Last-Click Attribution

Last-click attribution assigns 100% of the credit for a conversion to the final recorded touchpoint before the purchase. Everything earlier in the journey gets nothing.

Why it is still so widely used

It is unambiguous, cheap to compute, and stable — the same order always produces the same answer, and there is no model to explain or defend. That reliability is genuinely valuable, and it is why last-click remains the default in most reporting even where better options exist.

The bias it creates

Last-click does not reward the channels that create demand; it rewards the ones present when demand converts. In practice that means:

• Branded search looks excellent — it is usually the last click, and the customer was often coming anyway.
• Retargeting looks excellent — it is engineered to be last.
• Prospecting looks weak — it starts journeys it rarely finishes.
• Untracked and offline influence look non-existent — they cannot be a last click at all.

Optimising budget on last-click therefore shifts spend toward the bottom of the funnel until growth stalls, because the channels being defunded were the ones generating the demand the others were harvesting.

When last-click is the right choice

For short, single-session purchases with little consideration, last-click is close to accurate and the added complexity of a multi-touch model buys nothing. It is also a reasonable operational metric for in-platform bidding, where consistency matters more than truth.

What to pair it with

The fix is not usually a different attribution model — linear and data-driven models share last-click's core limitation of only seeing recorded touchpoints. The fix is evidence from outside the click stream: incrementality testing to establish what each channel actually caused, and marketing mix modeling to measure the whole mix. Causal attribution then uses that evidence to weight the click-based numbers rather than replacing them.