What is multi-touch attribution?

A first-touch report gives all the credit to whatever introduced a person to the company, and a last-touch report gives it to the demo form they finally filled out. Neither says much about the months in between, which is where a B2B deal spends much of its time. Multi-touch models fill that gap by dividing the credit among the interactions the system recorded, and the division depends entirely on the rule the team picks. Run one buyer's recorded touches through a linear rule and a time decay rule and the same blog post, webinar, and retargeting ad earn different shares. The model changes how the record is presented. It can only distribute credit among touches that were captured, which is why ABM attribution(opens in a new tab) usually starts with a harder look at what is missing.

How the common models split credit

Linear attribution gives every recorded touch an equal share. Time decay gives more to touches close to the conversion, on the logic that recent interactions mattered more. U-shaped models weight the first touch and the lead-creation touch heavily, with the remainder spread across the middle, and W-shaped models add a third peak at opportunity creation. Teams often split U-shaped credit 40, 40, and 20 and W-shaped credit 30, 30, 30, and 10, but those are conventions from tools and blogs and no standard exists, so check what your platform does.

Data-driven models learn weights from converting and non-converting paths. They need volume to do it: Google Ads recommends at least 200 conversions and 2,000 ad interactions in 30 days before data-driven attribution is dependable. Many B2B ad accounts record far fewer conversions than that, which pushes teams back toward fixed rules they can explain in a meeting.

GA4 no longer lets you compare the rules

Google announced in April 2023 that first click, linear, time decay, and position-based attribution would be removed from Google Ads and GA4, and GA4's own help now lists them as unavailable as of November 2023. GA4 reporting offers data-driven attribution and a few last-click variants. A team that wants to see how a U-shaped rule would change its numbers has to build that view in a warehouse, a spreadsheet, or a dedicated attribution tool, using its own touch data.

What a clickstream can't record

Every multi-touch model works on touches somebody logged. A rule can't distribute credit to an interaction that isn't in the data, and several kinds of buying activity routinely aren't:

• Research that happens in private channels, on review sites and in peer conversations, none of which produce a session you can tie to a name.
• Offline interactions such as events, dinners, and sales calls, unless someone logs them against the right contact.
• Ad impressions that were seen and never clicked.
• Activity from browsers and devices where the identifier didn't persist, so one visitor looks like several or none.
• Other people on the committee, when only one person's browser history is being stitched together.

The touches that were captured take all the credit, and the report still looks precise.

Account-level versus contact-level attribution

Most attribution tools build one person's path: a cookie or an email address, followed from first visit to form fill. B2B deals aren't decided by one person. The person who clicks is often not the person who signs. Attributing at the account level pools everyone's touches into one path, which captures more of the activity and can't say who was reached. Attributing at the contact level keeps each touch tied to a named person and a role on the opportunity, so the credit sits beside evidence of who saw what. Which model sits on top matters less than whether the touch data underneath names real people.

Using attribution next to other methods

Treat the output as one view among several. Influence reporting(opens in a new tab) asks a simpler question, whether any qualifying touch happened on the deal, and has fewer modeling choices to defend. Marketing mix models work from aggregate spend and outcomes, and incrementality tests compare an exposed group to a held-out group, so both can speak to channels that leave no click trail. Each has its own costs and blind spots, and none of them gives perfect attribution, so teams are better off holding several views loosely than treating one model as the answer. More on building that case is in proving marketing impact(opens in a new tab) and ABM measurement(opens in a new tab).