Why attribution is harder in B2B
In B2C the customer journey is short: see an ad, click, buy. One or two touches. Last-click attribution works acceptably.
In B2B the journey looks like this: LinkedIn ad (didn’t click) - organic article in search (read for ten minutes) - retargeting (clicked but closed the tab) - colleague recommendation (visited directly) - form on the site - three calls with an SDR - demo with an AE - 45 days of discussion - purchase.
Which channel gets credit for the sale? Every existing model gives a partially wrong answer.
The four main models
Last-click
100% of the credit goes to the last channel before conversion. Usually that’s organic search or a direct visit.
The problem: it undervalues channels that work at the top of the funnel (LinkedIn, display advertising, content). The marketer cuts budget from what actually builds demand.
Where it’s applicable: short deal cycle (up to seven days), one clear acquisition channel.
First-click
100% of the credit goes to the first touchpoint. The opposite extreme.
The problem: it ignores everything that happens before the purchase. If the first touch is organic, you’ll never know what pushed the final decision.
Where it’s applicable: when you need to understand what builds initial product awareness.
Linear
Credit is distributed equally across all touchpoints.
The problem: the first view of a LinkedIn ad (five seconds) gets the same weight as a demo call that lasted an hour.
Where it’s applicable: as a baseline model when you have no data on interaction quality.
Time-decay
Later touchpoints get more weight. The logic: the closer to the purchase, the more important.
The problem: it undervalues early channels that build demand. Especially unfair to content marketing.
Where it’s applicable: short deal cycles where the final touches genuinely matter more.
Data-driven attribution
Google Ads and GA4 offer a data-driven model: the algorithm determines each touchpoint’s weight based on historical conversion data.
Sounds like a solution. But in B2B there are problems:
- You need thousands of conversions to train adequately. Most B2B companies have dozens per month.
- The conversion Google sees (a lead on the site) is not the final conversion (a closed deal). The algorithm trains on the wrong goal.
- The model works within Google’s data. Touchpoints in LinkedIn, direct, referral are a black box.
What actually works for B2B
Instead of one “correct” model - a combination of approaches:
First + last touch in CRM
Store two fields in your CRM: the source of the first touchpoint and the source of the last touchpoint before the lead. Analyze both. This tells you what builds awareness and what triggers the inquiry.
Revenue by channel
The most important metric: how much ARR each channel brought in per quarter. Calculated manually: take closed deals for the quarter, look at their source in the CRM, sum by channel.
Not real-time, but an honest number based on real money.
Tools for closing the loop between ads and CRM
For automation: you need a tool that sees ad platform data (spend, clicks) and CRM data (leads, deals) simultaneously and builds end-to-end analytics.
Prooflytics solves exactly this - connects data from Meta, Google, and LinkedIn with the CRM funnel and shows cost per closed deal by campaign without manual spreadsheets.
Practical advice
Don’t spend weeks choosing the “ideal” attribution model. Start simple:
- UTM on every advertising link
- Source saved in CRM when the lead is created
- Once a month: export closed deals with their source, compare against ad spend by channel
This takes two to three hours per month and gives 80% of the information needed for budget decisions.
Add automation once the basic discipline is running consistently.
Takeaway
There is no attribution model that perfectly describes the reality of B2B sales. Last-click understates top-of-funnel, first-click understates bottom-of-funnel, linear is too mechanical.
The right approach: accept the limitations of any single model, use several metrics in parallel, and make decisions based on revenue by channel rather than just CPL or CPC.