ENG
VIE
+84 9323 494 34 
Multi-touch Attribution (MTA)
October 20, 2026 Multi-touch attribution (MTA) is a way of splitting the credit for a conversion across the several touchpoints a customer passed through, instead of giving 100% to the last click or the first click. MTA is one family of models within attribution modeling.
What is multi-touch attribution?
Someone buying a water purifier might watch a TikTok video, click a Facebook ad and open a promotional email before searching for the brand name on Google. Last click hands the whole order to Google. MTA takes the full path as the unit of analysis and divides the order among the touchpoints, and the shares always add up to exactly one conversion.
There are two ways to split the credit:
- Fixed rules: linear (equal shares), time decay (touchpoints closer to the conversion get more), U-shaped (40% to the first touch, 40% to the last, 20% split across the middle) and W-shaped.
- Algorithmic (data-driven): a model learns from paths that did and did not end in a conversion to estimate how much each touchpoint raises the odds of converting.
Adtek also has a longer Vietnamese-language article with a full comparison of the models, guidance on choosing one for a Vietnamese business and a checklist to run before you use attribution data to split budget.
Hypothetical example
A VND 2,000,000 order passes through four touchpoints: a TikTok video, a Facebook ad, an email and a branded search on Google. Credit by model, in VND:
| Model | TikTok | |||
|---|---|---|---|---|
| Last click | 0 | 0 | 0 | 2,000,000 |
| Linear | 500,000 | 500,000 | 500,000 | 500,000 |
| U-shaped | 800,000 | 200,000 | 200,000 | 800,000 |
For the very same order, TikTok either “brought in nothing” or “brought in VND 800,000,” depending on the model. That is why teams need to agree on one model before comparing channels.
Why MTA is harder in 2026
MTA only works when touchpoints can be tied back to the same person. That condition has weakened a lot:
- GA4 dropped the first click, linear, time decay and position-based models in November 2023, keeping only data-driven and two last click variants (Google Analytics Help, accessed October 2026).
- Meta stopped returning 7-day and 28-day view-through data through the Ads Insights API as of January 12, 2026 (Meta for Developers, October 2025).
- On iOS, AdAttributionKit postbacks carry no user-level or device-level data (Apple Developer, accessed October 2026).
- In October 2025, Google announced it would retire the Attribution Reporting API on both Chrome and Android, along with several other Privacy Sandbox technologies, because of low adoption (Privacy Sandbox, October 2025).
So the most complete MTA today usually runs on first-party data: websites with logins, CRMs, and apps with user accounts.
MTA vs MMM vs incrementality testing
| MTA | MMM | Incrementality testing | |
|---|---|---|---|
| Answers | How much each touchpoint contributed | How much each channel contributes to revenue | How many extra conversions the ads created |
| Data | Each person’s path | Aggregated weekly figures | A test group and a control group |
| Weakness | Loses track when users cannot be linked | Needs a long data history and updates slowly | Takes time and tests only one question at a time |
See also Marketing Mix Modeling (MMM).
Frequently asked questions
Can GA4 still do MTA?
Yes, but only in its data-driven form. If you want U-shaped or W-shaped models, you have to build them yourself on data exported from GA4 or inside your CRM.
Do small businesses need MTA?
If you run only one or two channels and customers buy on the same day, last click is usually good enough. MTA is worth the effort when the purchase cycle is long, several channels run at once and you have customer data to link the touchpoints.
If you need to decide whether MTA, MMM or incrementality testing suits your current budget, you can talk it through with the Adtek team via our contact page.
