Attribution

What is attribution?

Attribution is how you give credit to the marketing that led to a lead, a deal or revenue. It answers one question: which channels did the work? In B2B, the honest answer is almost always more than one.

Let's say you sell expense software. A finance manager finds your guide to month-end close through Google in March. In April, she sees your CFO's posts on LinkedIn a few times. In May, she clicks a Google ad for "expense software for startups" and books a demo. Who gets the credit?

That depends on the model you use. First-touch gives all of it to the guide. Last-touch gives all of it to the ad. Multi-touch models split it across the steps. Each one tells a different story about the same deal, and the story you believe shapes where next year's budget goes.

Why attribution matters for B2B SaaS

B2B deals take months and involve several people, so most of the research happens where your tracking can't follow. Buyers read reviews, ask peers in Slack groups, listen to podcasts and ask ChatGPT. Your analytics only meets them when they land on your site.

The tools changed too. In 2023, Google removed the first-click, linear, time-decay and position-based models from GA4 and Google Ads. GA4 now offers data-driven and last-click attribution, so most reports lean toward whatever happened right before the form fill.

Here's why I think attribution deserves real attention:

• Budget follows credit: If last-touch says paid search closes every deal, the blog and the founder's LinkedIn look useless. Teams that cut them often see paid search pipeline shrink a few months later, because fewer people know the brand when they search.
• Some channels leave no click: Podcasts, peer advice, private Slack threads and AI answers rarely show up as a visit. That's the dark social problem, and asking buyers is the only way to catch most of it.
• Sales and marketing need one story: When both teams read the same source fields in the CRM, arguments about who sourced a deal get a lot shorter.

How to set up attribution

You won't get a perfect answer, so build a few signals that check each other. Here's my process:

• Pick one software model and keep it: Use GA4's data-driven model if you have enough conversions, or last-click if you don't. Using the same model every month matters more than which one you choose, since trends only mean something when the rules stay put.
• Store first touch yourself: Save the first landing page and first source in hidden form fields, then pass them to your CRM. GA4 won't keep that history on the deal for you.
• Ask buyers directly: Add a required "How did you hear about us?" text field to your demo form. This is self-reported attribution, and it catches the podcast episode and the friend who recommended you.
• Keep sources on the opportunity: Copy the original source and the buyer's answer from the contact record to the opportunity. Then they stay with the deal until it closes, even if the contact changes.
• Send revenue back to ad platforms: Use offline conversion import to send qualified opportunities and closed deals to Google Ads and LinkedIn. Their bidding then learns from pipeline instead of from form fills.

An attribution example

The clearest public example I know comes from Graphite's study of n8n, the workflow automation tool. GA4 credited AI assistants with 0.9% of conversions. n8n's own post-signup survey put AI's share at 9%. Same buyers, two very different answers, and the survey is the one that saw the chats.

Now here's how each model would split the expense software deal from earlier:

Look at the last row. The channel the buyer remembers is the one the software models gave the least credit to. That's why I always read the form answers next to the dashboards. Our post on the AI search attribution problem shows the same gap for ChatGPT and Perplexity.

How to measure attribution

Attribution is a set of reports you compare, and the gaps between them are where the insight sits. Here's what I'd pull every month:

• Pipeline by original source: In your CRM, add up opportunities and pipeline value by the first source saved on each deal.
• Pipeline by self-reported source: Add up the same pipeline by what buyers wrote on the form. Put it next to the first report and look for channels that show up in one and barely in the other.
• Won ARR by source: Repeat both reports for closed-won deals. Some channels start lots of deals that never close, and some start few that almost always do.
• Branded search trend: Watch branded search clicks in Search Console. When channels you can't track are working, more people search your name.
• Share of Direct and unknown: Track how much pipeline lands with no clear source. If it climbs above a third, your hidden fields or UTM habits need fixing.

Common attribution mistakes

• Treating one model as the truth: Every model is a rule someone made up. Use it to spot trends, and check it against what buyers say.
• Cutting the channels that start deals: Last-touch makes early channels look weak. Before you cut content or LinkedIn, check how often they show up as first touch or in form answers.
• Using vague form options: A dropdown with "Social media" and "Online" tells you nothing. Use a free text field, or list specific options like LinkedIn, a podcast, a colleague and ChatGPT.
• Leaving sales out: Reps hear the real story on the first call. Ask them to confirm the source in the CRM before the deal moves on.
• Changing the model every quarter: A new model resets every trend line. Pick one and leave it alone for at least a year.

Here's a first step for this week. Pull your last 20 closed deals and compare the CRM source with what each buyer wrote on the form. How many match?