By Mehdi HassanWhat should marketing website analytics measure?
Marketing website analytics should measure whether the website creates and moves pipeline. For a B2B company that means qualified conversions (demo requests, contact forms, trials), the conversion rate of the pages that produce them, the pipeline and revenue those conversions become, and how visible the brand is where buyers now research. Everything else is there to explain changes in those numbers.
Most B2B dashboards get this backwards. They open with sessions, users and bounce rate, because those are what GA4 shows first, and they end before anyone knows whether a lead turned into a deal. The result is a report that grows every quarter while leadership trusts it less.
This hub sets out a short metric set that holds up in a board meeting, explains what GA4 can and can't see in 2026, and links to step-by-step guides for each piece: tracking AI referrals, measuring AI visibility, self-reported attribution, Search Console, reporting, attribution models and the CRM connection.
- Share of AI-attributed leads that landed on homepages, TripleDart study of 28 B2B SaaS firms
- 81.7%
- Leads from general blog posts in the same study, despite 12k sessions
- 0
- Lower top-result CTR when an AI Overview shows (Ahrefs)
- 58%
Why do sessions and traffic make poor goals for a B2B site?
Sessions make poor goals because they no longer track demand. Fewer searches end in a click, more research happens inside AI tools, and the traffic that does arrive varies wildly in value. A site can lose a large share of its blog traffic and gain pipeline in the same quarter, and a traffic goal would call that a failure.
The data from 2026 makes the gap obvious. Ahrefs found the top organic result's click-through rate falls 58% when an AI Overview is shown. SparkToro estimated that 68% of US Google searches in early 2026 ended without a click, though it changed its panel data that year, so the trend isn't a clean comparison. And in TripleDart's study of 28 B2B SaaS companies in Q1 2026, general blog posts drew about 12,000 sessions and zero leads, while homepages took 81.7% of AI-attributed leads.
Traffic still matters as a diagnostic. A sudden drop in organic sessions to your pricing page is worth investigating. But it belongs on the second row of the dashboard, under the numbers that show whether the site is doing its job. The same logic drives our view of what pages a SaaS website should have: build and measure the money pages before the blog.
Which B2B website metrics belong on the dashboard?
A B2B website dashboard needs four outcome metrics and four diagnostics. The outcome metrics answer "is the site producing pipeline?" The diagnostics answer "why did that change?" If a metric doesn't help with either question, leave it off.
Metric | Type | What it tells you | Source |
|---|---|---|---|
Qualified conversions | Outcome | Demo, contact and trial requests that sales accepts | GA4 key events + CRM |
Money-page conversion rate | Outcome | Whether homepage, pricing and product pages persuade | GA4 |
Website-sourced pipeline | Outcome | Dollar value of opportunities that started on the site | CRM |
AI visibility share | Outcome | How often you appear for buyer prompts in AI tools | Prompt audit or tool |
Traffic by channel | Diagnostic | Which sources grow or shrink, including an AI channel | GA4 |
Search impressions and clicks | Diagnostic | Demand for your topics in Google, before the click | Search Console |
Self-reported source | Diagnostic | What buyers say brought them, including dark channels | Form field + CRM |
Page speed and errors | Diagnostic | Technical problems that quietly cut conversion | Search Console, monitoring |
We cover how to build this as a live view in our guide to the marketing website dashboard. The order matters more than the tool. Put outcomes at the top, diagnostics below, and a sentence of commentary next to anything that moved.
What can GA4 see, and what does it miss?
GA4 sees what happens on your site after a visit begins: where the visitor came from, if a referrer was passed, which pages they viewed, and which events they triggered. It can't see an AI answer that mentioned you without a click, a buyer who read about you in ChatGPT and typed your URL later, or what happened after the form fill.
- It sees: sessions by source, landing pages, key events like form submissions, and referral traffic from AI tools that pass a referrer.
- It misclassifies: visits from AI mobile apps and some desktop apps that strip the referrer, which land in Direct.
- It misses: AI answers without clicks, word of mouth, podcasts, private Slack and community recommendations.
- It stops at: the lead. Without a CRM connection, GA4 can't tell you which visits became revenue.
The first gap is partly fixable. A custom channel group can pull ChatGPT, Perplexity, Claude, Gemini and Copilot referrals out of the general Referral bucket; our step-by-step guide to tracking AI traffic in GA4 gives the regex and the channel ordering. The other gaps need different instruments, covered in the next sections.
Why AI traffic often shows up as Direct
When someone clicks a link inside a mobile app, the app often opens it in an in-app browser or the default browser without passing a referrer. GA4 has nothing to attribute the visit to, so it files it under Direct.
The practical consequence is that your AI channel in GA4 is a floor, not a total. If Direct traffic to deep pages (not just the homepage) rises while organic is flat, some of it is likely AI-referred. A self-reported attribution field is the cleanest way to size that gap.
How do you measure AI search visibility?
You measure AI search visibility by asking AI tools the questions your buyers ask and recording whether, where and how your brand appears. Analytics can't do this because most AI answers don't send a click. A monthly audit of 20 to 30 buyer prompts, run in ChatGPT, Perplexity, Google AI Overviews and Gemini, gives a usable baseline.
The metrics that matter are simple: mention rate (how often you're named), citation rate (how often your pages are linked as a source), share of voice against named competitors, and accuracy (whether what the AI says about you is true). Engines differ a lot. The Semrush AI Visibility Index, built on 126 million prompts, found ChatGPT cites about 15 sources per answer against about 3 for Gemini, so expect different numbers per engine.
Our guide to measuring AI search visibility includes a prompt-tracking sheet you can run by hand before buying any tool. What moves these numbers is mostly off-site: in Ahrefs' study of 75,000 brands, web mentions correlated 0.664 with AI Overview visibility, against 0.218 for backlinks.
How should B2B teams handle attribution?
B2B teams should combine simple software attribution with what buyers say about themselves. Software attribution captures the touches it can see, usually first and last. A "How did you hear about us?" field captures the touches it can't, including AI tools, podcasts and peers. Together they give a picture you can act on.
Approach | Captures well | Misses | Best for |
|---|---|---|---|
First touch | Which channel introduced the buyer | Everything after the first visit | Judging awareness channels |
Last touch | What finally triggered the form fill | The research that built intent | Judging conversion pages |
Multi-touch | Tracked touches across a journey | Dark channels, offline and AI answers | High-volume, tracked journeys |
Self-reported | What the buyer remembers, including dark channels | Precision and small touches | Every B2B site |
Each of these has its own guide. B2B attribution when buyers research in ChatGPT explains why AI research breaks click-based models. Self-reported attribution shows how to add the field without hurting conversion. And multi-touch attribution for B2B helps you decide whether the expensive tools are worth it at your stage. Our honest view: most startups should run first touch, last touch and a self-reported field before paying for anything more.
How do you connect website analytics to pipeline?
You connect website analytics to pipeline by passing source data from the website into the CRM at the moment of conversion. Hidden form fields capture UTM parameters, the first landing page and the referrer, and the CRM stores them on the contact and the deal. Once that's in place, you can report revenue by source instead of visits by source.
Standardize UTMs
Agree a naming convention for source, medium and campaign, and use it on every paid, email and social link.
Capture source on the first visit
Store UTMs, referrer and landing page in a first-party cookie so they survive until the visitor converts days later.
Pass it through hidden fields
Add hidden fields to every form and populate them from the stored values on submit.
Map fields to CRM properties
Write the values to contact and deal properties in HubSpot or Salesforce, with first-touch values that never get overwritten.
Report on closed deals
Build pipeline and closed-won reports grouped by original source and by self-reported source, side by side.
The full setup, including HubSpot's built-in source properties and the Salesforce equivalent, is in our guide to connecting website analytics to HubSpot or Salesforce. This is the step most teams skip, and it's the one that turns a marketing report into a revenue report.
Which Search Console reports matter for a B2B site?
Search Console matters because it's the only source that shows Google demand before the click: impressions, average position and click-through rate by query and page. For a B2B site, the Performance report filtered to money pages, the Pages indexing report and URL Inspection cover most of what you need each week.
Impressions are especially useful now. If impressions on a query rise while clicks fall, you may be appearing in an AI Overview that answers the question without a visit. Google includes AI Overviews and AI Mode in the normal Web search data, so you can't filter them out, but the pattern is a useful signal.
Our Google Search Console guide for startups lays out a 15-minute weekly routine focused on the pages that produce revenue, rather than the whole site.
What does a monthly website report look like?
A good monthly website report fits on one page: outcomes at the top, diagnostics below, and three sentences on what changed and what you'll do about it. Leadership wants to know whether the site produced pipeline, whether that's going up, and what the team is changing. They don't want forty charts.
- Headline: qualified conversions, website-sourced pipeline and closed-won revenue, with last month and the same month last year.
- Money pages: conversion rate on homepage, pricing, product and comparison pages.
- Visibility: Search Console clicks and impressions, plus AI visibility share from the prompt audit.
- Sources: traffic by channel including the AI group, next to self-reported sources.
- Actions: what you shipped, what you learned, and what you'll change next month.
Our SEO reporting template for B2B gives this as a fill-in page. If you also publish content, measuring content ROI shows how to judge articles by assisted pipeline rather than pageviews.
What should you set up first?
Set up the pieces that connect the website to revenue first, then add visibility measurement. A team starting from a default GA4 install can get to a credible setup in a few focused weeks of work, in this order.
- Mark demo, contact and trial submissions as key events in GA4.
- Add hidden source fields to every form and map them to CRM properties.
- Add a "How did you hear about us?" field with an open text option.
- Create an AI channel group in GA4 for ChatGPT, Perplexity, Claude, Gemini and Copilot.
- Verify the domain in Search Console and filter the Performance report to money pages.
- Run a first prompt audit of 20 to 30 buyer questions and save the results as a baseline.
- Build the one-page monthly report and send it, even if half the cells are empty.
If the audit shows your site itself is the problem (slow, hard to update, or built so that AI crawlers can't read it), the fix is a rebuild rather than more tracking. Our SaaS website design guide covers what a site built for both buyers and answer engines needs. For the broader approach to AI search, start with our answer engine optimization guide and the AEO strategy for B2B SaaS.
