Funnel AnalysisFinding where value leaks

What funnel analysis is for#

Every business has a sequence of steps between a stranger's first visit and a paying customer. At each step, some people continue and some leave. Funnel analysis makes that sequence visible, so a team can stop arguing about whether 'the website' or 'the leads' are the problem and point to the specific step where value leaks.

It is a diagnostic tool. It does not tell you why people leave, and it does not by itself tell you what marketing caused. It tells you where to look, which is often the most valuable thing an analyst can offer a busy team.

Fix the narrowest point in the pipe before you pour more water in.

Step 1: define stages that reflect real commitment#

Choose stages that represent genuine increases in commitment, not just page loads. Each stage should be a distinct event you can measure reliably, ideally already defined in your GA4 events and CRM.

Two decisions shape the analysis. Open or closed funnels: does a person have to pass through every step in order, or can they enter mid-way? Time window: must all steps happen in one session, or within some days? Ecommerce funnels are often session-based; B2B funnels span weeks and must be joined across web analytics and CRM.

Step 2: measure step conversion, not just the total#

The overall conversion rate (customers divided by visitors) is a useful headline and a poor diagnostic. Two funnels with the same overall rate can have completely different problems. Calculate the conversion rate at each step, so you can see which transition loses the most people.

Keep the population consistent across steps. If the first step counts sessions and the last counts customers, the rates mix units and mislead. Choose users or sessions and stick with that choice throughout one funnel.

Step 3: segment before you conclude#

Averages lie in funnels more than anywhere else. A drop at the checkout step might be entirely caused by one payment method failing on one browser, or by one campaign sending visitors who were never going to buy. Segment every funnel by at least these dimensions before drawing conclusions.

• Traffic source and campaign, using clean UTM parameters. Low-intent sources depress early steps without anything being wrong with the site.
• Device and browser. Mobile and desktop funnels often differ sharply; a broken element on one device hides inside the average.
• New and returning visitors. Returning visitors usually convert better; a change in mix moves the total.
• Geography. Delivery coverage, payment preferences and language vary by region, particularly between metros and smaller cities.
• Product or service line. A low-converting category can drag down the whole funnel.

The point of segmentation is to separate traffic problems from experience problems. If one source's visitors leave at the first step while others continue normally, the issue is targeting, not the website. If every source drops at the same step on mobile, the issue is the mobile experience.

Segments with very few people should be read with caution.

Not every drop-off is a problem#

A funnel that narrows is not broken; it is doing its job. Many visitors arrive to compare prices, check a delivery date or read an article, and leave satisfied without buying. A B2B site that converts every visitor into an enquiry would be attracting only people who already decided, which means its marketing is not reaching anyone new.

The question is not 'why do people leave?' but 'are the right people leaving at the wrong step?' Focus on visitors who showed real intent (viewed pricing, added to cart, started a form) and then abandoned. Those are the leaks worth fixing. Early-stage exits by poorly matched visitors are a targeting question for media, not an experience question for the website.

Micro-funnels inside a step

When one step is clearly the bottleneck, zoom in. A checkout or enquiry form is itself a funnel: field by field, page by page. Tracking where people stop within the form (which field they last touched, which error they saw) often reveals a single cause: a mandatory phone field, an unexpected delivery charge, a payment option that fails on one device. Our guides to form optimisation and checkout optimisation go deeper.

Step 4: size the opportunity at each step#

Not every leak is worth fixing first. A step with a poor rate but few people passing through may matter less than a decent-looking step with huge volume. Estimate the value of improvement at each step: if this step's rate improved modestly, how many more customers would reach the end, and what would they be worth?

Step 5: find out why, then test#

Funnel data tells you where; it rarely tells you why. Pair it with qualitative evidence: session recordings and heatmaps on the leaking step, on-page surveys, sales call notes, customer service logs. Form a specific hypothesis ('mobile visitors cannot see delivery charges until the last step, so they abandon on discovery'), then test a fix through an A/B test where volume allows.

B2B funnels: joining web and CRM#

In B2B, the most important stages happen after the website: lead qualified, meeting held, proposal sent, deal won. Web analytics cannot see them. Build the funnel in the CRM, with lead source captured at the point of enquiry, so you can see whether a channel that produces many enquiries also produces won deals. Often it does not, and that discovery alone justifies the work.

Watch stage velocity as well as conversion. A step where leads wait for weeks before a first sales call loses buyers to faster competitors; speed of response is frequently the cheapest funnel fix available.

Use a shared lead ID that travels from the web form into the CRM, so that each won deal can be traced back to its original source and landing page. Without it, the web funnel and the sales funnel remain two separate stories that cannot be joined.

Common traps#

Finally, avoid optimising one step at the expense of the next. Removing all qualification questions from a form may raise enquiry rates and flood sales with poor leads. Judge every change on the end of the funnel, not only on the step you touched. For tracking how funnels change over time for different groups of customers, see cohort analysis.

Published by Fabulous.Media, a network of specialist marketing agencies. Updated 9 October 2026. Platform features change often; check current official documentation before acting on platform-specific detail.