How Attribution Models Actually Work, Without the Vendor Slideware

The rule-based models, and what each one hides

• Last-click credits the final touchpoint before conversion entirely. It's simple and matches most ad platforms' own default reporting, but it systematically overvalues bottom-of-funnel channels (branded search, retargeting) that tend to be the last thing a visitor clicks regardless of what actually persuaded them earlier.
• First-click credits whatever introduced the visitor. It surfaces which channels are good at discovery, but ignores everything that closed the sale, and can overvalue broad-awareness channels that rarely convert on their own.
• Linear splits credit evenly across every touchpoint. It's a fair-seeming default precisely because it makes no claim about which touch mattered more - which is also its weakness: a touch that did nothing gets the same credit as one that clearly drove the decision.
• Time-decay weights touches closer to the conversion more heavily, on the assumption that recency correlates with influence. It's a reasonable middle ground for longer sales cycles, but the decay rate is itself an assumption, not a measured constant.
• Position-based (U-shaped) assigns a fixed share (commonly 40/20/40) to the first touch, the last touch, and everything in between. It's an attempt to credit both discovery and closing explicitly, at the cost of an arbitrary split ratio.

Data-driven attribution: a different kind of model, not just another rule

Rather than applying a fixed percentage rule to every conversion path, a data-driven model looks at your actual historical conversions and non-conversions, and estimates each channel's real incremental contribution based on patterns in that data - comparing paths that included a given touchpoint against similar paths that didn't. It requires enough conversion volume to find a reliable pattern, which is exactly why platforms increasingly default to it for larger advertisers and fall back to a rule-based model for smaller ones.

Why every platform still disagrees with every other platform

Even with a chosen model, cross-channel comparison breaks down when each ad platform can only see its own touchpoints - Google Ads has no visibility into a Meta impression, and vice versa - so each platform's "attributed" conversions are calculated against a partial view of the actual path, typically defaulting to a self-favoring last-click, own-platform-only model unless you deliberately point every platform at one shared, cross-channel measurement layer instead.

Picking a model on purpose

The practical question isn't which model is universally right - it's which one matches the decision in front of you. Comparing channel performance for a budget conversation calls for a consistent, cross-channel model applied the same way to every channel (see our note on knowing which ad budget actually works). Justifying total marketing spend to finance calls for a defensible, documented methodology more than a specific model choice (see how to measure marketing ROI you can defend).