# Why Your Attribution Broke: A Short History of Losing the Signal | Attribution Playbook

Source: https://attributionplaybook.com/blog/why-attribution-broke.html

Sep 15, 2026 - Attribution Playbook


Most ecommerce teams know their attribution is worse than it used to be. Fewer can say precisely why, which layer failed, or what a fix actually restores. The erosion came in three distinct waves, and each needs a different response.

## Wave one: intelligent tracking prevention

Browser privacy features - led by Safari's ITP - capped the lifespan of cookies set by JavaScript. A shopper who clicked your ad on Monday and bought on Saturday suddenly looked like two different visitors, because the cookie that connected them expired mid-week. Effect: returning visitors were reclassified as new, and multi-touch journeys collapsed into single sessions.

## Wave two: ad blockers and tracking protection

Browser extensions and built-in protections began blocking analytics and ad pixels outright. These sessions never appear in your data at all - not misattributed, just absent. Effect: reported traffic and conversions fell below reality, and the missing share skews toward technical, higher-income audiences.

## Wave three: cookie deprecation and mobile identifiers

The long phase-out of third-party cookies and mobile ad IDs removed the connective tissue between platforms. Cross-site retargeting and view-through attribution - always the least reliable numbers - degraded most. Effect: platform-reported ROAS diverged further from anything a finance team can verify.

## Which fix restores which loss

Wave one's fix is first-party, server-set identifiers with durable lifetimes. Wave two's fix is server-side collection - events sent from your own domain and infrastructure rather than third-party scripts the blocker recognizes. Wave three has no full fix; the honest response is first-party data plus modeled or incrementality-based measurement. The common thread: the closer tracking lives to your own infrastructure and the network edge, the more signal survives.

## Frequently asked questions

### Why did attribution get so much worse so quickly?

Several supports fell at once: third party cookies declined, mobile platforms restricted tracking, consent rates cut the observable share, and journeys spread across devices. Each change removed signal the old models quietly depended on.

### Is last click attribution still useful for anything?

As a rough operational signal for bottom funnel activity, cautiously. As a budget setting truth it misleads, because it credits the last touch for demand other channels created and cannot see the touches privacy changes hid.

### What should replace a single attribution model?

A blend: platform data read with scepticism, controlled experiments and holdouts for big questions, and marketing mix thinking for budget level decisions. No single replacement carries the certainty the old dashboards pretended to.

### How should teams talk about attribution uncertainty?

Openly, with ranges and confidence stated. Reports that admit what cannot be observed produce better budget decisions than precise numbers built on a shrinking, biased sample presented as complete.