What multi-touch attribution means
Most customers don't buy on their first visit. They read a blog post from search, come back from a newsletter, click a retargeting ad and then pay. Multi-touch attribution (MTA) gives each of those touchpoints a share of the sale. Single-touch models give all of it to one: the first visit (first-touch attribution) or the last one (last-touch attribution).
The idea is fair on paper. Search started the relationship, email kept it going, the ad closed it. Each deserves something. The hard part is deciding how much, and that decision is the attribution model.
The common multi-touch models
| Model | How credit is split | Good for |
|---|---|---|
| Linear | Equal share to every touch | Long, even journeys where every step matters |
| Time-decay | More credit the closer a touch is to the sale | Short sales cycles and promotions |
| Position-based (U-shaped) | 40% first, 40% last, 20% split across the middle | Valuing both discovery and closing |
| W-shaped | Heavy credit to first touch, lead creation and the sale, the rest spread | B2B funnels with a clear sign-up step |
| Data-driven | An algorithm compares paths that converted with paths that didn't | Large ad accounts with lots of conversions |
Rule-based models (the first four) are transparent: you can redo the math by hand. Data-driven models are opaque and need a lot of conversions, so they mostly live inside the big ad platforms.
Worked example
A customer pays $300. Before paying they had four touches: an organic search visit on day 1, a newsletter click on day 9, a Meta ad click on day 13 and a direct visit on day 14, when they bought.
| Touch | First touch | Last touch | Linear | Position-based (40/20/40) |
|---|---|---|---|---|
| Organic search (day 1) | $300 | $0 | $75 | $120 |
| Newsletter (day 9) | $0 | $0 | $75 | $30 |
| Meta ad (day 13) | $0 | $0 | $75 | $30 |
| Direct (day 14) | $0 | $300 | $75 | $120 |
The same $300 sale gives search anywhere from $0 to $300 depending on the model. Under linear, the Meta ad earns $75. If that campaign cost $150 per customer, its ROAS is 0.5 under linear and 0 under last touch. Your budget decision flips with the model, which is why you should pick one on purpose and keep it.
Many tools drop direct visits from the path before splitting credit, because a direct visit usually means the customer already knew you. Without the direct touch, position-based gives search $120, the Meta ad $120 and the newsletter $60.
What you need to measure it
- Every touch, per person. Each visit needs its source: UTM parameters, a click ID or a referrer. Untagged visits fall into direct traffic.
- One identity across visits. A cookie, a logged-in user ID or an email has to tie day 1 to day 14. Without consent to store an ID, many visits can't be joined, and multi-touch paths shrink to single visits.
- The payment, tied to that person. Revenue from your payment provider, linked to the visitor who paid.
- A model and an attribution window (what a window is), applied the same way to every sale.
Even with all four, paths have gaps: people switch devices, clear cookies, decline consent and hear about you in places nothing records, such as a podcast or a group chat (dark social). Multi-touch attribution splits credit among the touches you saw, not the ones that happened.
When multi-touch helps, and when it doesn't
It helps when your journeys are long and you run several channels at once, for example content, email and paid ads for a B2B product with a two-month sales cycle. It shows that email nurtures deals that ads later close.
It helps less for small teams with short journeys. If most customers buy within one or two visits, every model gives nearly the same answer, and a simple first-touch view is easier to act on. It also can't prove that a touch caused a sale. A retargeting ad shown to people who were already about to buy collects credit under every multi-touch model. Only a holdout test, where some people don't see the ad, measures that.
Common mistakes
- Adding up platform-reported conversions. Meta and Google Ads each run their own attribution and each claims the sale. Their totals exceed your real revenue.
- Switching models to make a channel look better. Pick a model before you look at the results.
- Treating fractional credit as causation. Credit says a channel was present on the path, not that it made the sale happen.
- Ignoring missing touches. Paths without consent or across devices look shorter than they were, which inflates whatever touch you did record.
- Comparing tools with different windows. A 7-day and a 90-day window credit very different touches for the same sale.
How MetricRaven handles multi-touch attribution
MetricRaven doesn't run multi-touch or data-driven models. It credits each payment once, in full, to the customer's first tracked visit: the earliest pageview across every visitor ID linked to that customer. Every report uses that one rule, so channels, campaigns and pages add up to your real revenue. See revenue attribution model for why.
You can still see the touches that came after the first one:
- Assisted revenue in the Meta Ads and Google Ads reports counts payments from customers who clicked a campaign's ads before paying but first arrived some other way. It sits next to first-touch revenue and is never added to it.
- Paths to payment shows the pages each paying visitor viewed, in order, across all their visits, and the share of each path's visitors who paid.
- Customer panels on the Revenue tab show each customer's first touch, with a button that opens their full journey of visits.
- Insights compares channels on purchase rate, revenue, days to buy and visits to buy (buyer insights).
Payments link to visits when checkout passes the visitor ID (for example Stripe Checkout), when you identify users, or when you attribute by email. Payments with no tracked visit show as Unattributed instead of being guessed.
Questions
- What is the difference between single-touch and multi-touch attribution?
- Single-touch gives the whole sale to one visit, the first or the last. Multi-touch splits it across every recorded touch, by a rule such as equal shares (linear) or more weight near the sale (time-decay).
- Which multi-touch attribution model is best?
- None is correct in general. Linear suits long journeys where every step matters, position-based values discovery and closing, and time-decay suits short cycles. Pick one before you look at results and keep it, so trends stay comparable.
- Does Google Analytics 4 support multi-touch attribution?
- GA4 offers data-driven attribution, which spreads credit across touches using its own algorithm, plus last-click models. Google removed the rule-based first click, linear, time decay and position-based models in November 2023.
- Is multi-touch attribution accurate?
- It is only as complete as the touches you can record. Visits without consent, across devices or from untracked places like podcasts are missing, and credit says a channel was present, not that it caused the sale.
- Does MetricRaven support multi-touch attribution?
- No. MetricRaven credits each payment to the customer's first tracked visit, and shows later ad clicks as assisted revenue and page sequences in Paths to payment, without splitting the money.