Ask five analytics tools who deserves credit for a single sale and you’ll get five different answers. That’s not a bug — it’s the whole point of attribution modeling. A customer rarely buys on first contact; they read a blog post, see a retargeting ad, click an email, search your brand, and finally convert. An attribution model is just the rule that decides how much credit each of those touchpoints gets. Pick a different rule and the “winning” channel changes. Understand the rules and you stop arguing about which number is true and start asking which one answers your question.
The short version
- Attribution models assign credit for a conversion across touchpoints
- The big split is single-touch vs multi-touch
- No model is “correct” — each tells a different, useful story
- Lookback window and dark traffic quietly distort every model
- Pick the model that fits the decision you’re trying to make
The core split: single-touch vs multi-touch
Every model lands on one side of a line. Single-touch models hand 100% of the credit to one interaction — either the first or the last. They’re simple and they lie by omission, because they pretend the other touchpoints didn’t happen. Multi-touch models spread credit across several interactions. They’re closer to reality but harder to explain and more sensitive to how clean your tracking is. Most beginners start with single-touch because it’s what the platforms show by default, then graduate to multi-touch once they realize their best channels were hiding in the middle of the journey.
One journey, different credit
Picture a real path: a prospect finds you through an organic blog post, comes back a week later via a paid social ad, opens a nurture email a few days after that, and finally converts by clicking a Google search ad for your brand. Same sale, four touchpoints. Watch how the credit moves:
- First-touch: all credit to the organic blog post.
- Last-touch: all credit to the brand search ad.
- Linear: 25% to each of the four.
- Time-decay: most credit to the email and search ad (closest to the sale).
- Position-based (U-shaped): 40% to the blog, 40% to the search ad, 20% split across the middle.
Nothing about the customer changed. Only the rule did. That’s why two teams can look at the same data and reach opposite conclusions about which channel to fund.
How each model hands out credit
- Single-touchFirst-touch rewards discovery; last-touch rewards the closer. One winner, zero nuance.
- Weighted multi-touchLinear, time-decay, and U-shaped spread credit by position or recency across the path.
- Data-drivenAn algorithm assigns credit based on which touchpoints actually move conversions in your data.
Each model, briefly
First-touch answers “what makes people aware of us?” — ideal for judging top-of-funnel discovery. Last-touch answers “what closes the deal?” — useful but it over-credits brand search and retargeting, the channels that always sit near the finish line. Linear treats every touch as equal, which is fair and also a little naive. Time-decay leans toward recent touches, good for short sales cycles. Position-based (U-shaped) rewards the first and last touch most, which matches how a lot of B2C journeys actually feel. W-shaped adds a third weighted point (lead creation) for longer B2B paths. And data-driven attribution (DDA) uses your own conversion data to assign credit algorithmically — powerful, but a black box, and only as trustworthy as the data feeding it.
No attribution model is “true.” Each is a lens. The skill is choosing the lens that fits the decision in front of you.
The settings that quietly distort everything
Two things warp attribution results no matter which model you pick, and almost nobody talks about them.
The first is the lookback window — how far back a tool looks for touchpoints before a conversion. A 7-day window and a 90-day window will credit completely different channels for the same sale, especially if your sales cycle is long. Set it to match how people actually buy from you, not the default.
The second is dark traffic — the word-of-mouth, the Slack DM, the podcast mention, the “direct” visits that are really untracked referrals. None of it shows up as a clean touchpoint, so it silently inflates whatever channel gets the last click. If “direct” is one of your biggest “sources,” that’s usually a measurement gap, not a channel.
Reading attribution well: do this, not that
✓ Do this
- Match the lookback window to your real sales cycle
- Pick the model for the decision, not the vendor’s default
- Compare two models side by side to see the bias
- Sanity-check with holdout tests when stakes are high
✕ Avoid this
- Trusting last-touch as the full story
- Treating “direct” traffic as a real channel
- Switching models to make a channel look good
- Believing data-driven attribution because it’s automatic
How to choose a model
Start from the decision. If you’re trying to understand what fills the top of your funnel, lean first-touch. If you’re optimizing ad spend for immediate conversions, last-touch or time-decay will tell you what’s closing. If you want a balanced view for budget allocation across a multi-step journey, a position-based or data-driven model serves better. The mistake is picking one model and treating it as gospel. Mature teams keep two or three in view and read them against each other — the gap between them is often the most honest signal you’ll get.
Beyond attribution: incrementality and MMM
Attribution answers “who gets credit for conversions we tracked.” It can’t tell you whether a channel caused incremental sales or just took credit for sales that would’ve happened anyway. For that you need incrementality testing (turn a channel off in a region and watch what happens) or marketing mix modeling (MMM), a top-down statistical approach that estimates each channel’s contribution without relying on cookies or click paths. Attribution is your day-to-day lens; incrementality and MMM are the reality checks you run a few times a year.
Frequently asked questions
What is a marketing attribution model?
It’s the rule that decides how much credit each touchpoint in a customer’s journey gets for a conversion. Because a buyer usually interacts with several channels before purchasing, the model determines which channels look responsible for the sale — and different models produce different answers.
What’s the difference between single-touch and multi-touch?
Single-touch models give 100% of the credit to one interaction — either the first or the last. Multi-touch models spread credit across several touchpoints. Single-touch is simpler but ignores most of the journey; multi-touch is more realistic but needs cleaner tracking.
Which attribution model is best?
There isn’t a universally best model — each answers a different question. Pick the one that fits the decision you’re making: first-touch for understanding discovery, last-touch for optimizing immediate conversions, and a multi-touch or data-driven model for balanced budget allocation.
What is a lookback window?
It’s how far back a tool looks for touchpoints before a conversion. A short window and a long one will credit completely different channels for the same sale, so set it to match how long people actually take to buy from you.
Where this fits
Attribution is one piece of a larger analytics discipline. It feeds directly into how you measure marketing ROI — because ROI is only as honest as the revenue you attribute to each channel. It also underpins the idea of choosing one trustworthy number per channel, so your team stops gaming whichever metric a single model happens to reward. And none of it works without clean data collection, which starts with a solid GA4 setup.
If you want to pressure-test your attribution thinking against people running real budgets, that’s the kind of conversation that happens at SEO Spring Training — five days every April in Chandler, Arizona, where trainers walk through actual accounts and actual numbers. No theory, just the plays that work. When you’re ready to stop guessing which channel deserves the credit, come learn it in the room.