How a multi-million-euro B2B merchandise brand fixed the data its ad platforms were learning from, then taught Google, Meta and Microsoft the difference between a cheap lead and a valuable one.
Monday Merch · Corporate merchandise · Europe · mondaymerch.com
01 · The problem
Monday Merch sells branded merchandise to companies across Europe. Prospects research on the marketing site, move to a separate products storefront to build a quote, and the deal is then worked by a sales team in Odoo. Paid media on four platforms feeds the top of that funnel, and every platform was optimising toward whatever the tracking told it.
July
Audit the full tag estate, remove the defects producing false data, standardise the event model, and give Google Ads a real funnel to optimise toward instead of a single count.
August
Connect the CRM to every ad platform through a first-party server, so each opportunity is reported with its expected value and updated as the deal moves.
02 · Phase 1, fix the foundation
The container had accumulated years of tags across platform migrations, with a server-side layer added later. Nothing was obviously broken. The problems only show up when you compare what the tags send against what the business actually did.
Critical
Found. Two parallel tag sets per event, some of them double-triggered. Two competing GA4 configurations. And, entirely outside the container, the Odoo CRM was pushing its own lead events straight to GA4, a second pathway that no amount of container auditing would have surfaced.
Why it mattered. Every efficiency metric the business used was calculated on an inflated base. 22 duplicate and legacy tags became 13 single-purpose ones, and the CRM-side duplication was closed separately.
High
Found. The Conversion Linker listed only the marketing domain. The products storefront, where quotes actually happen, was not in the linked-domains list.
Why it mattered. This was the single largest driver of the 50% unattributed traffic. One configuration field, verified by a live click-through test, and the buyer's journey became one session again.
High
Found. Quote requests, contact forms and calendar bookings each sent a differently-named lead event. None matched GA4's standard, so a lead-led business showed zero leads in the report built for exactly that purpose.
Why it mattered. Standardised to one event fired from three sources. The downstream growth report, which had been counting the old names, was corrected in the same pass.
Medium
Found. A single generic quote action. No add-to-cart, checkout, booking or lead-form conversions, and no value attached to any of them.
Why it mattered. We built five dedicated conversion actions, each with its own trigger and label, and attached an estimated euro value to lead submissions derived from the budget range the prospect selects. The first step from counting to valuing.
| Phase 1 outcome | Before | After |
|---|---|---|
| Sessions with no identifiable source | Roughly 50% | 13% |
| GA4 tags in container | 22 duplicate and legacy | 13 single-purpose |
| Lead duplication pathways | 2, running independently | 0 |
| Google Ads conversion actions | 1 generic | 5, per funnel stage, value-bearing |
| Event naming | Non-standard and inconsistent | GA4 standard taxonomy across the container |
| Conversion delivery | Browser tags, per platform | One server-side stream, routed to every platform |
03 · Phase 2, teach the platforms what an opportunity is worth
Clean website data solves half the problem. The other half is that the most important events in a B2B funnel, an opportunity being qualified, moving stage, or being won, happen inside the CRM, days or weeks after the click, where no pixel can see them. Phase 2 built the pipeline that carries those moments back to every ad platform with a value attached.
04 · What happened next
Two months on, the tracking foundation is clean and the platforms are being fed opportunity value instead of raw lead counts. The early read comes from two places: a year-on-year comparison, and a set of controlled bidding experiments run inside Google Ads.
>3×
New-customer opportunities in July and August versus the same weeks the previous year
Client CRM reporting. Year on year, not attributed to any single cause.
9 vs 2
Won deals across five value-based bidding experiments, treatment versus control
Early results, with one market driving most of the lift. Directional, not conclusive.
~3×
Won value from the value-bidding treatment versus control, at roughly 75% lower spend per win
Same experiment set, as measured by Google Ads experiments.
05 · What's next
Older direct CRM-to-platform connections still run alongside the new pipeline. A deliberate cutover once the new signals are proven, so nothing gets counted twice again.
The early experiments are encouraging but concentrated in one market. Next is expanding across countries and campaign types, and moving the Meta and Microsoft budgets onto the new CRM signals.
At this level of spend, platform attribution is no longer enough. A Bayesian MMM will show which channels drive incremental revenue, and put a number on what the tracking work itself was worth.
How this was verified. Container version history across six releases; live Tag Assistant sessions under both granted and denied consent; GTM Preview with one qualifying and one non-qualifying event before every publish; a live cross-domain click-through test; inspection of outbound API request and response bodies for every platform, not just tag status; and reconciliation of the downstream growth report against the corrected event names. Results reported from the client's own CRM reporting and Google Ads experiments.
Most accounts we audit have at least one of the six defects in this study. Start with what people actually searched before they clicked. The free audit connects to your Google Ads account and shows you in minutes.
Run a free Google Ads audit