Almost nobody buys the first time they encounter you. They see a post on a Tuesday, think it is interesting, and do nothing. Three weeks later they see another one, click through, look around, and leave. A month after that they search for you by name and buy. Three posts contributed. One sale exists. Who gets the credit?
That question is what attribution models answer, and the two simplest answers are first-touch and last-touch. Understanding the difference takes about five minutes and will change what you conclude from your own numbers.
First-touch attribution: crediting discovery
First-touch gives all the credit to the earliest interaction you can identify — the post that first brought this person into your orbit. In the example above, the Tuesday post gets the sale, even though the customer bought months later after several other touches.
It answers one question well: what content brings new people in? If your bottleneck is reach, if you are trying to work out which topics or platforms introduce you to strangers, first-touch is the model that tells you. It rewards the top of the funnel and it is the right lens when the top of the funnel is what is broken.
Its weakness is that it systematically over-credits awareness content and under-credits everything that does the actual persuading. Run first-touch for a quarter and you will conclude that your broad, entertaining, widely-shared posts are your best performers, because they are the ones strangers meet you through. You will then make more of them and wonder why revenue does not follow.
Last-touch attribution: crediting the close
Last-touch gives the credit to the final identifiable interaction before the payment. In the example, that is the click three weeks before the purchase, or the post that immediately preceded it.
It answers a different question: what content converts people who already know who I am? If your problem is that you have an audience and not much revenue, last-touch is the model that shows you which posts move people over the line. It rewards specificity, product detail, proof, and offers.
Its weakness is the mirror image of first-touch. It under-credits the content that built the relationship in the first place. Taken literally over a long period, last-touch tells you to publish nothing but product posts, which works for a few months and then quietly runs out of audience, because you stopped doing the thing that brought new people in.
Why last-touch is the right default for a solo founder
Three reasons, and none of them are about theoretical correctness.
- It maps to something you can actually observe. This click, this cookie, this payment. First-touch requires you to have been tracking that person from the very beginning, which you almost certainly were not.
- It fails conservatively. Last-touch under-credits your top-of-funnel content, which is annoying. First-touch over-credits it, which is dangerous, because over-crediting makes you spend more on something that is not working.
- It matches the decision you are making. At small scale you are choosing what to publish next week, not allocating a media budget across channels. “What made someone buy” is the more useful input for that.
There is also a data-volume argument. Multi-touch models — splitting credit evenly, weighting by position, or fitting a statistical model — need a lot of conversions before their output means anything. With twelve sales a month, a model that assigns 0.3 of a sale to four posts each is producing noise with decimal places. Simple and unambiguous beats sophisticated and unstable.
Where first-touch earns its keep
Do not throw it away entirely. There is one situation where first-touch is clearly the better lens: you are trying to decide whether to keep investing in a new platform. A network you joined three months ago will look terrible under last-touch, because nobody there knows you well enough to buy yet. Under first-touch, you can see whether it is introducing you to people at all.
The same applies to top-of-funnel formats generally. If you want to know whether your long-form posts are worth the effort, last-touch will tell you they are not, because people do not buy directly off them. First-touch will tell you whether they are the front door.
If your tooling supports it, the useful move is not to pick one model permanently but to look at both occasionally and note where they disagree. A post that ranks high on first-touch and low on last-touch is a discovery asset. A post that ranks high on last-touch and low on first-touch is a closer. Both are worth having, and knowing which is which tells you what to publish when.
The window matters as much as the model
Attribution models are usually discussed without mentioning the other setting that changes the answer just as much: the attribution window, the length of time a click stays eligible to be credited for a later sale.
A last-touch model with a seven-day window and a last-touch model with a ninety-day window will produce quite different reports from identical data. The long window credits far more sales, catches longer consideration cycles, and flatters everything you published. The short window credits fewer and is much harder to argue with. Pick the model and the window together, then leave both alone long enough to compare months.
Pick one, write it down, stop changing it
The single most common attribution mistake is not choosing the wrong model. It is changing the model, or the window, halfway through and then comparing months that were measured differently. Every number moves, you cannot tell which change was real, and you lose the ability to see a trend.
So choose last-touch, choose a window that matches how long people actually take to buy your product, write both down, and hold them for at least a quarter. Consistency over time is worth more than accuracy in any single month, because what you actually want to know is whether things are getting better.
seenpaid defaults to last-touch for the reasons above: it is unambiguous, it maps directly to an observable click and an observable payment, and when it cannot confidently connect a sale to a click it credits nothing rather than guessing. An honest gap in the data is more useful than a confident number that happens to be wrong.