Attribution is the question "what actually drove this conversion?" — and it's deceptively hard. Customers touch many channels before they buy, and deciding how to give credit across those touches shapes budgets, careers, and the occasional heated meeting. Here's a practitioner's guide to the models, what each gets right and wrong, and how to choose without pretending any of them is perfect.
Why attribution is hard
A real customer journey might involve a blog post found via search, an ad seen weeks later, an email, a webinar, and finally a direct visit to buy. Which one gets the credit? All of them contributed; none of them alone closed the deal. Attribution models are just different rules for splitting that credit — and crucially, every model is a simplification. None of them is truth. The goal is a model that's useful, not one that's correct, because correct isn't on the menu.
Single-touch models
First-touch attribution
All credit goes to the very first interaction. This answers "what creates awareness?" and is useful for evaluating top-of-funnel and demand-creation efforts. Its blind spot: it ignores everything that happened between first touch and the sale, so it over-credits awareness channels and under-credits anything that closes.
Last-touch attribution
All credit goes to the final interaction before conversion. It's the simplest and most common model, and it answers "what closes?" — but it badly under-credits everything that built the relationship earlier. Last-touch makes brand and awareness work look worthless even when they're essential.
Multi-touch models
Linear attribution
Credit is split evenly across every touchpoint. It acknowledges that the whole journey matters, which is more honest than single-touch — but treating a throwaway email open as equal to a demo request is obviously too blunt.
Time-decay attribution
Touchpoints closer to conversion get more credit, earlier ones get less. This is intuitively appealing for businesses with longer sales cycles, where recent interactions plausibly matter more. The weakness: it systematically discounts the awareness touches that started the journey.
Position-based (U-shaped)
Gives extra weight to the first and last touches (often around 40% each) and splits the rest across the middle. The logic: the first touch (got them in) and the converting touch (closed them) are the most important, with the middle nurturing the relationship. It's a reasonable compromise for many B2B funnels.
Data-driven attribution
The most sophisticated approach uses statistical models to assign credit based on what actually correlates with conversion across your real data, rather than a fixed rule. It can be powerful, but it requires significant data volume to be reliable and it's a black box — harder to explain and defend in a meeting. It's a fit for high-volume businesses, overkill for smaller ones.
How to actually choose
Match the model to the question
If you're evaluating awareness investment, lean first-touch. If you're optimising closing tactics, last-touch tells you more. If you want a balanced view of a full B2B journey, a position-based or time-decay model is usually the pragmatic pick. The right model depends on the decision you're trying to make.
Don't marry one model
Sophisticated teams look at multiple models side by side. If a channel looks great under first-touch but invisible under last-touch, that's not contradiction — it's insight about where in the funnel that channel works. Comparing models often reveals more than any single one.
Be honest about the limits
Attribution can't capture offline conversations, word of mouth, or the brand impression that made someone click in the first place. Treat your numbers as directional evidence, not gospel. The marketer who says "this is our best estimate and here's why" is more credible than the one who defends a single number to the death.
The practitioner's takeaway
The goal of attribution isn't perfect accounting — it's better decisions. A simple, well-understood model that your team trusts and acts on beats a sophisticated one nobody believes. Start with something explainable, use multiple lenses, and keep reminding everyone that the map is not the territory. The teams that get this right argue less about credit and spend more time acting on what the models broadly agree on.
Attribution and measurement skills are in constant demand. Browse current MarTech and analytics roles to see how often attribution shows up in the requirements.