host. On a WordPress site, the header meta tag is usually the fastest route, as long as it gets added to the theme rather than to a single page that might get edited later. An unverified domain simply cannot set up prioritized conversion events, which means campaigns keep running, but with even less reliable data to optimize against.

Which events should you prioritize when you are limited to eight?

Once the domain is verified, Events Manager lets you configure up to eight conversion events per domain under aggregated event measurement, ranked from most to least important. For an iPhone user who declined tracking, Facebook can only report the single highest-priority event that person completed on their visit, not the full list of pages they viewed. For a restaurant client, that means putting the reservation or order at the top, the phone call click right after it, and leaving a menu page view at the bottom of the list, because a menu view with no booking or call attached says nothing specific about someone who actually spent money. A store selling online would put purchase first instead, followed by add to cart, with product view sitting at the bottom.

What changed in the default attribution window?

Before iOS 14.5, most accounts used a 28-day click and 1-day view attribution window, which counted a wider range of conversions toward each campaign. To fit within aggregated event measurement, Facebook shortened the default window for new ad sets to 7-day click and 1-day view, cutting the click portion from 28 days to 7 while leaving the 1-day view window unchanged. An account comparing July results to March results without accounting for that shortened click window can end up concluding campaigns are performing worse, when part of the gap simply comes from measuring a narrower click window. Before drawing any conclusion about performance, I always check which attribution window is set on each ad set, under the Attribution section of the ad set settings.

How do you read a report that now includes modeled conversions?

Ads Manager shows a results total under each campaign that blends directly counted conversions with conversions Facebook models from patterns it sees elsewhere in the account. That number is still useful for tracking a trend over several weeks, but it gets less reliable for comparing one specific day to another, especially in a smaller account. The practice I recommend: compare full weeks against each other rather than single days, wait three or four days before judging a budget or targeting change, and keep a second source of truth, such as the actual number of bookings that came in by phone or email, to check what the ad report is telling you.

One more limitation worth knowing before you dig into breakdowns: the placement and age-and-gender breakdowns that used to split conversions between Facebook and Instagram, or by device, are no longer available for events measured through aggregated event measurement. Ads Manager still shows those breakdowns for clicks and reach, just not for the conversion events tied to the eight prioritized slots. A local business trying to figure out whether Instagram Stories or the Facebook feed drove more reservations cannot answer that question from the conversion breakdown alone anymore, and has to rely on the overall trend and the secondary source of truth instead.

What should you do if you manage Facebook ads for a local business yourself?

First, confirm the domain is verified inside Business Manager. That is a separate setting from anything inside the ad account, and the ad account's event configuration will not work without it. Next, review the attribution window set on each ad set: plenty of accounts created before April 2021 are still running with a default window that no longer matches what Facebook can actually measure, which throws off the cost-per-result number on the dashboard. Finally, avoid launching a new campaign for every hypothesis during the first few weeks after a setup change: Facebook's learning phase restarts with any significant edit, and it needs several days of stable delivery before it produces representative results again.

What is the first mistake to avoid?

The most common mistake I see is cutting a campaign's budget the moment a conversion number drops on the dashboard, without checking whether the drop is a real loss of customers or just a loss of visibility into the data. Before touching a budget, I always check the numbers that do not depend on the pixel: calls received, reservation emails, or Page messages over the same period. If those numbers hold steady while the Facebook dashboard shows a drop, the problem is almost always in measurement, not in the campaign, and cutting the budget at that point costs more than waiting for the report to finish filling in.

This kind of check is now part of the standard setup I run on every Facebook and Instagram Ads account we manage for a restaurant or local business, alongside installing the Facebook pixel itself. Domain verification and picking the eight priority events happen once, but they decide the quality of every report that follows.

Drafted with AI assistance, checked and published by Marven Salgado.