iOS 14 and App Tracking Transparency: preparing Facebook ads

Apple announced the prompt at WWDC in June 2020 and shipped iOS 14 in September. After developer pushback, it delayed enforcement to "early 2021" without giving a date. When it arrives, any app that wants access to the device advertising identifier, the IDFA, must ask permission through a system dialog. Opt-in rates in early industry estimates range from roughly 10 to 40 percent depending on the app, and the Facebook app will be among those asking.

Facebook's response, announced in December, affects advertisers whether or not they run app campaigns:

The 28-day click window is already disappearing from Ads Manager for new reporting, and the comparison tool Facebook added lets you see both windows side by side for now. Use that window to understand your gap before it closes.

1. Verify your domain in Business Manager. Business Settings, Brand Safety, Domains. Add a DNS TXT record, upload an HTML file or add a meta tag. Only the verified business can configure the eight events for that domain, which matters if an agency, a franchisee or a platform partner also runs ads to it.
2. Pick and rank your eight events. Facebook will only optimize for the highest-priority event a person triggers on an iOS 14 device. For a store that is typically Purchase, Initiate Checkout, Add to Cart, View Content, and perhaps Lead and Complete Registration. For a service business, Lead, Schedule, Contact and a qualified lead custom conversion. Rank by business value, not by volume.
3. Audit your pixel events now. Many accounts fire duplicate or misnamed events, or have custom conversions nobody remembers building. Open Events Manager, use the Test Events tool, and remove anything that will not survive the cut.
4. Set up the Conversions API. Shopify, WooCommerce, BigCommerce and most major tag platforms now offer a native or partner integration. If you have a custom stack, your developer can send server events with the same event_id as the browser event so Facebook deduplicates them.
5. Change your reporting baseline. Export the last 90 days under 7-day click attribution and compare them with 28-day click. For most ecommerce accounts the difference is 10 to 30 percent of reported conversions; for considered purchases and B2B it can be more than half.
6. Review automated rules and bidding. A cost cap or ROAS target set against 28-day numbers will look like it is failing in January. Adjust targets before the system starts pausing winning ad sets.
7. Check iOS app campaigns separately. If you run app install ads, you will need SKAdNetwork configuration, a single ad account per app, and a limit of nine campaigns per app. That is a separate project with its own deadlines.

Website custom audiences built from the pixel will lose a large share of iOS visitors, and the retargeting pools that have quietly carried many accounts will be smaller and less accurate. Three moves help.

Build first-party audiences you own. Customer lists uploaded from your CRM or email platform, lead form submissions, and engagement audiences from on-platform activity, such as video views and Instagram interactions, are unaffected by browser restrictions. Grow your email and SMS lists now; the budget you spend on list building this quarter pays back as retargeting reach next year.

Lean on broader prospecting. Lookalikes seeded from purchaser lists and conversion-optimized broad targeting have been working well through 2020 as the algorithm improved. Expect that to remain more reliable than narrow interest stacking.

Consolidate. Fewer campaigns with larger budgets exit the learning phase faster, which will matter more when there are fewer signals to learn from. Campaign budget optimization across two or three ad sets is a sensible default.

For the first time in years, Facebook's own reporting will undercount rather than overcount. Decide now how you will judge the channel.

• Google Analytics with consistent UTM tagging gives a second, click-based view. It will also undercount, because it is browser based, but it is consistent across channels.
• Blended metrics. Marketing efficiency ratio, total revenue divided by total ad spend, and blended cost per acquisition across all channels are harder to game and unaffected by any single platform's attribution change.
• Post-purchase surveys. A "how did you hear about us" question at checkout is cheap and surprisingly informative when platform data goes dark.
• Geo holdouts. Pausing spend in a few matched regions for two or three weeks and comparing sales is the simplest form of incrementality testing a mid-sized business can run itself.

Write down the baseline for each of these in December so you can tell a measurement change from a performance change in February.

Doing nothing because the prompt has not shipped. The attribution and event changes are already rolling out and domain verification takes a day to propagate.

Letting the platform pick your eight events. If you have not configured them, Facebook will choose based on past activity, and it may not choose what matters to your margin.

Treating the Conversions API as a privacy workaround. It is a reliability measure for events you already have a legitimate basis to send. Review your privacy policy and consent banner at the same time; CCPA has been enforceable since July and the expectations keep rising.

Panic-switching budgets. Some advertisers will see reported ROAS fall 20 percent in January and move spend to a channel whose reporting has not changed yet. The underlying sales will often be the same. Judge on blended numbers.

Platform changes like this sit at the join between growth, web and technology, so we treat them as one project rather than an ads-team task. Discover inventories every pixel, tag, event and custom conversion on the site, who owns the domain, and which reports leadership actually looks at. Plan sets the event priority list, the Conversions API route for your stack, the new attribution baseline and the blended measurement model. Build is the implementation: verification, server-side events with deduplication, consent updates, audience rebuilds and rule adjustments. Improve runs through the first quarter of 2021 as the prompt rolls out, comparing modeled numbers against blended results and adjusting bids and creative as the data settles. Brand and creative work carries on alongside, because when targeting gets blunter, the message has to do more.

The performance marketing engagement is the closest pattern. A direct-to-consumer brand depended on pixel retargeting and 28-day attribution for most of its reported revenue. We rebuilt tracking with server-side events, consolidated a sprawling account into a handful of conversion-optimized campaigns, grew the owned email and SMS audience as the primary retargeting pool, and moved reporting to blended efficiency with a monthly geo holdout. Reported platform ROAS fell when the attribution window changed; actual revenue per dollar of total marketing spend went up.