How to measure marketing with AI: attribution and revenue by channel

To measure marketing with AI, use the CRM as the source of truth: every lead recorded with its source, every sale tied to that lead, and the analysis done in opportunities, sales, revenue and cost per customer by channel. Attribution tells you which channel gets the credit for each sale. AI doesn't change that logic, but it makes it mandatory, because it multiplies the number of assets, campaigns and messages that need to be evaluated.

This article shows how to set up measurement, which metrics to track and how to isolate the effect of AI. It's part of the guide to AI in marketing for B2B companies.

Why does AI make measurement more urgent?

Before, a small team launched a few campaigns and a few pieces of content a month. You could keep track by eye, even without a good measurement system. With AI, the same team produces far more ad variations, more content and more nurturing messages.

Without measurement, that creates two problems. The first is deciding based on whatever shows up fastest: clicks, likes, cost per lead. The second is not being able to say whether AI brought results or just more review work. Measurement is what turns volume into learning.

What attribution is, in simple terms

Attribution is the rule that decides which channel or campaign gets the credit for a sale. In B2B, a customer usually has several touchpoints with the company before buying: saw an ad, read an article, came back through search, messaged on WhatsApp. Each model distributes the credit differently.

For most small and mid-sized B2B companies, the practical path is to record each lead's first and last touch in the CRM and compare the two. That already answers most budget questions.

Build the foundation: the CRM as the source of truth

Ad platforms and website analytics tools show part of the story. Each uses its own attribution window and rules and tends to count the conversions it took part in. Adding up all their reports usually gives you more sales than the company actually made.

The CRM solves this because it records the real sale, just once. For that to work, it needs:

1. Source captured at entry. Campaign parameters, landing page and, when possible, the click identifier, recorded on the lead. Including when the conversation starts on WhatsApp.
2. Up-to-date stages. Lead, qualified lead, opportunity, sale, lost with a reason.
3. Sale value. To measure revenue, not just quantity.
4. Self-reported source. The "how did you hear about us" field, with options such as referral, event and AI tools, to capture what tracking can't see.

The technical link between media platforms and the CRM is covered in media and CRM integration.

The metrics that matter

Organize metrics into three levels and don't mix them when making decisions.

Activity shows what the team did: assets published, campaigns live, messages sent. It's useful for managing the work, not for evaluating results.

Response shows how the audience reacted: clicks, visits, leads, cost per lead, response rate. It's useful for diagnosing an asset or a campaign.

Business shows what turned into revenue:

• qualified leads by channel and by campaign;
• opportunities and sales by channel;
• revenue by channel;
• cost per opportunity and cost per customer by channel;
• time from lead to sale, by channel.

Budget decisions are made with business metrics. When volume is low, look at longer periods and use intermediate stages, such as opportunity, to get a signal before the sale.

Want to see how much sales and revenue each channel brings in, with the CRM as the source?
Discover Nexo's AI for marketing

How to isolate the effect of AI

Knowing that marketing sells is one thing. Knowing whether AI improved marketing is another. Four practices help:

1. Baseline. Before implementing, record where the metrics AI is supposed to change stand today: cost per opportunity, qualified lead rate, production time.
2. One initiative at a time. If AI creatives, a new landing page and personalized nurturing all start in the same month, there's no way to know which one made the difference.
3. Comparison group. When possible, keep part of the campaigns or leads on the old process and compare.
4. Full cost. Include tools, review hours and implementation. An initiative that brings more opportunities but eats up a lot of review time may not pay off.

The method for measuring AI's return in the sales funnel, with cost and gain calculations, is in how to measure the ROI of AI applied to sales.

Where AI helps with the analysis itself

AI is also useful for reading the data, as long as the numbers come from the CRM and the platforms, not from the AI. Good uses:

• summarizing the week's performance from a CRM export;
• flagging campaigns with an out-of-pattern cost per opportunity;
• grouping the loss reasons recorded by sales reps;
• cross-checking the tracked source against the source reported by the lead.

Always check the numbers it cites against the source. AI models can get math wrong and mix up columns, especially with large spreadsheets.

Platforms to optimize, CRM to decide

One rule that avoids a lot of arguments: use the ad platforms' numbers to optimize within them, and the CRM's numbers to decide budget across channels. The platform needs its own data to adjust bids and audiences. Leadership needs the real sale.

For the platform to also optimize for sales, send it the CRM events back, as we show in optimizing campaigns for sales.

A minimum dashboard

You don't need a complex system to get started. A dashboard with these columns, by channel and by month, already changes the conversation:

If any column can't be filled in, that's the first thing to fix in your tracking.

A review routine between marketing and sales

The dashboard only helps if it's read together. A simple routine:

• Every week: marketing looks at response metrics to adjust assets and campaigns. Sales checks that stages and loss reasons are up to date in the CRM.
• Every month: marketing and sales look at business metrics by channel together. This is where you decide where to increase or cut budget and which AI initiatives to expand.
• After each full sales cycle: review the qualified lead criteria, comparing them with who actually bought.

The monthly meeting is usually the one that changes the relationship between the teams the most. When both sides look at the same table, the discussion about lead quality stops being opinion and starts having numbers.

Common mistakes

• Adding up the conversions each platform reports.
• Deciding budget by cost per lead.
• Rolling out several AI initiatives at the same time.
• Not recording the source of leads that come in through WhatsApp.
• Leaving out the cost of reviewing AI-made assets.
• Asking AI to calculate numbers without checking them.

Conclusion

Measuring marketing with AI means measuring marketing with more discipline. The CRM becomes the source of truth, attribution defines who gets the credit, and budget decisions start using opportunities, sales and cost per customer by channel.

With a baseline, one initiative at a time and the full cost in the math, it becomes possible to say whether AI brought sales or just more assets. That answer is what justifies expanding, adjusting or stopping.

Want to know how much in sales each of your marketing channels brings in? Take Nexo's free assessment.
Take the free assessment