What your tags actually record
Measured live by wrapping dataLayer.push and fbq() in the page, walking the real funnel, then parsing the GTM container tag by tag across four fetches. Container version 10, as served 3 September 2026. No order was placed, so purchase is listed as configured rather than observed.
Tag ManagerGTM-WLHRZ3LF14 tags, container v10
GA4G-90BHGQN4VD7 ecommerce tags, client-side only
Google AdsAW-178028508521 conversion + linker
Meta Pixel1912695409662042PageView + 3 hand-written tags
Meta CAPI Gatewayarmedrunning on Cloud Run
Clarityt7vrr1xj2zin the app, not in GTM
The funnel, step by step
Of the seven steps between landing and checkout, Meta sees two. Google sees five of seven and is blind at the two points where shoppers actually drop out.
The five that cost the most
- One bracket is costing Meta its value optimisation. The Meta AddToCart tag reads its value from a Data Layer Variable named
ecommerce.items[0].price. GTM resolves dot paths only and never parses [0], so the value arrives empty. We captured the live call and there is no value field on it. The same container gets it right elsewhere with ecommerce.items.0.price.
- Tags only load after the visitor interacts. The GTM loader waits for a click, scroll or mouse move. We loaded the category page, sat still for 15 seconds after it finished rendering, and counted zero analytics requests. One scroll, and
gtm.js started 556 ms later followed by 16 analytics calls. Anyone who lands and leaves is never counted, so bounces do not report. The Google Ads conversion linker and Microsoft Clarity sit behind the same gate, so a fast-bouncing paid click never gets its gclid stored.
- Meta never fires ViewContent. Not on a fresh product page, not on an in-app one. Catalogue and Advantage+ ads cannot retarget anyone who looked at a saree and did not add it, which is the largest remarketing audience any store has.
- Meta never fires InitiateCheckout, so there is no checkout-abandoner audience. A
BuyNow event fires instead, which is not a Meta standard event, and the same trigger has no GA4 tag at all.
- Nothing records a product-tile click or a cart view. No
select_item and no view_cart trigger exists, and the search term is never captured. There is no way to tell which collection, rail or banner earns its place.
Credit where it is due
The GA4 ecommerce work is real and competent. The view_item payload carries brand, colour variant, discount and value correctly. A Meta Conversions API Gateway is running. This is a setup that is roughly 60% built, and the missing 40% happens to be the part that makes advertising measurable.
The catch nobody mentions
Colour-level reporting is impossible today, and not because of the tags. Because every colourway is a separate product rather than a variant, GA4 sees eight unrelated products where there are really two designs in four colours each. No tagging fix changes that. Which colour of a print sells best only becomes answerable once the catalogue is modelled as products with variants, which is the same change that fixes the SEO.
What we could not see from outside
Everything above describes the published GTM container and what we observed firing in a browser. GA4 Enhanced Measurement is switched on in the GA4 admin, not in the container, and is invisible from outside. A GTM workspace could also hold unpublished tags. So every gap above is stated as “no trigger exists for this”, never as “GA4 has no data for this”.
