An unfortunate truth
If you advertise anything on Meta that falls into a “restricted category,” your Meta Ad performance data is restricted by design.
If you sell any of the following, you cannot guarantee that your in-platform Meta reporting is accurate, ever.
Health and wellness products
Drugs or pharmaceuticals (including GLP-1s or ED medications)
A ton more, listed on Meta’s site
Because of complex inter-state and privacy laws, Meta will hide or obscure data from you on purpose. This means you cannot run creative experiments with the same rigor as other industries.
If you are an advertiser working with a restricted category, hopefully you already know this.
For example: one health and wellness brand we know was only permitted to access pageview events. Add to Carts, Checkout Initiations, and Purchases were all blocked and invisible in the Meta Ads Manager data.
What are you supposed to do?
If you can’t see attributable purchases, you can’t even calculate the ROAS of an ad campaign, let alone the MER.
Lots of people try to use GA4 instead. But GA4 quietly defaults to a model that over-credits your bottom-of-funnel channels.
All of this stacks into a disastrous problem. What we’ve discovered working with these types of brands: independent ad performance data is the solution.
There is no clever workaround. You need good data. You can and should work with Northbeam.
So what’s actually happening if you run ads in these categories?
1. Your Meta data is restricted by design, and the restrictions stack.
Meta is playing a funny regulatory game to stay compliant while also maximizing your ad spend.
They are looking at everything in your account, grading you on your compliance with their transparency rules. Here is a massive list of things Meta could punish you for.
Simply put: Meta classifies your business across three independent “surfaces”: your ad creative (literally the content of it), your landing page, and your “event payload”, the data that Meta collects for each ad.
If you cross too many lines into restricted territory, Meta will begin restricting your data.
Enforcement is cumulative. So a clean ad won't rescue a landing page whose URL mentions a condition, and a clean landing page won't override sensitive data in your event payloads. No funny business.
It’s somewhat obscured, but in our experience there’s three “levels” of restrictions that could appear on your account:
Warning zone: Meta strips custom parameters and anything appended to your URLs after the domain. Attribution and audience quality degrade gradually over days to weeks. Maybe you mentioned a few medical conditions in your ads.
Lower-funnel restrictions: Lower-funnel events like Purchase, Lead, and Schedule get restricted. Lookalikes built on those events stop refreshing. Audiences go stale and the algorithm falls back to broad demographic targeting.
This directly creates the CPM inflation y’all keep complaining about. And it just compounds, getting worse over time into a death spiral. This usually happens when Meta figures out you’re a health or wellness brand.
Full restriction: An immediate blackout. Swaths of your data are fully obscured, like in the case mentioned in the intro. You might only be able to see pageview events, for example. This happens to medical industries mostly.
On top of the event layer, since September 2025 Meta has proactively flagged custom audiences and custom conversions whose names imply a sensitive trait. So if you uploaded a list with “_diabetes_” in the name, good luck.
A previously approved audience can get disabled by its label alone, after which it stops matching and your campaigns quietly stop receiving conversion data.
If you advertise into the EU, assume enforcement is stricter still.
The tl;dr: you can be spending heavily on Meta and still have no trustworthy lower-funnel read on in-platform data.
2. Falling back to GA4 and last click makes it worse
Many advertisers default to GA4 as a cheap/free solution to this tracking issue. But GA4's defaults are built to over-credit the channels that close demand, not the ones that create it.
Google has an incentive to prioritize search performance, after all. Plus the models in GA4 are naturally biased towards results that are easier to track.
Last non-direct click hands 100% of credit to the final touchpoint, usually branded search, direct, email, or SMS.
Someone sees your Meta ad Monday, researches Tuesday, then Googles your brand name Friday to buy. Last click gives Google Search the whole sale and Meta nothing.
The nature of GA4 makes it a dangerous reporting tool for restricted-categories advertisers for two reasons:
GA4's "data-driven" model often isn't running. Data-driven attribution requires roughly 400 conversions per month to activate. Below that, GA4 silently reverts to last click without telling you. Your settings still say "data-driven," your reports still generate, but you’re convincing yourself to tilt budgets away from the awareness channels like Meta that are actually driving the conversion.
The default window is too short. GA4's acquisition lookback defaults to 30 days, so any touchpoint older than a month is excluded outright. If your customers take weeks to decide on your product, that means all your ACTUAL demand generation information is erased from this report.
Listen. Friends don’t let friends use GA4 to run their ecommerce business. Don’t do this to yourself.
3. Your funnel is longer and weirder than a t-shirt brand's
So quit acting like it should. Restricted categories rarely convert in a single checkout. A GLP-1 or telehealth customer path looks more like:
Quiz
Medical review
Prescription approval
Fulfillment
First shipment
Subscriptions
Refills, and so on
Regulatory pressure is pushing these flows to get more structured, not less. Meta will probably clamp down even more. The FDA issued warning letters to dozens of telehealth companies over GLP-1 marketing in early 2026.
This stuff all comes together to make two timing problems that Meta’s (or any ad platform’s) pixel can’t solve.
First, revenue timing and marketing timing are not the same thing. The ad that created demand may have run weeks before revenue actually lands.
Second, much of your lifetime value lives in subscriptions and refills that client-side tracking never sees cleanly. If your reporting can't separate when cash arrived from when demand was created, you'll keep mis-pricing CAC and under-investing in the channels doing the real work.
4. What to build instead: a first-party measurement layer
The brands scaling in these categories in 2026 didn't find a loophole, cause there ain’t one. Meta is too focused on their regulatory situation.
Please stop trying to hack it. Loopholes always close, and if you’re playing in legal gray areas, you’re playing with borrowed time. Your boss is looking to you for stable, predictable revenue. You can’t get there by hacking Facebook ads.
You need a stable source of truth on your ad performance so you can run real experiments.
Five pieces matter:
A first-party data foundation. Combine your own pixel, first-party cookies, UTMs, and backend order data as the revenue source of truth. Don’t rely on in-platform ad performance data any longer, build your own from your own customer and order data. This is the core functionality of Northbeam and why it will be the cleanest representation of performance you have.
Clicks + Deterministic Views, if you can. Northbeam's C+DV approach pairs click data with platform-verified views to read impression-heavy upper-funnel channels. Basically, you can attribute revenue to views of a TikTok ad, for example. No clickthrough required. Deterministic views are verified against real orders, not guessed or modeled.
Long, configurable lookback windows. Match your attribution window to your actual consideration cycle instead of forcing every decision into a 30-day platform default. This way earlier paid touchpoints stay eligible for credit when the conversion finally lands. All of you need this, right now.
Cash and accrual reporting. Cash answers "what revenue came in today?" It’s best for finance and the board. Accrual answers "which touchpoints drove that revenue?" This is better for day-to-day media decisions. Legendary operators use both models because it makes sure you grade your marketing against the results that actually matter.
Unbiased reporting on your ads. Google's data-driven attribution only sees Google's own touchpoints; Meta only sees Meta's. A neutral layer like Northbeam can measure Meta, Google, TikTok, affiliates, creators, email, SMS, and direct all equally. Stop asking each ad platform to grade it’s own homework, that’s why your conversions don’t line up across Shopify, GA4, and your ad platforms.
We didn’t invent these ideas, we’re just building the tool that helps you master them. The zeitgeist of marketing is settling on “triangulation”: multi-touch attribution for daily optimization, marketing mix modeling for budget allocation, and incrementality tests to prove causal lift. We’ve got it all for you.
Bottom line
A restricted-category brand can't run the same measurement stack as a hacky apparel or beauty business. It isn’t the same business, why would you follow the same logic?
You need an independent, first-party layer that handles restricted tracking, captures long journeys, separates cash timing from demand creation, and compares channels fairly.
We built Northbeam for this exact use case. It makes your advertising future-proof. So what are you waiting for?
A practical guide
If you’re still reading, here’s a practical checklist if you’re just starting out a new restricted-category ad account on Meta.
Find out how Meta classified you. Check your domain classification in Events Manager. Remember restrictions stack across ad, landing page, and event payloads. Fix the things that are actually triggering Meta.
Rename anything that implies a condition. Custom audiences and custom conversions get flagged by their names. "GLP-1 buyers" or "anxiety_purchase" will trip the classifier even if the underlying data is compliant. Even if you sell compliant products don’t use “anxiety” as a label or something.
Move conversions server-side. A Conversions API setup with neutral custom event names keeps sensitive health data out of the browser pixel while still feeding Meta an optimization signal.
Build lookalikes from general, non-condition data. Condition-specific audiences are out; genuinely general wellness signals can still work. Just watch the labels.
Expect CPM inflation and lookalike decay after you get flagged, and budget 4-8 weeks for audience quality to recover after a compliance fix.
Stop optimizing against the platform's broken lower-funnel. Anchor your decision-making in Northbeam data, it’s worth the onboarding time.
Extend your lookback to match reality, and split cash reporting from accrual so your CFO and growth leads aren't fighting over one number.
Are you a seven-figure brand that wants to use Northbeam?
Are you looking for full marketing execution and media buying help at the same time?
Through Northbeam’s partnership with Common Thread Collective, qualified DTC brands can now get enterprise attribution plus a dedicated CTC Prophit Engineer managing your insights every day.
This is my preferred solution for high-growth brands that want to join the big leagues. It’s enterprise attribution and marketing execution for aspirational operators.
💸 AppLovin's Axon opens self-serve to every advertiser worldwide this month. The cheap-CPMs window will close once everybody figures this out - be a first mover.
🎯 Meta is testing "Push Delivery to This Ad" to force spend onto one creative. Can’t wait to see this functionality get abused.
🗑️ Is this the end of Shopify? AI-native brands are moving away from Shopify for increased functionality.
🔗 Pinterest added Amazon Storefront linking for creators on June 10. Affiliate revenue is about to go nuts, I hope.
Love this newsletter? Please forward it to your friends, coworkers, or every single resident of New York City. Subscribe here to get this data in your inbox every week.