The Attribution Problem Nobody Wants to Talk About
Every agency has had the conversation: the client wants to know which channel is "really" driving conversions. The honest answer — "it's complicated, and your current data doesn't support a definitive answer" — is not what they want to hear.
The more common (and problematic) answer: pull last-click conversion data from each channel's native platform and present it as truth. The result: every channel takes credit for the same conversion, total attributed conversions exceed actual conversions by 2–5x, and nobody is making good decisions.
Why Attribution Data Is Hard
Cross-device journeys: A customer sees a Facebook ad on their phone, searches on a laptop, reads a review on a tablet, and converts on their phone again. Most attribution models see only the last touch.
Walled garden reporting: Meta, Google, and TikTok each report conversions using their own measurement — and they each take credit. Adding up their reported conversions gives you more conversions than actually happened.
Cookie and tracking limitations: iOS privacy changes, cookie deprecation, and adblockers mean 20–40% of digital journeys are invisible to pixel-based tracking. Models built on incomplete data produce unreliable conclusions.
Sohovi automatically detects PII in your datasets — emails, phone numbers, SSNs — all processed client-side so your data never leaves the browser.
Lookback window inconsistency: A 30-day lookback on one channel and a 7-day lookback on another means you're not comparing the same thing when you compare performance across channels.
What Good Attribution Data Requires
For multi-touch attribution to produce trustworthy insights, you need:
-
A single conversion source of truth: Your actual conversion data should come from your analytics platform (GA4) or your backend, not from each ad platform's native reporting. Reconcile ad platform conversions to actual backend conversions.
-
Consistent tracking parameters: Every ad from every channel uses UTM parameters consistently. Inconsistent UTMs mean some touchpoints are invisible in your data.
-
A unified customer ID: The ability to connect touchpoints from the same user across sessions and devices. This typically requires a logged-in user journey and is unavailable for anonymous browsing.
-
A defined attribution model: Choose a model (linear, time-decay, data-driven) and apply it consistently. Changing models mid-analysis to get a better story is the agency equivalent of cherry-picking.
Setting Client Expectations Honestly
The most valuable thing an agency can do on attribution is be honest about its limits:
"Our best data shows Google Search drives the most last-click conversions. Meta and display appear earlier in the funnel based on view-through data. We can't tell you definitively how much each channel contributed to the final decision — no one can with certainty given current tracking limitations. Here's what we're confident about, and here's what we're measuring more closely."
Clients who are given honest, well-reasoned analysis trust their agencies more — not less.
