Last-click vs Multi-touch

The real difference

Last-click hands the entire conversion to whatever the customer touched immediately before buying. Simple, stable, and easy for everyone to agree on. It is also structurally biased towards channels that sit at the bottom of the funnel, which is why brand search always looks brilliant and why a YouTube campaign that actually caused the demand shows almost nothing.

Multi-touch spreads credit across touchpoints. Linear, time decay, position based, or a data-driven model that assigns weights from observed paths.

The catch is that a multi-touch model can only see what it can track. Cookie restrictions, app-to-web gaps, consent choices, people researching on a phone and buying on a laptop. Half the journey is invisible, so the model distributes credit across the visible half with total confidence. Precise output, incomplete input.

Privacy changes have made this harder in ways that will not reverse. Browser restrictions, app tracking permissions and regional consent requirements all strip signal out of the journey, and every model built on that signal inherits the gaps without ever announcing them in the interface. Assume undercounting. Then design decisions that survive it.

Where last-click still works

Small accounts. Short buying cycles. Single-channel spend. If you run paid search only, and buyers convert within a day or two, last-click is close enough and arguing about models is a distraction from writing better ads.

It also works as a common language. Everyone understands it, it does not shift under you when a platform updates its model, and it is consistent month to month, which makes trend reading reliable. Use it for tactical decisions: which keyword, which product, which city, which creative to cut. Just stop treating it as a verdict on whether a channel deserves to exist, because on that question it is reliably wrong in one direction.

It is also the model your finance team can reconcile. Orders, revenue, source. No weighting arguments, no fractional credit, no meeting spent explaining why a channel shows two-point-four conversions. That simplicity has real organisational value and is worth keeping even after you add better measurement alongside it.

Where multi-touch adds something

Long consideration cycles and several channels running at once. Real estate is the obvious case. A buyer sees a hoarding, then a Meta ad, then reads a locality guide, then searches the project name six weeks later and fills a form. Last-click credits brand search for all of it, which tells you to cut the very channels that started the process.

Multi-touch at least shows those earlier touches exist. Use it for budget allocation across channels and for understanding path length, not for judging individual keywords. And read it as directional. If the model says display assisted a fifth of conversions, the useful conclusion is that display assists, not that the number is a fifth.

Longer funnels also make view-through and assisted paths worth examining, provided you read them as evidence of contribution rather than as a settled figure. A hospitality brand where guests research for three weeks before booking, or a developer whose buyers take four months, will always look strange under a last-click report. That strangeness is the point.

What should actually settle arguments

Run last-click as your operating dashboard, keep a multi-touch view for quarterly allocation, and resolve genuine disagreements with a holdout. Switch a channel off in two comparable cities for three or four weeks and watch total revenue, not platform-reported conversions. If nothing moves, the channel was taking credit rather than creating demand.

That test costs a little revenue and settles arguments no model can. Pair it with one blended number, total spend divided by total new customers, tracked weekly. Blended cost is harder to game than any attribution window and it is the number a founder should be watching anyway.

Keep one caution in mind. Attribution debates absorb enormous amounts of senior time and rarely change the decision that follows. Get the measurement good enough to avoid obvious mistakes, then spend the remaining energy on offer, creative and landing page, which move results far more than any model ever will.

Set the reporting rhythm and stop relitigating it. Weekly, read last-click plus blended cost. Monthly, read channel mix. Quarterly, run one holdout on whichever channel somebody in the room insists is either wasted or badly undervalued, because that argument returns every quarter in every marketing team and the only thing that has ever settled it is a period where the channel was switched off and the revenue line either moved or it did not. Then act on what you saw.