Tracking diagnosis — what GA4 misses today
Every number elsewhere in this deck is biased by five tracking conditions. This page consolidates them so a reader can calibrate the rest of the deck without surprise.
The five conditions below are not isolated bugs. They sit in different parts of the platform — key-event flags, UTM hygiene, duplicated events, the LinkedIn Insight Tag, and a property with no key events configured at all. Each one is fixable, and each one has a fix routed to the roadmap in §10.
1. form_submit is not flagged as a key event
GA4 records 1,714 form_submit events on ProDir and 5,389 on GSI in the 12-month window, but isKeyEvent is (not set) on both. The only flagged key events on the property are cart_view (a partial slice — see condition 3 below), financial_literacy_email_sent, and thermogenics_article_conversion. The result is that GA4's "key events" column under-counts real conversion intent everywhere it appears in the platform UI. The page-by-page picture of which form submits matter lives in §03 (ProDir form pages) and §04 (GSI form pages); the fix is in §10-P0.1.
2. UTM hygiene gap on owned campaigns
Direct is 79.1% of ProDir sessions and 74.6% of GSI sessions. The (direct) / (none) source/medium row alone equals the Direct channel total, so there is no intermediate misclassified source absorbing the volume. Brand awareness genuinely drives a large share of this — visitors typing the URL or arriving via saved bookmarks. Owned email and LinkedIn outreach without UTM tags also fold into Direct invisibly, and from GA4 alone we cannot separate true Direct from leak. The fix is in §10-P1.
3. cart_view appears with two key-event flags simultaneously
The same event name cart_view is recorded twice on both sites — once with isKeyEvent=true and once with (not set). The split persists in the year-over-year window, so this is not a recent regression; it points to two streams or two GTM containers sending the same event under different configurations.
| Property | (not set) | key=true | Total |
|---|---|---|---|
| ProDir 12m | 659 | 210 | 869 |
| GSI 12m | 2,240 | 1,133 | 3,373 |
GSI's flagged share is 1,133 of 3,373, which is roughly 33.6% of the actual cart_view volume. The "purchase intent" measurement reported by GA4 reflects only that third on GSI. The fix is in §10-P0.2.
4. LinkedIn paid attribution is broken
GA4 sees 12 of LinkedIn's 754 landing-page clicks (about 1.6% of what LinkedIn says it sent). LinkedIn's own platform reports 0 conversions across the $861 USD spent on 13 campaigns in the 12-month window. GA4 separately records 313 form_submit events on sessions sourced from linkedin.com / referral — real activity that LinkedIn cannot see because the Insight Tag is either not installed on the conversion pages or not configured to fire on conversion events. The reconciliation is in §06; the fix is in §10-P0.4.
5. BoardConnex marketing site has no key events
1,494 sessions on boardconnex.com over the 12-month window, 0 key events configured. There is no form_submit, no cart_view, no book_a_demo event in the property. The November–December 2025 traffic spike (663 sessions across the two months, against a baseline of 53–124 per month) is unaccounted for. The marketing site is being measured for traffic but not for outcome. The fix is in §10-P0.3.
Reading the rest of the deck through these conditions
If these five gaps stay open, the consequences ripple through every dashboard a reader will open after this deck.
- Every "key events" column inside GA4 understates real conversion intent on both ProDir and GSI, because
form_submitis the dominant lead signal and is not flagged. - LinkedIn cannot optimize delivery toward results while the platform sees zero conversions, so the $861 USD spent in the window produced no algorithmic feedback loop.
- The November–December BoardConnex spike of 663 sessions is invisible in attribution. Whether it produced demos, pricing-page contacts, or outbound interest is not knowable from the platform.
- UTM-less email and outreach disappear into Direct, so a real campaign cannot be distinguished from genuine brand-awareness traffic in any channel report.
- GSI's
cart_viewvolume reads as 1,133 in standard reports when the actual figure is 3,373 — a roughly 67% understatement of purchase intent.
Where to go next
The roadmap in §10 sequences the four P0 items so the four conversion-measurement gaps close first. The four P0 items are independent and can run in parallel without conflict.