Why Your Website Gets Traffic but No Leads: A Layer-by-Layer Diagnosis

Table of contents

• If you are short on time: five quick checks
• Layer 1. Measurement: define what a customer is before asking why they are missing
• Layer 2. Traffic quality: are the people who can buy actually arriving?
• Layer 3. Message: what the page says in the first seconds
• Layer 4. Offer: the step you ask the visitor to take
• Layer 5. Friction: what makes acting costly
• Layer 6. Trust: reasons to believe
• Reading the layers together
• What not to do with this diagnosis
• Next step

“We have traffic but no customers” is not a diagnosis. It is a symptom, and its cause can sit in at least six different places: how things are measured, who arrives, what the page says, what it offers, how hard it is to act, and how little it is trusted. Each calls for a different intervention, and fixing the wrong one costs time and money.

Our thesis is operational: diagnosis runs bottom-up, starting with measurement, and every layer has a specific data point that confirms or rules it out. Only when one layer is in order does it make sense to move to the next. The method works for an online store, a local service business, a B2B company or an independent professional; the criteria are the same, and only the type of “customer” and “action” changes.

If yours is a B2B site with long sales cycles, this article complements why most B2B websites don’t generate leads and CRO: what it is and how to apply it to a B2B website, which we do not repeat here. What we add is the order of checks and the data to decide with.

If you are short on time: five quick checks

Before the full diagnosis, these five checks, about 30 minutes in total, rule out the cheapest problems to fix. They are our own practical criteria, not a Google rule.

1. Send a test inquiry through each channel (form, phone, WhatsApp), on mobile and desktop, and confirm it reaches a real person (layer 5).
2. Open GA4 and check that the action you care about is marked as a key event (layer 1).
3. Look at conversion by channel for the last 30 days, not the overall average (layer 2).
4. Show the first screen of your main landing page to someone who does not know you and ask what you do and for whom (layer 3).
5. Note how many inquiries came in over that period and how many ended in a sale (layer 1).

Layer 1. Measurement: define what a customer is before asking why they are missing

The first check is an uncomfortable one: many sites “without leads” do generate them, and nobody records it. Before changing anything, answer two questions.

Which action counts as a conversion, and is it measured? In Google Analytics 4, the action that counts as success is defined as a key event, described by Google as an event that measures an action particularly important to the success of your business. You mark an event as a key event once it exists. For a lead-generation site, Google lists generate_lead as a recommended event, triggered when a user submits a form or a request for information. GA4’s enhanced measurement also records form_start (the first time a user interacts with a form in a session) and form_submit, and Google notes that comparing the two shows how many users started a form versus how many submitted it. Not every form is detected automatically, so verify with a real test that the event fires.

For phone calls, WhatsApp clicks, bookings or purchases, you usually need to define custom events. Each business has its own: the store measures purchases, the clinic measures appointments, the firm measures quote requests.

Does that action connect to real sales? A submitted form is not a customer. If you do not record what happened to each inquiry (answered, qualified, sold), you will know how many forms come in but not whether they are good ones. At a minimum, note each contact’s source and outcome. This is the same underlying problem we cover in how to build a marketing measurement system.

A note on GA4 engagement metrics, because they are often misread. According to Google, an engaged session is one that lasts longer than 10 seconds, includes a key event, or has at least two page or screen views; bounce rate is the percentage of sessions that were not engaged. The definition rests on interaction, not just leaving the page, so interpret it that way.

Verdict for this layer. If you cannot answer “how many people took the important action in the last 30 days, and how many became customers?”, stop the diagnosis and fix measurement. Everything else is a guess.

Layer 2. Traffic quality: are the people who can buy actually arriving?

An overall conversion rate hides more than it shows. The right question is not how much the site converts but how much each traffic source converts.

A hypothetical example, with made-up figures used only for illustration: a site gets 10,000 visits a month. Eight thousand arrive through very broad informational content and produce 8 inquiries (0.1%). Two thousand arrive through specific service searches and produce 32 inquiries (1.6%). The total is 40 inquiries, an overall 0.4% conversion. With that average, it looks like “the site converts badly.” With the breakdown, you see the site converts reasonably well for people who arrive with intent, and the problem is the 80% of traffic that has none. Redesigning the site would fix nothing; changing where the visits come from would.

What to look at:

• Acquisition channel in GA4. Default channel groups depend on campaigns carrying UTM parameters; Google warns that without them “(not set)” values appear and traffic is categorized incorrectly. If you do not tag campaigns, your channel breakdown will be unreliable.
• Landing pages. Which pages do converters enter through? And non-converters?
• Search Console queries. What searches do you appear for? If they are very general questions unrelated to what you sell, the traffic is volume, not intent.
• Location and device. A local business with heavy traffic from another country, or a service site converting poorly only on mobile, points to different causes.

Verdict for this layer. If conversion by channel and landing page is reasonable on sources with commercial intent and poor on the rest, the problem is acquisition. The action is to redirect content or spend, not to touch the site. It is the same reasoning as in when digital advertising is the wrong answer.

Layer 3. Message: what the page says in the first seconds

If the traffic is right and still does not act, the next suspect is that people do not quickly understand what you offer. Nielsen Norman Group summarizes a Microsoft study (Liu et al., 2010) covering more than 205,000 pages and over two billion dwell times: users often leave pages within 10 to 20 seconds, and pages with a clear value proposition can hold them much longer. This is general behavioral evidence, not a number that applies to your page, but it supports a principle: the essentials must be understood immediately.

The five-second test. Show your home page (and your most important entry pages) for five seconds to five people from your market who do not know you. Ask: what does this company do? For whom? What would you do next? If answers diverge or stay vague, you have a message problem.

Signals in the data. Entry pages with many non-engaged sessions from good-intent channels; little scrolling toward the sections where the offer is; menu clicks toward basic information pages (who are they? what do they do?) instead of moving toward the action.

Frequent errors: describing the company instead of the customer’s problem; using the same generic copy on every page; hiding the specific service behind abstract categories.

Layer 4. Offer: the step you ask the visitor to take

Message and offer get confused often. The message says what you do. The offer is what the person gets if they take the next step, what they commit to, and what they risk. A page can be very clear and still ask for a disproportionate step. The decisions that come before the page (priority customer, position, offer) are developed in the four strategic decisions that come before channels.

Ask yourself:

• Is the next step proportionate to what the person knows about you? Asking for a “sales call” from someone who just arrived from an informational search is asking too much. A guide, a short assessment, a ballpark quote or an initial consultation are smaller steps.
• Is it concrete? “Contact us” does not explain what happens next. “Tell us about your case and we will reply within one business day with proposed next steps” does, provided you keep the promise.
• Are the terms visible? For a store: prices, shipping, returns. For a local service: coverage area, hours, how quotes work. For B2B: what the first conversation includes. Uncertainty about cost or process is a classic brake.
• Is there an option for someone not ready? A visitor who does not want to buy today should have something to do (subscribe, save, download), not just leave.

Verdict for this layer. If traffic is good, the message lands and visits reach the pages where the step is, but almost nobody takes it, the bottleneck is the offer or how it is presented.

Layer 5. Friction: what makes acting costly

Friction is anything that adds effort, doubt or error between intent and action. It is the layer most often reviewed first and, taken alone, the one that explains the least.

First, verify the basics work. The most profitable finding in any audit is often the simplest: the form does not submit, the notification email lands in spam, the WhatsApp button points to an old number, the consent banner covers the contact button on mobile. Send a test inquiry yourself through every channel, on mobile and desktop, and verify it reaches a real person.

Then measure where people drop off.

• form_start versus form_submit. If many people start a form and few submit, the form is the friction: too many fields, unnecessary questions, unclear validation errors.
• Mobile performance. If mobile conversion is far below desktop for the same traffic intent, review the mobile experience. Loading speed, visual stability and responsiveness have thresholds published by web.dev, which we develop in technical SEO and user experience.
• Process steps. In a purchase or booking, every extra screen is a chance to abandon.
• Session recordings and user tests. Watching a handful of real sessions often reveals problems no report shows: buttons that do not look like buttons, fields nobody understands, silent errors.

Verdict for this layer. If people try to act and cannot, or drop out midway, friction is a cause. If they never even try, the cause sits higher up.

Layer 6. Trust: reasons to believe

Even with the right traffic, a clear message, a reasonable offer and a smooth process, a person will not act if they doubt you will deliver. Google puts the idea in its helpful content guidance: of the E-E-A-T components, trust is the most important, and it asks whether information is presented in a way that makes you want to trust it, with clear sourcing and evidence of the expertise involved. How that credibility is built beyond the website is explained in what builds a brand’s credibility.

Trust is built with verifiable evidence, proportionate to the risk of the decision:

• Who you are. Real people, contact details, address or service area, legal information.
• What other customers experienced. Real reviews, with a way to verify them, such as a link to your Google profile. Never invented or purchased reviews. If your business depends on being found nearby, Google’s rules on profiles and reviews are in local business marketing.
• What happens if something goes wrong. Return policies, guarantees, clear terms.
• Proof you know what you are talking about. Bylined content, an explained method, work you can show with permission. More in E-E-A-T for mid-sized companies.
• Basic technical signals. A secure site with no visible errors or broken elements.

Verdict for this layer. If people arrive, understand and see the offer but do not move forward, and the questions reaching you through other channels (calls, messages) are about credentials, guarantees or other customers’ experiences, the problem is trust. The answer is never to fabricate social proof; it is to document and show what already exists.

Reading the layers together

The table summarizes what data to look at and what decision to make for each layer. Read it top to bottom.

Two methodological points.

This data points to hypotheses, not causes. A correlation in GA4 does not prove a page fails because of its message. When volume allows, a controlled A/B test is how you verify a change. When it does not, which is typical for smaller sites, combine quantitative data with qualitative observation: recorded sessions, conversations with customers who bought and with those who did not.

Problems tend to stack. It is common to find incomplete measurement, mixed traffic and a vague offer at once. Resolve them in order, layer by layer, and measure the effect of each change before the next. If you change everything at once, you will not know what worked.

What not to do with this diagnosis

• Redesign by default. It is the most expensive answer and the least likely to resolve layers such as traffic or offer. We develop this in why most B2B websites don’t generate leads; here we only add that a redesign should follow from the diagnosis, not replace it.
• Buy more traffic. If conversion by channel is poor, more of the same volume adds cost, not customers.
• Copy another site’s percentages. An “average conversion rate” means nothing without industry, channel, intent and value per customer.
• Fabricate social proof. Invented reviews, logos of non-clients or unsupported figures not only violate platform rules; they destroy the trust you are trying to build.
• Optimize without a measurable hypothesis. Every change should answer to a layer and a data point you expect to move.

Next step

If you want a layer-by-layer diagnosis of your site, using your real GA4 and Search Console data, it is part of our conversion optimization and marketing analytics work. You can also read our CRO methodology. To talk about your case, get in touch.

Sources verified as of October 9, 2026. The Nielsen Norman Group piece summarizes 2010 data: it is cited as general behavioral evidence, not as a threshold.

Sources

• Google Analytics Help, “Key events”: support.google.com/…/9267568
• Google Analytics Help, “Recommended events: generate_lead”: support.google.com/…/9267735
• Google Analytics Help, “Enhanced measurement events” (includes form_start and form_submit): support.google.com/…/9216061
• Google Analytics Help, “Engagement metrics: engaged sessions, engagement rate and bounce rate”: support.google.com/…/12195621
• Google Analytics Help, “Traffic sources and UTM parameters”: support.google.com/…/10917952
• Google Search Central, “Creating helpful, reliable, people-first content”: developers.google.com/…/creating-helpful-content
• Nielsen Norman Group, “How Long Do Users Stay on Web Pages?” (2011), summarizing Liu, Chao et al. (Microsoft, 2010): nngroup.com/…/how-long-do-users-stay-on-we…