Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.datalyr.com/llms.txt

Use this file to discover all available pages before exploring further.

Meta, TikTok, and other ad platforms classify ad accounts as “Health & Wellness” based on the semantic content of the events you send them — content_name, content_category, URL paths, etc. Once classified, your account loses advanced targeting, lookalike audiences off purchases, and delivery priority. Sensitive Category Redaction strips the trigger fields before events leave DATALYR. Your hashed email/phone, click IDs, value, and order_id still go through, so attribution and matching keep working.

​
What It Does

Strips category-trigger fields from outgoing CAPI payloads:
  • content_name, content_category
  • Health-keyword product titles (Vitamin D 5000 IU, Ashwagandha 600mg, Magnesium Glycinate, etc.)
  • Health-keyword URL paths (/products/sleep-gummies-30ct → /event)
  • Sensitive tokens inside contents[] and cart_data.items[]
Preserves everything Meta needs to match and attribute:
  • Hashed email and phone
  • Click IDs (fbc, fbp, ttclid, gclid)
  • Purchase value and currency
  • Order ID
  • External ID
  • IP address and user agent
Applies across every CAPI destination at once:
  • Meta Conversions API
  • TikTok Events API
  • OpenAI Ads Conversions
  • Google Ads Conversions

​
When to Use It

Turn this on if:
  • You sell supplements, vitamins, protein, collagen, probiotics, CBD, hemp, or similar wellness products
  • You sell products for specific health conditions (sleep, anxiety, hair loss, joint pain, weight management, hormonal support)
  • You ship via telehealth, pharmacy, or prescription workflows
  • Your Meta Ads Manager shows a “Health & Wellness category restriction” notice
  • Your EMQ score dropped without any tracking changes on your end
Leave it off if your store sells non-sensitive categories (apparel, electronics, home goods, beauty without wellness claims). The redactor uses substring matching and may catch ambiguous keywords like “vitamin” in a juice product name — see Tradeoffs below.

​
Enabling Redaction

Go to: Settings → Meta CAPI Toggle Strip health-vertical fields on. 
1. Sign in to app.datalyr.com
2. Settings → Meta CAPI
3. Toggle "Strip health-vertical fields" → ON
4. Done — effective on the next event sent
One toggle controls all four CAPI platforms. No per-rule configuration is required.

​
What Gets Redacted Per Platform

PlatformField-strip targetURL scrub
Metacustom_data object (content_name, content_category, etc.)event_source_url, referrer_url
TikTokproperties object + each entry in contents[]page.url, page.referrer
OpenAIdata objectsource_url (web only)
Googlecart_data.items[] (each item’s product fields)N/A — Google CAPI has no URL field
When a URL scrub fires, the path is replaced with /event while the origin is preserved (so Meta’s attribution still has a valid domain). DATALYR’s keyword library covers four categories: Product categories — Health & Wellness, Vitamins & Supplements, Sports Nutrition, Sexual Wellness, Pharmacy, Telehealth Active ingredients — vitamin, magnesium, creatine, ashwagandha, collagen, probiotic, melatonin, CBD, glucosamine, lions mane, reishi, and ~30 more Condition-implying terms — anxiety, insomnia, weight loss, fertility, thyroid, joint pain, arthritis, migraine, acne, hair loss, and similar Dosage patterns — 500mg, 5000 IU, 60 capsules, 30 gummies, 50 billion CFU, 5g creatine A field with any match is stripped. URL paths with any match are scrubbed.

​
How Your Attribution Stays Intact

Because the matching signals (hashed em/ph, fbc/fbp, ttclid, gclid, IP, UA, external_id, value, order_id) are not redacted, your ad platform can still:
  • Match the event to a user — identity resolution via hashed PII works the same
  • Attribute the conversion — click IDs and visitor matching work the same
  • Track revenue — value and currency flow through
  • Deduplicate vs. pixel — order_id and external_id are preserved
What changes: the ad platform’s classifier loses the semantic context it would have used to flag the account, but the targeting and bidding work the same.

​
Verifying

After turning it on, fire a test purchase: 
1. Open any conversion rule (Meta, TikTok, OpenAI, or Google)
2. Click "Test postback"
3. Pick a recent purchase from a health-vertical product
4. Send
In Meta Events Manager → Test Events, the received event should show custom_data without content_name or content_category. The value, currency, hashed email, and order_id should all be present. Repeat in TikTok Events Manager and OpenAI Events Dashboard as needed.

​
Tradeoffs

Broad substring matching. A product named Daily Vitamin Smoothie matches the vitamin keyword and gets redacted, even if it’s a juice product. This is intentional — opt-in workspaces accept conservative behavior over selective edge cases. If false positives become a problem, contact support to adjust the keyword list for your workspace. Cross-platform application. The same toggle drives Meta + TikTok + OpenAI + Google. You can’t enable redaction for Meta only — by design. The cost of running it on platforms that aren’t actively flagging health verticals (Google primarily, OpenAI defensively) is negligible. Worker cache TTL. When you flip the toggle, the change propagates immediately. If our background cache hits a brief outage on save, the change can lag up to 5 minutes — in practice, this almost never happens. No per-rule override (in dashboard). If you need surgical control (redact one Meta rule but not another), reach out to support. Per-rule overrides are programmatically supported but not surfaced in the dashboard.

​
Why This Matters

Meta’s 2025 enforcement is payload-semantic, not event-name-based. Renaming a Purchase event to event_123 doesn’t help — Meta inspects the content_name, content_category, and URL of every event. A single event with content_name: "Testosterone Booster 90 caps" can flag the entire ad account. Once flagged, the ad account loses:
  • Lookalike audiences off purchases (the highest-value optimization target)
  • Detailed Targeting expansions
  • Custom audience matching at full match rate
  • Delivery priority in auction (the algorithm deprioritizes flagged accounts)
Brands typically see 20–40% drops in attributed revenue after a flag lands. The flag is reversible after a quiet period (no flagged signals for ~30 days), but the lost ad spend during that window is unrecoverable.