The Attribution Trap: Why Optimising for Last-Click Is Making Your Growth Worse
A founder called me last month with a plan. “Google Search shows a much higher return on ad spend than Meta. Let’s move the whole budget into Search.”
I asked him one question. “Do we know if Search also brought in the buyers? Or did it just close the ones Meta warmed up?”
He paused. “Maybe? I’d have to check.”
He could not check. His dashboard did not measure that. It measured which ad someone clicked last before buying. He was about to defund the channel creating demand to feed the channel capturing it.
This is the attribution trap. Last-click attribution does not measure marketing effectiveness. It measures which touchpoint happened to come last. Those are different things. Confusing them quietly wrecks your growth.
This article argues one point. Your attribution model is not a reporting choice. It is a budget-allocation weapon. Pick the wrong one, and it will starve the channels that actually build your business.
1. Last-click credits the closer, not the creator
Google’s own definition is blunt. Last-click gives 100% of the credit to the final non-direct touchpoint before conversion. Everything before that click gets nothing.
Think about what that means. A buyer sees your Meta ad, reads a review, searches your brand name, clicks the Search ad, and buys. Last-click hands all the credit to Search. Meta gets zero. The review gets zero.
Search did not create that customer. It caught one already on the way in. Google now recommends data-driven attribution instead, because it assigns credit using actual touchpoint data.
It works like a personal record in the gym. The final rep proves you finished. The training block built the strength. Last-click credits only the final rep.
💡 Key Takeaway: Last-click measures the order of clicks, not the source of demand.
2. Your “best” channel is often just harvesting
Brand-keyword search is the classic offender. Someone already wants you. They type your name. Your ad sits above the free link. They click the ad. Last-click calls that a win.
eBay tested this properly. In controlled experiments, brand-keyword ads showed no measurable short-term benefit, and paid search returns were a fraction of the usual non-experimental estimates.
Read that again. The channel with the cleanest ROAS was mostly paying for clicks eBay would have won for free. Last-click could not tell demand capture from demand creation. Neither can your dashboard.
3. Moving budget to the winner defunds the funnel
Here is the mechanism that hurts you. Reallocation.
When you shift budget to the high-ROAS channel, you pull it from the channels feeding that channel. Fewer people discover you. Fewer warm leads reach the bottom. The harvesting channel keeps posting strong numbers, right up until the funnel above it runs dry.
- The top of funnel loses budget first
- Demand creation slows on a lag, not immediately
- The closing channel looks healthy for weeks
- Then total sales fall, and no single report shows why
By the time the drop appears, you have already cut the thing that caused it.
4. A clean dashboard is not sound logic
The trap survives because it looks trustworthy. One number per channel. Easy to sort. Easy to defend in a meeting.
Marketers already doubt it. Only 21.5% of marketers trust last-click as an accurate measure of long-term business impact. The confidence is low, yet the budgets still move on it.
The independent evidence is worse. WARC found that 35% of last-click-attributed spend generates zero incremental sales. A third of the spend the model rewards changes nothing.
Even the giants act on this. Procter & Gamble cut about US$200 million in digital ad spend in 2017 over viewability and audience-quality concerns. A company with the best tools decided its reported efficiency was not real impact.
💡 Key Takeaway: A clean report can hide broken logic. Simplicity is not accuracy.
5. Triangulate before you reallocate
The fix is not a better single model. It is more than one lens.
The strongest teams combine three:
- Attribution for direction
- Marketing mix modelling for the full picture
- Incrementality tests for cause
Yet only 44% of senior marketing analytics professionals use all three together. Most still rule with one number.
Airbnb ran the harder test. It cut performance marketing and leaned into brand, and Q1 2021 sales and marketing spend fell 28% year on year while the business held. Much of that spend was not creating the demand it claimed.
You do not need Airbnb’s budget. You need one incrementality test before your next big reallocation.
Final Thoughts: Your attribution model is a budget weapon, not a report
“Last-click is simple, cheap, and good enough.” I hear this often. Simple is fine. Wrong is not.
The cost of last-click is invisible. You never see the customers you stopped creating. You only see a tidy report that keeps pointing at the closer.
You do not need a measurement science project. You need one habit. Before you move budget on an ROAS number, ask whether that channel created the demand or just caught it. Run one holdout test. Cut one channel for two weeks and watch total sales, not the dashboard.
The founder who wanted to pour everything into Search did not have a Search problem. He had a measurement problem that looked like a Search win.
If your best channel keeps winning while growth stalls, your attribution model may be lying to you. Book a call, or connect with me on LinkedIn. Let’s find where your budget is funding the wrong end of the funnel.
A note before you close this tab. The fact that you read this far tells me something. You already sense that the way you’ve been thinking about growth might be incomplete. That instinct is worth following.
Mervyn Chua is a growth-transformation consultant helping founders and CEOs build the strategic clarity and systems to grow in an AI-first world. If this raises questions worth exploring for your brand, let’s talk.
