Why Attribution Is Broken and What to Do About It

The Attribution Illusion

Every attribution model tells a story about which channels drove your conversions. The problem: these stories are all wrong in different ways. Attribution models don't measure true causal contribution — they measure correlation between touchpoints and conversions using rules that are either arbitrary (position-based) or opaque (data-driven).

Understanding attribution's fundamental limitations is the starting point for doing measurement better. Our marketing analytics team takes a measurement triangulation approach that goes beyond attribution.

Why Attribution Models Are Fundamentally Limited

The Incrementality Problem

Attribution models assume every touchpoint was necessary for the conversion. But many conversions would have happened anyway. A user who was definitely going to buy your product and happens to click a retargeting ad on their way to your site — did the ad cause the purchase?

Attribution says yes. Incrementality testing (which measures what would have happened without the ad) often says no — 30–50% of attributed conversions happen without the channel doing anything causal.

The Cross-Device and Cross-Browser Problem

Most tracking is device and browser-specific. A user who sees your ad on their phone, researches on their laptop, and buys via their tablet appears as three separate users in your analytics. Any attribution model applied to this fragmented data is working with incomplete paths.

iOS and Privacy Changes

iOS 14.5+ App Tracking Transparency, Safari ITP, Firefox anti-tracking, and ad blocker usage have severely degraded pixel-based cross-site tracking. The customer journey you can observe in your analytics is significantly shorter and less complete than the actual journey.

Platform Attribution Wars

Each ad platform's attribution shows that platform as responsible for the majority of conversions — because each uses their own pixels, their own attribution windows, and counts their touchpoints without seeing others'. Google Ads + Facebook Ads + email combined attribution often sums to 200%+ of actual conversions.

Better Approaches

Triangulation

Don't rely on a single attribution model. Use three independent signals and look for consistency:

• Platform-reported attribution (Google Ads, Facebook Ads)
• GA4 attribution (independent measurement layer)
• Revenue-channel correlation (when you spend more in a channel, does revenue increase?)

Where all three agree, you have confidence. Where they disagree, investigate before making budget decisions.

Incrementality Testing

Run holdout experiments where a control group doesn't see your ads. Compare conversion rates between exposed and holdout groups. The difference is true incremental lift from that channel. This is the most reliable measurement of channel effectiveness but requires significant volume and organizational discipline.

Marketing Mix Modeling

Statistical model using historical spend and revenue data to estimate the marginal contribution of each channel. Doesn't require tracking users across channels — works at the aggregate level. More expensive to build but captures offline spend and privacy-safe channels.

Server-Side Tracking

Implement server-side tracking to reduce pixel data loss from ad blockers and browser restrictions. Better first-party data improves the quality of all attribution models you apply on top of it.

The honest truth: there is no perfect attribution solution. The goal is reducing uncertainty, not eliminating it. Our marketing analytics team builds measurement programs that use multiple signals intelligently. Contact us to develop a practical attribution approach for your business.