Why Meta Stopped Optimising Your Wellness Brand's Ads
Performance degraded over a few weeks. CPAs drifted upward. Lookalikes stopped refreshing properly. Reporting lost detail nobody remembers switching off.
Nothing in Ads Manager says why, because the cause is not in Ads Manager. Meta has classified your website as a health and wellness data source, and is now restricting what it accepts from your pixel and Conversions API.
For Indian D2C this catches far more brands than expect it — Ayurvedic skincare, hair oils, supplements, sleep and digestion products, intimate wellness. None of them think of themselves as healthcare. Meta's classification does not ask.
The Symptom Before the Diagnosis
What brands notice, roughly in order:
Cost per result climbs without a corresponding change in auction conditions
Custom audiences built on URL rules or custom parameters stop updating
Reporting breakdowns that used to be available return less detail
Advanced matching coverage falls
Catalogue items stop updating through the pixel
A custom conversion you have used for months becomes unusable in new campaigns
The last one is the clearest tell. Meta's own developer documentation states that any custom conversion suggesting specific health conditions — it gives arthritis and diabetes as examples — "will be flagged and prevented from being used to run ad campaigns."
The error reads: "This custom conversion is blocked because it may contain information (e.g., health, financial) not allowed under Meta's terms."
If you have seen that message, you are not dealing with an algorithm dip.
What Meta Classifies as Health Data
Meta describes a health and wellness data source as one associated with medical conditions or specific health statuses, provider-patient relationships, services for accessing personal health information, or health-related products and services.
Its prohibited information list covers diseases and medical conditions, mental and behavioural health, sexual and reproductive health, medical procedures and treatments, prescription medications — and, critically for this category, over-the-counter products and supplements for specific medical conditions.
That last clause is where most Indian D2C wellness brands get caught.
A hair oil marketed for hair fall. A supplement for immunity or sleep. A churna for digestion. A product page URL reading /collections/acne-treatment. None of this is healthcare in the brand's own understanding, and all of it reads as a health-status signal to a classifier.
The classification applies to the data source — your domain — not to individual products. One product line can place the whole site in the category.
How Your Own Setup Makes It Worse
Classification determines whether Meta scrutinises your payload. What is in the payload determines how much gets removed.
Four places restricted data hides, and brands rarely put it there deliberately.
URL paths and query strings. Meta restricts transmitting anything in a URL after the domain under core-setup restrictions, and notes that UTM parameters may carry prohibited data from the landing pages viewed. A URL like /quiz/result?concern=hairfall sends a health signal with every page view.
Custom event names. The name itself is data. An event called AcneQuizCompleted is a condition signal regardless of what it carries.
Custom parameters. Anything describing a condition, symptom, treatment, or inferred health status.
Product and content fields. content_name carrying a product title with a condition keyword, or content_category set to a therapeutic classification.
Parameter patterns to avoid
Pattern | Example |
|---|---|
Condition or symptom names |
|
Treatment or regimen |
|
Quiz and assessment outputs |
|
Body and health metrics |
|
Service or appointment detail |
|
Product names implying a condition |
|
Therapeutic categories |
|
Any personally identifying field | email, phone, name in plain text |
That last row is a separate problem with the same symptom — and it also sits under DPDP, which is covered in DPDP for D2C analytics.
What Actually Happens to a Restricted Event
Meta applies restrictions at three levels, and the level determines the damage.
Level | What Meta restricts | Practical effect |
|---|---|---|
Core setup | Custom parameters, and anything in a URL after the domain | Only predefined standard parameters survive — |
Standard event restrictions | Specific standard events — Lead, AddToCart, Purchase | Optimisation loses the conversion signal it was bidding toward |
Full restrictions | All event sharing from the source, in certain regions | No signal at all |
Most brands land in core setup, which is why the symptom is degradation rather than failure. Campaigns keep running. They just learn from less.
Worth knowing: Meta notes that repeated prohibited-information notifications may place a Business Tool into core setup for at least 90 days. This is not a setting you can switch back after cleaning the payload — there is a dwell period.
The downstream effect on match quality is covered in our post on improving Meta Event Match Quality. A restricted source and a poorly matched source produce similar symptoms through different mechanisms, and it is worth establishing which you have before fixing either.
A Before and After Payload
Illustrative, built from Meta's documented CAPI structure rather than any specific engagement.
Before — a Purchase event from a wellness brand
json
Five problems. A custom event name describing a condition. A URL carrying the concern and a quiz result. A content name and category stating a treatment. A quiz output. And plaintext PII that should never be in custom_data at all.
After — the same purchase, cleaned
json
What changed: standard event name, URL truncated to domain, identifiers moved into user_data and hashed, product referenced by SKU rather than descriptive name, and every condition-linked field removed.
What is lost: granular product reporting in Meta. What is gained: events that arrive intact and continue optimising.
The Remediation Sequence
Order matters — cleaning the payload before you know what is being flagged wastes a cycle.
1. Find out what Meta is telling you. Events Manager → Diagnostics for restriction notices. Settings → Manage data source categories to see your classification. Settings → Data controls → Manage event blocking for blocked events. Where you see "Event parameters blocked," click Review — Meta names the parameters and pages affected, though not the values.
2. Audit what you actually send. Fire a test purchase and inspect the full payload — event name, source URL, every custom parameter. Most brands find at least two fields they did not know were being sent, usually added by an app rather than deliberately.
3. Rename events to Meta's standard set. Purchase, AddToCart, ViewContent, Lead. A descriptive custom name is a liability for no gain.
4. Strip condition-linked parameters. Everything in the table above. Replace descriptive product fields with SKUs.
5. Truncate URLs. Send the domain. If you need campaign attribution, carry it in your own analytics rather than in the Meta payload.
6. Move identifiers into user_data, hashed. Email and phone belong there, SHA-256 hashed, never in custom_data as plaintext.
7. Re-verify, then wait. Confirm the clean payload in Events Manager, then allow for the dwell period before expecting restrictions to lift.
What Server-Side Does and Doesn't Solve
A common and expensive misunderstanding: Conversions API is not a route around this.
Meta's data-sharing checks apply to both the pixel and the Conversions API. Sending restricted data server-side does not make it permitted — it makes it restricted data arriving by a different channel.
What CAPI genuinely helps with: it gives you a single controlled point to construct the payload, which makes stripping fields far easier than editing tags across a theme. It also improves delivery of permitted data that browser-side collection loses to blockers and tracking prevention.
So the sequence is: clean the payload first, then use server-side to deliver the cleaned version reliably. Doing it the other way round delivers the problem more efficiently.
Meta also states that its filters are not a substitute for your own compliance mechanisms — the payload remains the advertiser's responsibility.
How to Tell Whether It's Fixed
Four checks, in order.
Restriction notices cleared. Events Manager Diagnostics should no longer show active restriction messages. Allow for the dwell period.
Event Match Quality. Check the score per event before and after. A cleaned payload with identifiers properly hashed into user_data should hold or improve match quality even while custom parameters are removed — if it falls, identifiers were lost in the cleanup.
Delivery and parameter coverage. Events Manager shows which parameters are arriving. Confirm the standard set is intact.
Reconciliation. Compare Meta's reported purchases against your platform order count over 30 days. Our post on why Meta Ads Manager and GA4 never show the same number covers the expected gap and how to read a change in it.
Then monitor monthly. New apps, new product lines and new landing page structures all reintroduce the problem. A clean payload in November is not a clean payload in March.
Why This Costs More Than It Looks
A restricted source does not fail loudly. Campaigns keep spending, reports keep populating, and the degradation looks like market conditions or creative fatigue.
So brands respond by changing creative, adjusting budgets and rebuilding audiences — none of which address a classifier removing the signal those decisions depend on. Months of optimisation work gets spent on the wrong problem.
That is the same pattern we describe in what broken tracking actually costs a business: the expensive failures are the ones that keep producing plausible numbers.
Running a wellness or Ayurvedic brand and seeing performance degrade without explanation? Get a GA4 and Meta audit from FunnelFreaks.