Sales funnel: the real leak is usually somewhere else

What a funnel is, and what it is not

A sales funnel is a name for the route somebody travels from never having heard of you to buying: seeing, wanting, comparing, deciding. The name is useful for planning and close to useless for diagnosis, because it does not tell you where the drop between two stages came from.

Real diagnosis asks three simpler questions. Is the traffic the right traffic? Does the page open fast enough for anyone to stay? Is the thing you want them to do easy to do? If the answer to the second is no, neither of the other two matters.

Numbers taken from those same pages

We opened three of the top results in Chrome. One made 59 network requests, moved 3 MB, recorded an LCP of 2,048 ms, and sent 32 requests to other domains. Our own home page in the same test made 10 requests, moved 286 KB and recorded an LCP of 248 ms.

Two of the three did not complete loading within thirty seconds. One of them returned just 24 KB to our automated browser, which is consistent with a bot protection layer sitting in front of it. That is a choice rather than a defect, and we are not calling it one; it simply means we could not measure that page, and we are saying so.

Those 32 third-party requests are something we see on most projects and nobody budgets for. Each one is a fresh connection, a new opportunity to be slow, and a dependency on a service you do not control. On a page whose job is to get a form filled in, that number should be a single digit.

Where the leak actually is

In the work we have done, three points account for most of it.

None of them concerns the funnel as a concept.

First, how fast the landing page opens. A visitor arriving from an ad has less patience than an organic one, because they have not yet committed to you in any way. Second, the length of the form: every extra field costs a percentage of the people who would have completed it, and most forms we see carry two or three fields nobody ever uses. Third, an unclear next step, meaning a page offering three buttons of equal importance.

All three are measurable and all three are fixable in an afternoon. the loading and stability metrics turns the first into a number, and the other two only require counting.

The right order of work

Start by checking that the traffic arriving is your audience at all. If the phrase you rank for, or buy ads against, belongs to somebody with no purchase intent yet, no amount of page work will rescue it. That is what search work is for: the quality of the traffic, not its volume.

Then make the page fast.

Then shorten the form. Only after those three should you reach for the things usually suggested first: headline A/B tests, button colour changes, adding a video. On a page that does not open in two seconds, those produce measurement noise rather than improvement.

When the funnel really is the problem

One limit worth stating honestly: the page is not always at fault. If your traffic is precisely your target audience, the page is fast, the form has two fields, and still nobody converts, then the problem genuinely is the offer. What you sell, at the price you name, does not appeal to this person.

In that situation no technical work helps, and we have nothing to sell you except a sentence: change the offer, not the page. Recognising that case is straightforward too, because it only becomes meaningful once the previous three are fixed.

What are the stages?

Usually awareness, interest, consideration and decision. The labels differ between sources and it hardly matters; what matters is knowing which page on your site answers each stage.

How does a marketing funnel differ from a sales funnel?

The boundary is a convention. Typically the marketing funnel runs up to lead generation and the sales funnel takes it from there. In a small business that separation tends to generate meetings rather than clarity.

What is a good conversion rate?

There is no credible global figure, and anyone handing you a specific percentage is quoting an average from a different industry. Take your own baseline and measure improvement against that.

How do I find the leak?

Start with landing page speed, because it is the quickest thing to measure, then count the fields in the form, then count how many equally weighted exits the page offers. Speed work usually pays off before any other test.

If the leak persists after those three, your landing page was probably not structured to sell in the first place. Building a page that asks the visitor for one specific thing is different from adding a form to an existing page, and digital marketing without that page is buying traffic for somewhere not ready to receive it.

Method: Google organic results for the Persian query, depth 20, read on 16 August 2026 from a server in Helsinki; 16 URLs answered 200, and server response time, HTML weight and tag counts were taken from those responses. Three URLs were then opened separately in headless Chrome with an empty cache. The server is not in Iran, so do not read the timings as an Iranian visitor's experience; request and byte counts are location-independent.

The responsiveness metric is defined on the INP page at web.dev, and form guidance is in its payment and address form best practices.