Attribution is the hardest unsolved problem in marketing. In 2026, three things broke the old model: cookie deprecation, AI-mediated search, and platform walled gardens. Here's the stack that works now.
Layer 1: First-Party Data Foundation Server-side GA4 with enhanced conversions CRM linkage (HubSpot, Salesforce, Pipedrive) Customer Data Platform (Segment, RudderStack) if scale demands it
Layer 2: Platform-Native Signals Enhanced conversions in Google Ads Conversions API for Meta LinkedIn Conversions API TikTok Events API All fed server-side, all tied to first-party identifiers.
Layer 3: Causal Validation Geo-incrementality tests every quarter on top channels Conversion lift tests inside ad platforms Marketing Mix Modelling (MMM 2.0) — Bayesian, AI-accelerated
Layer 4: Multi-Touch Attribution (MTA) Used only as a directional signal — never as the source of truth.
Trusting any single platform's reported ROAS Last-click attribution as the basis for budget decisions Client-side-only conversion tracking Treating MTA as gospel
No attribution system is "right." The goal is to be directionally correct, fast, and continuously validated — not perfectly accurate. Spend more energy on causal tests than on attribution dashboards.
Triple Whale (eCommerce) — best modern MTA + MMM stack Northbeam — strong MMM for DTC Funnel.io — solid data warehouse for marketing Recast — AI-driven MMM for mid-market and up
Book a free attribution audit — we'll map your stack and find the leaks.