Apple's tracking restrictions did not kill Facebook advertising. They made measurement less certain and made the platform more dependent on the data you deliberately send it.
Five years on, the accounts that perform are not the ones that found a clever workaround. They are the ones that installed server-side tracking properly, optimised for events they can actually verify, and stopped treating in-platform attribution as fact.
Quick Info
What changed permanently
Attribution is partly modelled, not observed
Default window now
Seven-day click, one-day view
Most important fix
Conversions API alongside the pixel
Second
Sending offline and CRM conversions back
Expect
In-platform and analytics numbers to disagree by 20% to 40%
What did not change
Creative and offer decide results
What actually happened
Apple required apps to ask permission before tracking users across other companies' apps and sites. Most people said no. That removed a large share of the deterministic link between an ad impression and a later conversion on iOS devices.
—Attribution windows shortened, with seven-day click and one-day view becoming the practical default.
—A portion of reported conversions is now statistically modelled rather than directly observed.
—Audience sizes for retargeting shrank, because the events that populate them are partly missing.
—Reporting delays lengthened, so same-day numbers are less reliable than they were.
—Breakdowns by age, gender and placement became less complete for iOS traffic.
None of this stopped the advertising working. It made the feedback loop noisier, which is a real cost but a different one from the collapse that was predicted.
Why your numbers disagree
Meta reports conversions it attributes to ad exposure within its window, including modelled ones. Google Analytics reports sessions with a last-click or data-driven model and no view-through. Your CRM reports what actually closed. All three are answering different questions.
01Meta over-reports relative to analytics in most accounts, because it counts view-through and models the gaps.
02Analytics under-reports Meta's contribution, because it credits the last click and much Meta influence is earlier.
03Your CRM is the truth about revenue and knows nothing about which impression caused it.
04Reconciling them exactly is not possible. Using each for its own job is.
Stop trying to make the dashboards agree. Use Meta's numbers to compare ads with each other and your own numbers to decide whether the channel pays.
The setup that recovers the most signal
This is the practical work, and it is the difference between an account that adapted and one that has been slowly declining since 2021.
01Install the Conversions API alongside the browser pixel. Server-side events are not subject to the same browser restrictions, and the two together give the fullest picture.
02Deduplicate properly using event IDs, or your conversions inflate and the optimisation degrades.
03Verify your domain in Business Manager and configure your priority events, which determines what survives when tracking is limited.
04Send offline conversions back: closed sales, qualified enquiries, phone bookings. For service businesses this is the single highest-value item here.
05Upload CRM outcomes periodically if you cannot integrate directly. Even a weekly file helps.
06Collect first-party data deliberately: email addresses, phone numbers, consented identifiers. That is what the API has to match on.
Most agencies will tell you they have set this up. Check the events manager yourself and confirm server events are arriving with a good match quality score.
How to measure now
A small business does not need an econometric model. It needs three habits.
—Blended cost per acquisition: total marketing spend divided by total new customers. Crude, unarguable, and the number your business actually runs on.
—Ask every enquiry where they found you. Imperfect and frequently more accurate than any dashboard.
—Watch the trend of total enquiries against total spend rather than per-campaign attribution.
—Run occasional off periods on a channel and watch total volume. This is the only real test of incrementality available to you.
Directional truth from four crude measures beats false precision from one dashboard.
What to optimise for
Shorter windows and partial data mean the platform learns from less. Give it the most meaningful signal it can get enough of.
—Optimise for the deepest event with sufficient volume, roughly fifty a week per ad set.
—If purchase volume is too low, optimise for a verified qualified lead rather than a raw form fill.
—Send value with the event where order values vary, so the system optimises for revenue rather than count.
—Consolidate ad sets. Fragmented budgets learn badly at the best of times and much worse on partial data.
What stopped working
—Narrow behavioural retargeting funnels with six stages. The audiences are too incomplete to populate them.
—One-day and same-day performance judgements. Reporting lag makes them meaningless.
—Micro-optimising by placement and demographic breakdown on iOS traffic, where the data is incomplete.
—Very long attribution windows as a way of claiming credit. The default windows are shorter and that is the reality now.
—Treating in-platform return on ad spend as a fact to be reported to a board.
What matters more than it used to
01First-party data. An email list is now an advertising asset as well as a marketing channel, because it is what you can match on.
02Creative and offer quality, because when the platform's targeting signal is noisier, the creative does more of the work.
03Landing page conversion rate, which is fully observable and entirely under your control.
04Follow-up speed, which converts the leads you already paid for.
05Simplicity of account structure, because complexity needs data to justify it and the data got thinner.
Notice that four of those five are things you own rather than things the platform provides. That is the durable lesson from the whole episode.
Where privacy is heading
The direction is consistent: less cross-site tracking, more modelling, more reliance on first-party data and server-side integration. Browser changes, regulation and platform policy all point the same way.
So the right investments are the ones that survive further restriction: your own customer list, your own site analytics, clear consent, and creative that persuades rather than targeting that intercepts.
Frequently Asked Questions
Do iOS privacy changes still affect Meta Ads?
Yes, permanently. Attribution is partly modelled, windows are shorter, and audiences are less complete. The workaround is better first-party signal, not clever targeting.
What is the Conversions API and do I need it?
It sends conversion events from your server to Meta rather than from the browser. Yes, you need it, alongside the pixel, and it is the highest-return technical fix available.
Why does Meta report more conversions than Google Analytics?
Different attribution models, view-through counting and modelled conversions on Meta's side. Both are partly right; use each for what it is good at.
What attribution window should I use?
Seven-day click and one-day view is the practical default. Longer windows mostly serve to flatter reporting rather than to inform decisions.
Is retargeting still viable?
Yes, with smaller and less precise audiences. Keep the funnel to three stages and expect lower reach than pre-2021 guides describe.
How do I know if my ads are actually working?
Blended cost per acquisition, asking enquiries where they found you, and turning the channel off occasionally to see whether total volume moves.
Before You Go
The adaptation is not mysterious: install server-side tracking properly, send your real outcomes back, optimise for events you can verify, simplify the account and judge the channel on blended numbers.
Everything else that changed was reporting comfort. The main guide covers the structure this assumes.
Send the real signal. Judge on blended numbers.
WRITTEN BY TARIQ SALLAM
Marketing Consultant. Entrepreneur. Content Creator.
I'm a marketing consultant, entrepreneur and content creator. I help businesses grow through practical marketing, websites, SEO, content and AI.
More About Tariq →
Keep reading