--- title: Meta's incremental attribution model — should you switch from standard? slug: meta-incremental-attribution-model summary: What Meta's ad-set-level "incremental" attribution model does (predicts which conversions the ad caused, and optimises for them) versus standard 7-day click / 1-day view / 1-day engaged-view, why results look worse after switching, and who should use it — Jon Loomer says most advertisers, Ben Heath says only larger advertisers with a customer base. domain: marketing tags: [measurement, attribution, incrementality, meta-ads, reporting] related: [meta-audience-segments-and-remarketing, last-click-attribution-biases, meta-weekly-cadence-and-attribution, meta-budget-scaling-methods, meta-creative-fatigue-diagnosis, measurement-stack-attribution-incrementality-mmm, meta-bid-strategy-min-roas-vs-cost-cap] sources: [yt:Xx_hBs9Re8s, yt:6TRy9eUAPVg, yt:du6rDgRf06g, yt:Ha0gOItSZaU, yt:tcQ_1mm_bHU, yt:dJlkNgpXhVQ, yt:DBvtLzi_j7o, yt:C0R5YJiWpig] updated: 2026-10-24 --- # Meta's incremental attribution model ## TL;DR - **Standard** (the default): 7-day click, 1-day engaged-view and 1-day view. Meta credits any conversion in those windows, whether or not the ad caused it.[^1] - **Incremental** (an ad-set setting under *Attribution model*): Meta models whether a conversion was *caused* by the ad. Announced Aug 2024, it now works for **optimisation as well as reporting**: Meta can't pad results with view-through conversions.[^2][^3][^4] - Expect **reported results to drop** after switching. Most of the drop is perception: view-through and engaged-through conversions still happen, they just aren't counted.[^5][^6] - In Loomer's accounts incremental results roughly match standard's **click-through** results. If most of your conversions are clicks, little changes.[^7] - **Who should use it.** Jon Loomer: most advertisers, full-time, because most never dig into what standard results mean. Ben Heath: not beginners or businesses with a small customer base; yes for larger advertisers with a big warm base. At ~$100K/month he'd consider switching and running proper incrementality tests.[^8][^9][^19] - Either way, before scaling or pausing anything, check how many results Meta considers incremental. Add the column via **Compare attribution settings**.[^10] ## Details **Why standard misleads.** An impression is served, and the person converts within a day through another channel, such as an email. They may never have looked at the ad. That conversion is credited anyway. Loomer calls it remarketing in disguise. If you isolate remarketing audiences, the inflation gets worse. A high share of view-through and engaged-through conversions is a strong sign that many results would have happened without the ad.[^1][^11] **What it fixes.** Ben Heath: incremental attribution shows what you actually get for the spend, and it also *trains* Meta to find incremental buyers. Otherwise a "huge proportion" of the budget can go to warm audiences who buy regularly anyway.[^3] Loomer's list of bad decisions made on standard numbers: - scaling an ad set propped up by view-through conversions, then being surprised when it fails - turning off an ad because it loses to one that view-through inflates[^12] **Why it can understate.** Ben: incremental only counts what happens **inside the attribution window**. Considered or expensive purchases often convert later, so the model is "pessimistic" for long sales cycles. Advertisers who switch, see worse numbers and cut budget can hurt themselves.[^13] **If you stay on standard, read the layers** (Loomer): - break results down by attribution setting; you want a heavy concentration of click-through, ideally 1-day click - use **first conversions** vs all conversions; 100 conversions may come from 80 customers - add 28-day click to see down-funnel conversions - add the incremental column alongside[^14][^15] For scaling decisions specifically, judge only 1-day click or incremental results. The other windows tend to depend on other channels, which means remarketing, and remarketing doesn't scale.[^16] **Agency practice.** CTC sets its on-platform target from **7-day click**, adjusted by an incrementality factor measured for the channel (see [[meta-bid-strategy-min-roas-vs-cost-cap]]).[^17] ## Disagreements / caveats - **Default for everyone vs experienced only.** Loomer believes Meta should make incremental the default and admits he doesn't always use it himself, because he reads the breakdowns. Ben recommends against it for beginners and small customer bases.[^8][^9] - **Low-volume accounts.** Loomer's volume advice: keep standard or use incremental, but don't narrow to 1-day or 7-day click only, because you need the volume.[^18] - Incremental is still Meta modelling its own impact. Cross-check it with lift or geo tests, and with MMM at scale: [[measurement-stack-attribution-incrementality-mmm]], [[incrementality-test-design]]. ## See also - [[meta-audience-segments-and-remarketing]]: the audience-side view of the same inflation - [[last-click-attribution-biases]] - [[meta-weekly-cadence-and-attribution]] - People: [[people/jon-loomer]], [[people/ben-heath]], [[people/tony-chopp]] [^1]: yt:Xx_hBs9Re8s @ 00:47 — Jon Loomer [^2]: yt:Xx_hBs9Re8s @ 01:38 — Jon Loomer [^3]: yt:6TRy9eUAPVg @ 23:01 — Ben Heath [^4]: yt:Xx_hBs9Re8s @ 02:30 — Jon Loomer [^5]: yt:Xx_hBs9Re8s @ 06:33 — Jon Loomer [^6]: yt:6TRy9eUAPVg @ 23:01 — Ben Heath [^7]: yt:Xx_hBs9Re8s @ 05:46 — Jon Loomer [^8]: yt:Xx_hBs9Re8s @ 07:20 — Jon Loomer [^9]: yt:6TRy9eUAPVg @ 23:01 — Ben Heath [^10]: yt:Xx_hBs9Re8s @ 08:13 — Jon Loomer [^11]: yt:Xx_hBs9Re8s @ 01:38 — Jon Loomer [^12]: yt:Xx_hBs9Re8s @ 04:53 — Jon Loomer [^13]: yt:6TRy9eUAPVg @ 24:35 — Ben Heath [^14]: yt:tcQ_1mm_bHU @ 03:13 — Jon Loomer [^15]: yt:Ha0gOItSZaU @ 06:22 — Jon Loomer [^16]: yt:dJlkNgpXhVQ @ 02:22 — Jon Loomer [^17]: yt:DBvtLzi_j7o @ 14:55 — Tony Chopp [^18]: yt:C0R5YJiWpig @ 04:10 — Jon Loomer [^19]: yt:du6rDgRf06g @ 14:46 — Ben Heath