Attribution Modeling

Attribution modeling is the set of rules or the algorithm that decides which marketing touchpoints get credit for a conversion, and how much credit each one gets. Take the same order, change the model, and the channel that “earned” it changes too.

What is attribution modeling?

A customer watches a TikTok video, clicks a Facebook ad two days later, then searches the brand name on Google over the weekend and places an order. The attribution model decides who gets credit for that order through three settings:

• Which touchpoints count: clicks only, or ad views as well.
• The lookback window (see attribution window): touchpoints that happened too long before the conversion are ignored.
• How credit is split: all of it to one touchpoint, a fixed split, or a split calculated by an algorithm.

Types of attribution models

The percentages in a rule-based model are set by a person. No data proves them right.

What changed from 2023 to 2026

Google. Since November 2023, GA4 no longer offers four models: first click, linear, time decay and position-based. GA4 now has three options: Data-driven, Paid and organic last click, and Google paid channels last click (Google Analytics Help, Attribution models, accessed October 2026). Google Ads dropped the same four models, moved the conversion actions that used them to data-driven, and made data-driven the default for most conversion actions (Google Ads Help, About attribution models, accessed October 2026).

Meta. As of January 12, 2026, the Ads Insights API stopped returning data for the 7-day view and 28-day view windows. The remaining windows are 1, 7 and 28 days after a click, 1-day engaged view, and 1 day after a view (Meta for Developers, October 2025).

Apps. Mobile measurement platforms usually give credit to the last touchpoint inside the window. Adjust defaults to 7 days for device-matched clicks, and an install that falls outside the window with no other valid engagement is recorded as organic (Adjust Help Center, Attribution windows, accessed October 2026). On iOS, AdAttributionKit postbacks (iOS 17.4 and later) carry no user-level or device-level data (Apple Developer, AdAttributionKit, accessed October 2026), so splitting credit per user is hard.

Why platform reports don’t match

Each platform credits itself under its own rules. A hypothetical example: a customer sees a Meta ad, clicks a Google ad the next day, and buys an order worth VND 1,500,000. Meta (1-day view) records 1 order, and Google Ads also records 1 order. Add the two reports together and you get VND 3,000,000, twice the actual revenue.

The usual fix is to pick one shared source (GA4, an MMP or the CRM) for splitting budget across channels, and to use each platform’s own numbers only for optimizing inside that channel.

Attribution vs. MMM and incrementality testing

Attribution only splits credit for conversions that were recorded. It cannot answer “how many orders do we lose if we turn this channel off?” That question needs marketing mix modeling, which works on weekly aggregated data, or an incrementality test with a control group.

FAQ

Should I use last click or data-driven?

An account with few conversions gives data-driven little data to learn from. The Model comparison report in Google Ads puts two models side by side, so you can see how cost per conversion and ROAS shift before you switch for good.

Does changing the model affect bidding?

Yes. Smart Bidding optimizes to the conversions column, so when credit is redistributed across campaigns, your old CPA or ROAS targets may no longer fit. Google recommends testing first and checking the impact on profit.

If your Meta, Google and GA4 numbers don’t line up, the Adtek team can review your measurement setup with you. Reach us through the contact page.