Your traffic report looks fine. Sessions are up, rankings held, the ad account is spending. The lead count has not moved.

That gap has two possible explanations and they call for opposite fixes. Either the traffic is wrong, in which case no amount of button color testing will help, or the traffic is fine and something after the click is losing people. Most advice on this problem assumes the second explanation and then only examines the first half of it, stopping at the landing page as though a lead becomes a customer the moment the form clears.

It does not. Between a search query and a closed deal there are seven stages, and a lead can die at any of them. The stages after the form are the ones almost nobody audits, which is exactly why they are where we find the biggest leaks. Here is the whole path, what failure looks like at each point, and how to tell which one is costing you.

First, separate a traffic problem from a conversion problem

Before anything else, work out which half of the funnel is at fault. Pull impressions, clicks, and average position from Search Console for the last twelve months, then put form submissions and calls next to them.

If impressions and clicks are steady while conversions fall, the traffic is arriving and something downstream is failing. Start at stage two.

If impressions hold while clicks fall, you may not have a site problem at all. Ahrefs re-ran its study of 300,000 keywords using Search Console data and found that the presence of an AI Overview now correlates with a 58% lower clickthrough rate for the top-ranking page, up from 34.5% in the earlier version of the same analysis. Your rank did not move. The clicks did.

If clicks are up and conversions are flat, look hard at what changed in your keyword mix. Growth in informational traffic looks identical to growth in buying traffic on a line chart and converts nothing like it, which is the whole argument behind judging traffic by intent rather than volume.

Stage one: the queries you rank for

Rankings are not the goal. Rankings for queries that precede a purchase are the goal, and the two get conflated constantly because one is easy to report.

A page that ranks first for a definition question will pull traffic and produce nothing, and that is not a failure of the page. It is a failure of expectation. Someone asking what a term means is not ready to hire anyone, and treating that visit as a lost conversion sends you off optimizing a page that is doing its job.

The test is simple. Sort your landing pages by sessions, then look at the conversion rate of each. If your highest-traffic pages convert near zero and your commercial pages convert fine but get little traffic, you do not have a conversion problem. You have a visibility problem on the pages that matter, which is a different project entirely. We go deeper on sorting commercial queries from informational ones separately.

Stage two: the page they land on

Assume the traffic is right. The next question is whether the page answers the thought the visitor arrived with.

Message match is the whole game here. Someone who clicked an ad for emergency repair and lands on a general services page has to re-orient before they can act, and a percentage of them will not bother. The mismatch is usually invisible to the person who built the page, because they know what the business does and cannot un-know it.

Paid traffic suffers worst, since the promise in the ad is specific and the page it points at often is not. That pattern is common enough that we wrote up why paid traffic frequently converts worse than organic on the same site. For everything else, the usual suspects are well documented, and most of them show up in our list of conversion problems that survive good traffic.

Stage three: the form

The form is where intent turns into a record, and it is the most commonly broken thing on any site we audit.

Field count is the obvious lever and the least interesting one. The failures that actually cost money are quieter. Validation errors that appear below the fold on mobile, so the visitor taps submit, sees nothing happen, and leaves. Notification emails routing to an address nobody monitors. Spam filtering tuned aggressively enough that real submissions disappear alongside the junk.

Test your own form from a phone, on cellular data, using an email address you can check. Do it monthly. We have watched a site go six weeks with a silently broken form while everyone upstream congratulated themselves on the traffic. The full checklist lives in our breakdown of the ways a working form still loses submissions.

Stage four: the handoff

Here is where most audits stop and most leaks start.

A submission is not a lead until a person has it. In between sit routing rules, assignment logic, enrichment, and notification, and every one of them adds minutes. When we time this stage for clients, the delay is rarely the salesperson. It is the gap between the form clearing and anyone being aware a form cleared.

Then there is coverage. A lead arriving at 4:50pm on Friday enters a queue that opens Monday, which is roughly sixty-four hours of a competitor having the same person's attention. Ask yourself who specifically owns the first reply, what the standard is in writing, and what happens when that person is on vacation. If the answer involves the word "we" rather than a name, the answer is nobody, which is the pattern we describe in what happens when follow-up has no owner.

Stage five: the response

The research on response time is old, consistent, and widely ignored.

The Lead Response Management study conducted by Dr. James Oldroyd with InsideSales found that the odds of qualifying a lead are 21 times greater when the first contact attempt comes within five minutes rather than thirty, and the odds of making contact at all are about 100 times greater. Separately, Oldroyd, McElheran, and Elkington audited 2,241 US companies for Harvard Business Review by submitting test leads and found that 23% never responded at all, and the average response among those who did was 42 hours.

Those two studies get blended together across the internet, so to be precise: the 21x and 100x multipliers come from the InsideSales and MIT research, and the 42-hour average comes from the Harvard audit. Both point the same direction.

The number that matters is not the industry average. It is yours, measured from the submit timestamp to the first human contact, reported as a median and a worst case. Almost nobody has it. Measuring your own response time and fixing what the number exposes is usually the highest-return hour in this entire list, and for teams that cannot staff the coverage internally, our pipeline response team answers and qualifies every inbound lead as an alternative to hiring.

Stage six: qualification and nurture

Speed gets a lead into a conversation. What happens in that conversation decides whether the lead was worth generating.

Two failures dominate. The first is qualifying live, with no criteria agreed in advance, which turns every first call into an interrogation and leaves the definition of a good lead different for every person on the team. The second is treating interested but not ready as the same thing as unqualified. Someone who is buying in six months is not a bad lead. They are a lead with a date on it, and dropping them means paying to generate that person again later.

Write the criteria down before the next call. Three or four factors is plenty. Then decide what happens to everyone who does not meet them today, because the default is that nothing happens and the record goes quiet.

Stage seven: what you count as a win

The last stage is the one that quietly corrupts all the others.

If your ad platforms optimize toward form fills, they will find you more form fills. Not more customers. The algorithm honors the goal you gave it, and if that goal stops at the form, it will happily spend your budget on the audience most willing to fill one out.

Connecting closed-won data back to the campaign that sourced it changes what the platforms chase, which is the difference between a cheaper cost per lead and a cheaper cost per customer. Those two numbers frequently move in opposite directions. Start by setting up GA4 to track revenue rather than form submissions, then look at what it takes to measure the stages after the form and how to feed lead quality back into the campaigns generating it.

One more wrinkle worth planning for. Visitors arriving from ChatGPT, Perplexity, and AI Overviews behave differently from search visitors, usually landing further along in their decision and converting on different pages, which means AI-referred traffic needs its own measurement approach.

Run the diagnostic on your own funnel

Work down this table in order. The first row where your answer is bad is where to spend the next month.

Where to start

You cannot fix seven stages at once, and trying produces a quarter of activity with nothing to show for it. Find the largest single drop and work there until the number moves.

In our experience the biggest and cheapest wins sit in stages three through five, because they require no new traffic, no new content, and no new budget. A form that works, a named owner, and a response standard will often outperform a quarter of ranking improvements, which is the pattern behind our conversion-focused work for an environmental services client.

If you have read this far and cannot confidently answer the diagnostic table for your own business, that is the actual finding. XRAY is a monthly report that puts a full day of our team inside your marketing and hands back the wins, losses, gaps, and what to fix first, for $200 a month with no contract. If you would rather start with the channels themselves, our free audit of your search and paid performance is a reasonable first step.

Either way, stop optimizing the top of the funnel until you know the bottom of it holds.