Multi-touch vs last-click: what each model hides from you
Every marketing meeting has the same ghost in the room: the channel that gets credit doesn't match the channel that did the work. Last-click says paid search closed the sale. Multi-touch says five channels touched it and splits the credit evenly, or by some decay curve nobody can explain in a sentence. Both are wrong in specific, predictable ways. The problem is nobody tears down how they're wrong before building a dashboard on top of them.
If you run ecommerce or DTC ops, you've probably had a CMO ask "which model is right?" The honest answer is neither — they hide different parts of the customer journey, and if you don't know which part, you'll misallocate budget with total confidence.
What last-click actually hides
Last-click attribution gives 100% of the credit to the final touchpoint before conversion. It's simple, it's fast to compute, and it's baked into Google Analytics and most ad platform dashboards by default. That simplicity is exactly why it lies to you.
- It hides upper-funnel work. A customer sees a Facebook ad on Monday, reads a blog post on Wednesday, gets retargeted on Instagram Friday, then googles the brand name and clicks a paid search ad on Saturday. Last-click hands the entire sale to paid search. Facebook and the blog post get nothing, even though they built the awareness that made the branded search happen.
- It hides retargeting's real job. Retargeting ads convert warm traffic, not cold traffic. When last-click credits retargeting for a sale, it's really crediting whatever channel put that person in the funnel originally. Scale retargeting spend based on last-click numbers and you're often just paying to re-capture demand you already created elsewhere.
- It hides branded search inflation. Branded search looks like your best-performing channel because it catches everyone who already decided to buy. That's not acquisition, that's harvesting. Last-click can't tell the difference between a channel that creates demand and one that just closes it.
- It hides cross-device and cross-session gaps. If someone sees your ad on mobile and buys on desktop three days later, standard last-click tracking frequently can't stitch that together at all. The conversion shows up as "direct" or "organic," and the ad spend that actually drove it disappears from the report entirely.
The net effect: last-click systematically overvalues bottom-funnel, high-intent channels and undervalues everything that builds awareness or consideration. Teams that optimize purely on last-click data tend to starve top-of-funnel spend over time, then wonder why total revenue growth stalls even as "efficiency" metrics look great.
What multi-touch hides instead
Multi-touch attribution (MTA) was built to fix the last-click problem by spreading credit across every touchpoint in the journey. Linear, time-decay, U-shaped, W-shaped, algorithmic — pick your model. The pitch is "fairer" attribution. The reality is a different set of blind spots, and they're harder to spot because the output looks more sophisticated.
- It hides the data collection gap. MTA models are only as good as the touchpoints they can see. Email opens, SMS clicks, organic social views, word-of-mouth, podcast ads, affiliate links with no tracking parameter — if it's not captured in your tag manager or CDP, it doesn't exist in the model. You end up with a precise-looking percentage breakdown built on an incomplete dataset, and precision on incomplete data is more dangerous than honest uncertainty.
- It hides the identity resolution problem. Multi-touch requires stitching together a single person across devices, browsers, and sessions. Since iOS 14.5 and the broader cookie deprecation push, that stitching relies more on probabilistic matching and first-party data quality than deterministic IDs. If your identity resolution is shaky, your "40% credit to paid social" figure is really "40% credit to whatever our matching algorithm guessed was the same person."
- It hides the arbitrary weighting decision. Someone picked the attribution window (7-day click, 30-day view, 90-day lookback) and someone picked the decay curve. Those choices aren't neutral. A 30-day window dramatically inflates upper-funnel channels compared to a 7-day window. Change the window, change the budget allocation, with zero change in actual customer behavior. Most teams never audit which window their platform defaults to or why.
- It hides correlation dressed as causation. MTA tells you which channels were present in a path to purchase. It does not tell you whether removing that channel would have changed the outcome. A channel can show up in 60% of converting paths and still be contributing nothing incremental, because those customers would have bought anyway.
Multi-touch feels more rigorous because the math is more complex. But complexity isn't accuracy. A model built on patchy data and an arbitrary decay curve is still a guess, just a guess with more decimal places.
The pipeline problem underneath both
Here's the part that doesn't get discussed enough in marketing meetings: both models depend entirely on data pipeline quality, and most teams have never audited theirs. Before arguing about which attribution model is "right," check these:
- Is UTM tagging consistent across every campaign, every channel, every agency touching your ads? One inconsistent naming convention and your model is quietly splitting credit that should've been merged.
- Are server-side events deduplicated against client-side pixel fires? Double-counted conversions inflate whichever channel fires first or last, depending on your setup.
- Does your order management or ERP system pass clean conversion data back to your ad platforms and analytics tools, or is there a lag that causes mismatched attribution windows?
- Are returns and refunds backed out of attributed revenue? A channel that drives high-return products will look more profitable than it is if refunds never flow back through the attribution layer.
Fix the pipeline before fixing the model. A last-click report built on clean, deduplicated, properly tagged data will give you better decisions than a multi-touch model built on broken tracking.
Incrementality is the actual answer
The honest fix isn't picking a better attribution model. It's running incrementality tests that measure what happens when you actually turn spend off. Geo holdouts, matched-market tests, and platform-native conversion lift studies answer the question both attribution models dodge: would this sale have happened anyway?
This doesn't mean throw out attribution entirely. Attribution is still useful for day-to-day optimization and catching obvious problems fast. But it should be treated as a directional signal, not gospel. Run quarterly holdout tests on your top two or three channels. Compare what the test says about incremental lift to what your attribution model says about credit share. The gap between those two numbers tells you exactly how much your model is lying, and in which direction.
The real takeaway
Last-click hides the upper funnel. Multi-touch hides its own assumptions behind a more complicated-looking number. Neither one tells you what would've happened if you'd spent the money differently. Treat both as approximations, fix your tracking pipeline before arguing about models, and run real incrementality tests on the channels where the budget actually matters. The team that knows exactly what their attribution model is hiding will always out-decide the team that just trusts the dashboard.