Every conversion, attributable.
Server-side tracking, Conversions API across every platform, Consent Mode v2, and an attribution model your CFO can defend — the measurement spine the post-cookie era requires. If we can’t measure it, we don’t trust it.
The market reality in 2026
Third-party cookies in Chrome are functionally gone — Privacy Sandbox is rolling out and the browser-side pixel is no longer a reliable signal. iOS 17+ tightened the ATT framework further, and Meta + TikTok have moved most attribution to modeled conversions backed by Conversions API signals. The EU has made Consent Mode v2 a hard requirement for using Google’s ads ecosystem, and Privacy Sandbox + AT&T-style restrictions are spreading to other markets.
What used to be a “configure GA and move on” job is now infrastructure. Server-side tagging, deduplicated cross-platform CAPI, first-party data pipes, modeled-conversion configuration, and a defensible attribution model are the table-stakes spine. Brands without this spine are silently bleeding signal — their attribution looks fine on the dashboard, but the data their algorithms optimize against is increasingly fictional.
How I deliver measurement work
-
01
Audit before any wiring
Map the current pipeline end-to-end: what pixel fires where, what server endpoint receives it, what the dedupe key is, what the consent state actually is at the moment of fire, what the data-loss rate looks like compared to back-end source of truth. The audit produces a leak list ranked by revenue impact — not a tech-debt wishlist.
-
02
Build the spine
GA4 + GTM Web Container + GTM Server Container (Google Cloud Run or stape.io), with Meta CAPI / Google Enhanced Conversions / TikTok Events API / LinkedIn Conversions API wired through the server — deduplicated by event_id so the platforms credit the same conversion exactly once. Consent Mode v2 wired in. UTM taxonomy locked. First-party data infrastructure prepared for Privacy Sandbox.
-
03
Stabilize + document
Debug across browsers, devices, and consent states. Validate end-to-end — from click to landing pixel to server endpoint to platform dashboard. Document the data model so the next operator (or your in-house team) can run it. Build data-quality alerts on volume drops, dedupe failures, consent state anomalies.
-
04
Operate
Monthly attribution model review — data-driven attribution vs last-click vs custom. Quarterly QA across the platform stack. Pre-flight checks for every new campaign launch. Annual data-pipeline audit so the spine doesn’t calcify as platforms evolve.
The stack
GA4 · GTM Web Container · GTM Server Container (Google Cloud Run / stape.io) · Firebase · BigQuery export for raw events
Meta Conversions API · Google Enhanced Conversions / Enhanced Conversions for Web · TikTok Events API · LinkedIn Conversions API · Pinterest Conversion API · Snap Conversions API
Google Consent Mode v2 · Cookiebot / OneTrust / custom CMP · IAB TCF v2.2 · first-party data infrastructure for Privacy Sandbox readiness
Looker Studio · BigQuery + SQL · GA4 attribution models · data-driven attribution · incrementality / lift studies · custom dashboards for executive reporting
What gets measured
- Attribution accuracy% of conversions correctly attributed to the click that produced them
- Data-loss ratebrowser-pixel signal lost to ATT / Privacy Sandbox / ad-blockers
- Server vs browser event match rate98%+ target on dedup-by-event_id
- CMP consent acceptance ratebenchmarked against vertical norms; banner UX tuned for it
- Time-to-insighthours from new campaign launch to clean conversion data
- Pipeline uptimedays/month with full conversion fidelity; alerts on degradation
Proof from the work
Every flagship and high-ticket engagement runs on a custom measurement spine. A selection:
A global skincare leader
Multi-market measurement layer custom-built so awareness reach could be tied to branded-search lift and downstream consideration — not just impressions.
Multi-market attribution modelA premium D2C tea brand
Measurement stack rebuilt to survive iOS attribution loss — Meta CAPI + server-side GTM + event_id dedup. Recovered ~24% of signal that was previously dropping to ATT.
~24% signal recoveryA vertical-SaaS platform
Multi-touch attribution model bridging Google Ads, LinkedIn, and content-led inbound — with closed-ARR feedback into HubSpot. Demo-to-close trace was the deciding factor in budget defense.
4.2× demo requestsCommon questions
What does server-side tagging actually cost?
Hosting runs roughly USD 50–200/month for most mid-market clients — stape.io is the easiest entry point, Google Cloud Run is the path for higher-volume sites. Setup time is 2–3 weeks for a clean build, longer for complex e-commerce or multi-domain setups. The ROI is measured in recovered conversion signal, which usually pays back the build cost in the first month of optimized media spend.
Do I need a Consent Management Platform (CMP)?
If you serve EU traffic — yes, it’s mandatory. Google requires Consent Mode v2 to use audience features for EU users. For non-EU traffic, it’s strongly recommended: privacy regulation is expanding (Brazil LGPD, India DPDPA, California CCPA/CPRA), and showing consent intent improves trust and signal quality.
Will modeled conversions hurt my ad performance?
No — if anything, they recover signal that ATT and Privacy Sandbox removed from the browser-side pixel. Consent Mode v2 in particular leans on Google’s machine learning to fill in conversion data for users who didn’t consent to ad cookies. The math improves; the visibility into individual conversions decreases. That’s the trade.
How long does a measurement audit take?
The audit itself takes about 1 week — pipeline mapping, gap analysis, leak quantification. Remediation depends on what we find: typically 4–8 weeks to fully rebuild the spine, including server-side container, CAPI wiring across platforms, Consent Mode v2 implementation, and dedupe validation.
The other modules in the system
// CARD 03 / MEASUREMENT Back to the Measurement card on the home page
Want to audit your measurement spine?
The audit takes a week and produces a leak list you can hand to any operator. The first 30 minutes are on me.