First-Touch vs Last-Touch Attribution: Which Lies Less

First-touch and last-touch attribution both lie to you. They just lie in opposite directions. Last-touch overcredits the final click before the sale and buries everything that created the demand. First-touch overcredits the discovery moment and ignores everything that actually closed. Pick one and you will systematically over-fund one part of your funnel and starve the other. The fix is not finding the honest model. It is knowing exactly what each one hides and reading them together.

What last-touch attribution hides

Last-touch gives all the credit to the final interaction before conversion. Someone discovers you through a blog post, follows you for weeks, gets nurtured by email, then finally clicks a branded search ad and buys. Last-touch hands 100 percent of the credit to that branded search click and zero to the blog post and email that did the real convincing.

The result is predictable. Bottom-of-funnel channels, branded search, retargeting, look like heroes, because they are always standing closest to the sale. Top-of-funnel channels that create demand look worthless, because they are never the last click. Budget flows to the harvesting channels and away from the demand-creating ones, and eventually the pipeline dries up because nothing is filling the top anymore.

This is why last-touch quietly kills growth. It rewards the channels that capture existing demand and punishes the ones that generate it. I see brands defund the exact content and awareness work that was feeding their whole funnel, then wonder why branded search volume fell. It fell because you stopped creating the demand that fed it.

What first-touch attribution hides

First-touch has the opposite bias. It gives all the credit to the discovery moment, the first time someone found you. That makes awareness channels look great and closing channels look useless.

The problem is that discovery is not conversion. A channel can be excellent at getting attention and terrible at driving sales. If you optimize purely on first-touch, you pour money into top-of-funnel reach that fills the pipeline with people who never buy, while starving the nurture and closing work that turns interest into revenue. You get traffic and no orders.

First-touch is more useful than last-touch for one specific question, which is where demand originates. But used alone to allocate budget, it over-invests in the top and ignores whether any of that attention ever converts.

Neither is truth, and that is the point

Both models take a multi-touch reality and force it onto a single interaction. Real buyers touch you many times, across channels, over days or weeks. Any model that assigns all credit to one touch is discarding most of the story. The question "which is more accurate" is the wrong question, because both are wrong by design.

The right move is to read them together and let the disagreement teach you. When first-touch and last-touch disagree sharply about a channel, that gap is information. A channel that scores high on first-touch and low on last-touch is a demand creator, keep funding it even though it never gets the closing credit. A channel that scores low on first-touch and high on last-touch is a demand harvester, useful, but it needs the top-of-funnel channels to feed it. This is the same blended-view discipline I apply to ecommerce in ecommerce attribution after iOS privacy changes.

The check that beats both models

Attribution models describe correlation. To know what actually drives revenue, you test. Turn a channel off for a slice of your market, hold everything else steady, and watch whether revenue moves. If it drops, the channel was incremental. If it holds, the channel was taking credit for sales you would have gotten anyway.

That incrementality test cuts through the model wars entirely. It does not care about first-touch or last-touch, it measures cause. One good holdout tells you more than a year of arguing about attribution windows. It is the same logic behind judging marketing on the one number that cannot be gamed, which I cover in how to measure marketing agency ROI honestly.

How to actually use this

Do not adopt one model and trust it. Look at first-touch and last-touch side by side, treat the disagreement as a map of your funnel, and confirm the big budget decisions with holdout tests. Fund demand creators even though last-touch ignores them. Keep harvesters, but do not let them starve the top.

That layered read is how I set up measurement for clients at Girard Media, because a single attribution model, whichever one, will always push you to over-fund one half of your funnel. The goal is not a perfect number. It is a clear enough picture to spend confidently, and that comes from reading the models against each other and testing the answer. If your reporting still runs on one model, that is the gap to close with Girard Media.