Prescient vs. last-click: What your attribution model can't see

A row of dominoes falls, one after another, until the last one topples the object at the end of the line. It's tempting to give that last domino all the credit. It's the one that actually made contact, after all. But that domino only fell because dozens of others tipped into each other first, and anyone who only ever watches the last one will never understand why the row fell the way it did, or how to set up the next one to work just as well.

That's last-click attribution in a nutshell. It's not wrong to look at the last domino. It's just not the whole picture, and for a lot of brands, that gap is costing them budget they don't even know they're leaving on the table. The decisions built on top of last-click data (what to cut, what to scale, what "worked") are only as good as what the model can actually see. And for any brand that's grown past pure D2C, what it can see keeps shrinking.

Key takeaways

• Last-click attribution isn't a flawed model so much as a limited one, and plenty of genuinely profitable brands have scaled fast using it as their primary source of truth.
• Last-click can't show you what's driving Amazon purchases when the spend behind them happened somewhere else, like Meta or CTV, because there's no click connecting the two.
• The same is true for physical retail. If your paid media sends someone into a Target instead of onto your site, there's no click event to attribute, so last-click has nothing to work with.
• Consumers reach brands through multiple paths and devices, sometimes clicking, sometimes going direct on a completely different screen later, and last-click only ever captures a fraction of that.
• Halo effects, meaning the revenue a campaign drives indirectly through another channel, can come from any campaign, not just the big awareness plays at the top of your funnel.
• The more a business grows beyond pure D2C into retail or marketplaces, the more last-click's blind spots start showing up in the budget decisions built on top of it.
• Prescient's model works backward from what actually happened (the sale, the visit, the purchase) rather than trying to trace a person's path the way pixel-based tools attempt to.

Why last-click became the default

Before getting into what last-click misses, it's worth saying plainly: there's a good reason so many brands use it. If you're reporting on last-click right now, it's probably because leadership requires it, or because you're at a stage where it's the simplest tool available and it's worked. Last-click comes standard in most ad platforms and analytics tools, it's easy to explain to a room full of stakeholders, and it doesn't require a data science team to set up or maintain.

That simplicity is also exactly why so many brands have built real, profitable businesses on it. Last-click didn't hold those brands back from getting off the ground. It's a legitimate way to get a read on what's working when you're small, scaling, or need a number the whole company can agree on without a lot of explanation.

None of that changes what the model is actually built to see, though, and that's where things start to get more complicated as a brand grows.

What last-click structurally can't show

Last-click was built to answer one specific question: what was the final touchpoint before someone clicked "buy"? It's good at answering that question. The trouble is that a lot of revenue doesn't have a final click attached to it, and those cases aren't rare edge cases so much as entire categories of purchase behavior.

Two show up constantly for brands that have grown past pure D2C:

• Amazon. If your paid spend on Meta, CTV, or anywhere else influences someone to go buy your product on Amazon, last-click has no way to connect those dots. There's no click running from your ad to your Amazon listing, so the sale shows up as if it came from nowhere, or worse, gets credited entirely to Amazon's own advertising.
• Physical retail. The same problem shows up at stores like Target. Your ad might be exactly what got someone to swing by and pick your product off the shelf, but a walk into a store doesn't generate a click. Last-click simply has nothing to attribute that purchase to.

Both of these are versions of the same issue: last-click needs a click to work with, and a growing share of real purchase behavior doesn't produce one.

Why sessions tell a bigger story

Consumers don't move through a single, tidy path from ad to purchase. They reach a brand through multiple channels and multiple devices, and the path they take often has nothing to do with what a last-click report can see.

Someone might see your ad on their phone during a commute, do nothing about it in the moment, then go direct to your site on their laptop that evening to buy. There's no click connecting those two moments in most attribution setups, so the whole interaction shows up as "direct traffic," with no credit going back to the ad that actually started it. Multiply that across a customer base and you start to see why direct clicks only ever capture a fraction of a channel's real effectiveness.

It's also worth clearing up a common misconception here: this kind of spillover, usually called a halo effect, isn't just something that happens with big brand-awareness campaigns at the top of the funnel. A retargeting ad someone scrolls past without clicking can still be the thing that gets them to search for your brand later. Halo effects can come from any campaign, which means last-click is missing this kind of influence across your entire media mixt.

What this costs a growing brand

Put those two problems together (the missing retail and marketplace purchases, and the missing cross-device sessions) and the real cost of last-click starts to come into focus. It keeps a brand's view of performance limited to whatever activity happens to generate a clickable ad unit and a direct click to a website.

For a pure D2C brand with no retail footprint and a simple, single-device customer journey, that ceiling might not matter much. But very few brands stay that way as they scale. The moment a brand adds a retail presence, sells on Amazon, or picks up customers who research on one device and buy on another, last-click starts missing a bigger and bigger share of what's actually driving revenue. And because those budget decisions get made off whatever the model can see, brands in this position often end up cutting or underfunding the exact campaigns doing the most work behind the scenes, simply because that work doesn't show up as a click.

This isn't an argument that last-click needs to be thrown out. It's a case for understanding exactly where its walls are, especially once a brand looks more like an omnichannel business than a single-channel one.

Where Prescient comes in

Prescient's marketing mix model takes a different starting point than last-click. Instead of trying to trace an individual's path from ad to purchase the way pixel-based tools attempt to, it works backward from what's actually observable, like a sale happening on Shopify, on Amazon, or in a retail partner's data, and estimates what combination of marketing activity most likely drove it. That structural difference is what lets it surface Amazon halo effects, retail halo effects, and cross-device sessions that a click-based model was never built to catch in the first place.

None of this is about replacing last-click or telling you it was the wrong call to start there. It's about giving you a fuller picture as your business grows past what a single click can capture. If you want to see what that fuller picture actually looks like for your own brand, book a demo and we'll walk you through it.

FAQs

Is last-click attribution wrong?

Not exactly. Last-click is genuinely good at answering the specific question it was built to answer: what was the final touchpoint before a click-based purchase? Plenty of brands, especially smaller or scaling ones, have built profitable businesses using it as their primary measurement tool. The issue isn't that it gets things wrong, it's that it can only see activity that ends in a click, which leaves out a growing share of how customers actually shop as a brand scales.

Can last-click attribution show Amazon or retail sales?

Generally, no. Last-click needs a click connecting your ad to the purchase in order to attribute credit, and purchases on Amazon or in physical retail stores like Target often don't have that connection. Someone might see your ad and then buy on Amazon later, or walk into a store without ever clicking anything, and last-click has no mechanism to trace either of those paths back to the marketing that drove them.

What's the difference between an Amazon halo effect and a retail halo effect?

They describe the same basic idea (marketing spend driving a purchase that doesn't happen through a click) but in two different contexts. An Amazon halo effect refers to non-Amazon marketing spend, like a Meta or CTV campaign, influencing a purchase on Amazon. A retail halo effect refers to marketing driving a purchase at a physical retail location, like a Target or Walmart store, where there's no click involved at all.

Do I need to switch off last-click attribution to see these effects?

No. Last-click can stay exactly where it is for what it's good at. Getting visibility into Amazon halo effects, retail halo effects, and cross-device sessions requires a different kind of model, one built to work from observable outcomes rather than click trails, but that's additive information rather than a replacement for what you're already tracking.

Why do halo effects matter for campaigns that aren't top-of-funnel?

Because halo effects aren't limited to big awareness campaigns. A retargeting ad or a lower-funnel campaign someone sees without clicking can still lead to a purchase later, whether that's a branded search, a direct visit, or a sale on Amazon or in retail. If a brand only credits halo effects to its awareness spend, it risks undervaluing plenty of other campaigns doing similar work throughout the funnel.