---
title: How last-click attribution biases channel results
slug: last-click-attribution-biases
summary: Last-click flatters low-funnel, intent-capturing channels (search, Apple Ads, Google UAC standard) and under-credits view-based ones (YouTube, CTV, chatbot ads); incrementality tests have reversed channel rankings — and platform-reported numbers are biased toward the platform.
domain: marketing
tags: [measurement, attribution, incrementality, apple-ads, google-uac, ctv]
related: [meta-incremental-attribution-model, meta-audience-segments-and-remarketing, measurement-stack-attribution-incrementality-mmm, when-last-click-is-good-enough, incrementality-test-design, apple-ads-as-first-paid-channel, apple-ads-attribution-without-mmp]
sources: [yt:qZj6_mJGOMg, yt:tJzzjP29f-k, yt:iD_YbEW0Y3w, yt:Xx_hBs9Re8s, yt:tcQ_1mm_bHU]
updated: 2026-10-24
---

# How last-click attribution biases channel results

## TL;DR
- Last-click is "clicky". It credits lower-funnel channels and misses view-based upper-funnel ones: YouTube, CTV, ChatGPT ads.
- Haus example: **Google UAC standard** looks efficient and **UAC video** looks terrible on last-click. Incrementality tests show the reverse.
- **Apple Search Ads** looks amazing on last-click in the MMP but tends to be less incremental. Budgets are shifting out of low-funnel search.
- Meta and Google can deliver mostly to people who would have converted anyway, so their platform ROAS overstates the lift.
- Read every platform report by asking "how is this biased, and why?" Then apply a test-derived adjustment factor rather than throwing last-click away.

## Details
**Direction of the bias.** [[people/olivia-kory]]: attribution picks up lower-funnel channels and gives you signal there, but it misses upper-funnel and view-based impact.[^1] Her clearest case is Google App Campaigns: the standard campaign looks extremely efficient, while the video campaign, delivered mainly on YouTube, looks horribly inefficient. "You run an incrementality test and it's actually the reverse."[^2] Apple Search Ads is her example of a channel that "looks amazing on a last click basis in the MMP" but tends to be less incremental, and she sees budgets moving accordingly.[^3]

**Harvesting existing demand.** There is a widespread belief, which Kory thinks has "some truth" to it, that Meta's and Google's algorithms are so good at finding intent that they keep delivering to existing customers and people who would have come anyway.[^4] Haus customers' number one goal is reaching new audiences. They pursue it in two ways: signal engineering (optimising Meta/Google toward shallower events) and expanding into video channels.[^4]

**Platform self-reporting.** [[people/michael-kaminsky]]: "if you log into Meta's ad platform you should always be asking yourself… how is this biased?" Platforms are incentivised to make their ads look good, and their attribution and "incrementality-driven" optimisation code is closed.[^5] [[people/thomas-petit]] adds that networks now *model* a lot of conversions, which pushed people to question last-click in the first place.[^6]

**Other blind spots**
- **Interaction effects.** CTV may work only in combination with other channels. Many MMMs deliberately don't model interactions, so they miss this too.[^7]
- **Claiming credit.** Seufert cites a cookie-stuffing startup that took credit for conversions it never drove. Last-click can't catch this; incrementality can.[^8]
- **False precision.** Last-click tools trained marketers to expect one exact ROAS number. That certainty "was never justified".[^9]

**Inside Meta: view-through and remarketing inflation.** Meta's standard attribution (7-day click, 1-day engaged-view, 1-day view) credits conversions whether or not the ad caused them; Jon Loomer wants a heavy concentration of click-through (ideally 1-day click) and warns that a high share of view/engage-through usually means you're remarketing — results that are "misleading, often inflated, and lack incrementality".[^13] Meta's own incremental attribution model is one in-platform corrective: [[meta-incremental-attribution-model]].[^11][^12]

**What to do.** Keep last-click for daily, ad-level operation. Run periodic incrementality tests per major channel, then turn the gap into a correction factor ("incrementality-adjusted attribution").[^10]

## Disagreements / caveats
- [[apple-ads-as-first-paid-channel]] (App Masters) recommends Apple Ads as the first paid channel. Kory's point doesn't contradict starting there: the claim is that ASA's last-click ROAS overstates its *incremental* value once you have organic demand. Check with a holdout before scaling ASA on brand or own-category keywords.
- The UAC reversal is Kory's generalisation from Haus clients, not a published benchmark.

[^1]: yt:qZj6_mJGOMg @ 22:17 — Olivia Kory
[^2]: yt:qZj6_mJGOMg @ 24:49 — Olivia Kory
[^3]: yt:qZj6_mJGOMg @ 25:34 — Olivia Kory
[^4]: yt:qZj6_mJGOMg @ 26:22 — Olivia Kory
[^5]: yt:tJzzjP29f-k @ 02:39 — Michael Kaminsky
[^6]: yt:iD_YbEW0Y3w @ 02:25 — Thomas Petit
[^7]: yt:qZj6_mJGOMg @ 29:37 — Eric Seufert
[^8]: yt:iD_YbEW0Y3w @ 09:03 — Eric Seufert
[^9]: yt:tJzzjP29f-k @ 46:40 — Eric Seufert
[^10]: yt:qZj6_mJGOMg @ 23:05 — Olivia Kory
[^11]: yt:Xx_hBs9Re8s @ 01:38 — Jon Loomer
[^12]: yt:tcQ_1mm_bHU @ 03:13 — Jon Loomer
[^13]: yt:tcQ_1mm_bHU @ 02:24 — Jon Loomer