Here is a purchase that happened yesterday in your analytics. Someone searched your brand name, clicked a paid search ad, and bought. Google Search gets 100% of the credit.
Here is what probably happened before that: two weeks ago, they saw a display ad while reading an industry newsletter. A few days later, a colleague mentioned your product. Last week they watched a YouTube review. Yesterday morning they searched "your brand name review" and spent 15 minutes on your website. Then they searched your brand again and clicked the paid ad.
In your attribution report, Google Search drove the conversion. Display: zero credit. Organic search: zero credit. YouTube: zero credit. Word of mouth from a colleague: not even in the data.
Last-click attribution is an accounting system that was designed for a simpler internet, when the typical customer journey was short and linear. It was never accurate. We just tolerated it because there wasn't a better option at scale, and the inaccuracy was roughly consistent over time so you could still see trends. That reasoning is now breaking down.
Last-click gives all credit to the final touchpoint, ignoring every channel that built intent
What last-click attribution actually rewards
Last-click systematically rewards channels that appear late in the customer journey, regardless of whether they caused the journey to happen.
Branded paid search is the clearest example. If you run branded search campaigns, ads that show up when someone searches your company name, those ads will almost always appear right before a purchase, because people who are about to buy tend to search for you by name. Last-click attribution will credit those ads heavily.
But ask yourself: is someone who's already decided to buy from you, searching your name with intent to purchase, being caused to convert by your branded search ad? In most cases, they would have found your website organically, or directly, without the ad. The ad isn't creating demand. It's just intercepting existing demand at the moment it surfaces.
Run a holdout test on your branded search campaigns and this becomes apparent fast. Turn off branded search for a matched set of users or in a matched geographic region. Measure what happens to conversions. In the vast majority of cases, the incremental impact is small, most of the "conversions" that attributed to branded search still happen, just through the organic result instead. See why your Google Ads attribution is probably wrong for more on this dynamic.
The retargeting illusion
Retargeting has the same problem, often more severely.
Your retargeting audience consists of people who have visited your website. By definition, that's an audience with above-average purchase intent. Some percentage of them were always going to buy, regardless of whether they saw your retargeting ads. When those people eventually convert, your attribution system says the retargeting ad caused it. But it might have been coincidental, they were going to convert anyway.
This is measurable. Multiple large brands have run holdout tests on their retargeting audiences and found that 70-90% of the people in the "converted" group would have converted without seeing any retargeting ads at all. The attributed ROAS on retargeting looked excellent. The incremental ROAS was a fraction of that.
This doesn't mean retargeting is worthless, that 10-30% incremental lift is real, and for some businesses the economics still work. But it does mean that if you're scaling retargeting spend based on attributed ROAS, you're probably significantly overspending.
What gets hurt
The real cost of last-click attribution isn't that it overcredits the wrong channels. It's that it undercredits the right ones, specifically, upper-funnel channels that do important work early in the customer journey but rarely appear as the last touchpoint. This problem exists across every attribution model, though last-click makes it most severe.
Campaigns designed to drive awareness, YouTube, connected TV, display, podcast, outdoor, generate demand that surfaces days or weeks later as branded searches, direct visits, and word of mouth. Last-click attribution doesn't see any of this. The result: brands consistently underfund awareness and overfund the bottom-of-funnel channels that capture demand but don't create it.
Over time, this hollows out your customer acquisition. You're harvesting the demand that exists while starving the channels that fill the top of the funnel. Conversion rates stay good but volume declines. CAC creeps up. You spend more on retargeting trying to solve a problem that retargeting can't fix.
What to do instead
The answer isn't to switch to a different attribution model. Data-driven attribution, time-decay, linear, these all have the same fundamental limitation: they can only assign credit based on observed touchpoints, and they can't distinguish causation from correlation.
The real answer is to use attribution as one input among several, rather than as the primary measurement system.
For channel budget allocation, use MMM or incrementality testing. These methods can capture channel effects that attribution misses (offline, cookieless channels) and can test causal claims rather than just correlating touchpoints with outcomes.
For day-to-day optimization within channels, attribution is still valuable. If your DDA model shows that a particular creative or audience segment consistently appears in converting paths, that's useful optimization signal, even if the exact credit allocation is imprecise.
The specific change for most teams: stop using attributed ROAS as the primary KPI for budget allocation decisions. Replace it with incrementally tested lift for the channels where you run tests, and MMM-estimated channel efficiency for annual planning. Keep attribution data for creative testing, audience management, and channel sequencing. Use it for what it's actually good at.
Last-click attribution isn't wrong because it's last-click. It's wrong because it assigns 100% credit to one touchpoint in a multi-touchpoint world, and because the touchpoint it credits is systematically biased toward channels that are close to conversion rather than channels that cause conversion. That's a bias that compounds with every budget decision you make based on it.