Skip to content

Cosmetics — France — 2026

Pariderm

Ad spend running against tracking that had quietly stopped telling the truth.

Engagement
Digital transformation
Status
Live
Live
Not public yet
Pariderm event tracking, before and after the rebuildBefore: three separate Meta Pixel installations — one in the theme header, one through a plugin, one through the tag manager — all reporting into Events Manager, producing duplicated and dropped events. After: a single audited Pixel plus a server-side Conversions API feed, reporting deduplicated events.BeforePixel via theme headerPixel via pluginPixel via tag managerEvents Managerduplicated / droppedAfterOne audited PixelConversions API (server-side)one source, verified in the pagesource — deduplicated, andresilient to browser blocking
Three competing Pixel installations, audited back to one, with events reported server-side.

The situation

Pariderm, a French cosmetics brand on WooCommerce, had accumulated years of tracking sediment — Meta Pixel installed through several routes at once, tag manager containers layered over plugin integrations, events firing twice or not at all. The consequence isn't cosmetic: every optimisation decision and every reported conversion downstream inherits the error.

What we built

  • A rebuilt measurement layer: the conflicting installations audited and stripped back to one
  • The Pixel reconnected cleanly, with each stage verified in the live page source rather than trusted
  • Conversions API set up so events are reported server-side and survive browser-level blocking

How we approached it

We removed before we added.

The instinct with broken tracking is another integration. We paused and audited every existing source first, then rebuilt from a verified-clean state — which is why the fix held.

We treated marketing performance as an engineering problem.

Attribution accuracy is a data pipeline question, and it belongs with the people who can read the page source.

Stack

WooCommerce, Meta Pixel, Conversions API, Google Tag Manager