Last-Click Attribution

Last-click attribution assigns 100% of the credit for a conversion to the last click the customer made before converting — the final touchpoint gets all the glory, every touch before it gets none. It's the oldest and most common ATTRIBUTION model, the historical default of analytics platforms, and it persists because it's simple, deterministic, and universally available. It's also systematically wrong in a specific, predictable way: by crediting only the final touch, it over-rewards the channels that happen to be near the finish line and erases everything that drove the customer there.

Why it's biased and how, because the distortion is predictable: by giving all credit to the last click, last-click systematically over-credits demand-HARVESTING channels (branded search — the customer typing your name was already coming; RETARGETING — shown to people already deciding; the final email or discount code at checkout) and systematically under-credits demand-GENERATING channels (brand campaigns, broad prospecting, the awareness content and upper-funnel touches that created the demand the harvesters later collect) — the exact INCREMENTALITY-VS-ATTRIBUTION bias, now baked into the default model. The downstream damage when budgets follow it: optimizing on last-click drifts spend toward the bottom of the funnel (feeding the channels that take credit, starving the ones that create demand — the attribution doom loop), it makes branded search and retargeting look like heroes (and they get scaled until they're just harvesting demand other channels manufactured), and it under-funds the brand and prospecting that actually fill the top. The cousins and the honest comparison: FIRST-CLICK (the opposite bias, all credit to the first touch — over-credits discovery, ignores closing), linear and time-decay (spreading credit across touches — better, but still correlational rules, not causal truth), DATA-DRIVEN-ATTRIBUTION (algorithmic credit allocation — more sophisticated but still attribution, still answering 'which touches were near conversions' not 'what caused them'), and the real arbiter, INCREMENTALITY testing (which last-click can't approximate). When last-click is actually fine (because it's not always wrong): for genuinely single-touch or short, simple conversion paths (a direct-response click that converts immediately with no journey), for within-channel optimization where its relative comparisons are more trustworthy than its cross-channel credit, and as a cheap directional default when the alternative is no measurement at all — the error is not using last-click, it's using it for cross-channel BUDGET decisions it's structurally incapable of getting right. The practical stance: last-click as a simple operational and directional tool, never as the basis for allocating budget across channels, with multi-touch for a richer (if still correlational) picture and incrementality as the truth that overrules all of them.

Last-click attribution matters as the default most analytics still ships with — so it matters most as a bias to recognize and correct, especially at budget-allocation time, where its over-crediting of branded search and retargeting and under-crediting of brand and prospecting will misdirect spend toward the bottom of the funnel if trusted. It's acceptable for short single-touch paths, within-channel optimization, and as a cheap directional default; it's dangerous as the basis for cross-channel budget decisions. The discipline is knowing the bias (bottom-funnel over-credit, top-funnel erasure), using last-click only where its limits don't bite, layering multi-touch for a fuller correlational view, and letting incrementality testing overrule it for the channel-level decisions that matter — never letting the simplest model win an argument it's structurally unequipped to settle.

Last-click attribution is the historical default of web analytics - the simplest rule the early platforms could implement, crediting whatever touch immediately preceded the conversion - and it persisted long past the point its bottom-funnel bias was understood, becoming the model the incrementality and multi-touch movements exist to correct.

Etymology: source.

Search interest for this term over the last five years:

View interest-over-time on Google Trends →

Areas of marketing where last-click attribution is a core concern: