Somebody describes a site to an agent, the agent publishes it, and there is a form on the contact section. Does anything actually arrive?
We can answer that from our own data rather than from opinion. Across published projects, 29.5% carry a form at all. Among the 707 that do, 10.84% of form views become a submission and 5.74% of visits become a lead. The median project with a form has collected four leads. And 19.4% have live traffic and nothing at all.
Those last two numbers are the interesting ones, and they say opposite things.
Most published sites have no form
Seven out of ten projects never declare one. That is worth sitting with before any conversion talk, because it says what people mostly publish: a page that exists to be read and linked, not a page that exists to capture something.
A portfolio, a landing page for an event, a one-off explainer, a demo sent to one client. None of those need a form, and adding one would not improve them.
So the population below is self-selected. These are the people who decided they wanted an inbox, which is exactly the group whose conversion is worth measuring.
What the forms look like
| Share of the 707 forms | |
|---|---|
| Median number of fields | 4 |
Carry a <label> | 76.8% |
Have at least one required field | 68.5% |
Four fields is the median, and that matches what the forms are for. Name, email, one message box, and one thing specific to the job: a date, a budget band, a phone number.
The label figure is the one we did not expect. When a form is written by an agent rather than assembled from a widget, the markup comes out with labels attached most of the time, because the model writes the pattern it has seen most often in training rather than the shortest thing that renders.
Worth one caveat before anyone quotes that as a win over the open web. Published surveys of the web measure a different unit: HTTP Archive's 2025 accessibility chapter reports the share of input fields whose accessible name comes from a label, around 34% on desktop. Ours is the share of forms with a label present. The two are not comparable, and the honest statement is the narrow one: on these sites labels are usual rather than exceptional.
What actually converts
10.84% of form views end in a submission. 5.74% of visits end in a lead.
The gap between those two is the funnel in miniature. Roughly half the people who reach a page with a form never scroll to it or never engage with it, and of those who do see it, about one in nine fills it in.
We are not going to put those next to a benchmark from commercial checkout funnels, and that is deliberate. Those forms live on sites where the visitor arrived with intent and often with a card in hand. A single page for a ceramics studio is a different animal, and a comparison would tell you something about the difference in populations rather than about your form.
The median project with a form gets four leads
Median four. The 90th percentile is 33.
That distribution is the shape of the whole thing. A few projects collect a real stream, most collect a handful, and a handful is often the correct outcome. Four enquiries for a made-to-order ceramics studio that fires six pieces a week is not a disappointing number, it is a full order book.
The mistake is reading a small number as failure. The right question is not "how many" but "how many did this page need to be worth publishing", and for a lot of these pages the answer is one.
Traffic and nothing arriving
19.4% of projects have live traffic and zero leads. One in five.
The tempting conclusion is that their forms are broken. We cannot claim that and neither should you: nothing in this data tells us why a form stayed empty, and there are several honest reasons it would.
The visitors were not buyers. A link shared in a community brings readers, not clients. The page asked for the wrong thing at the wrong moment, like a project enquiry from somebody who came to look at pictures. Or the offer needs a conversation the form cannot start, and the visitor emailed instead, which counts as a success nowhere in this dataset.
What is true is narrower and more useful: a form with traffic and no submissions is a signal worth investigating, not a verdict. Send yourself one submission on the live address and read what arrives. If your own test lands intact, the form works and the question moves to the page and the audience.
A sixth of what arrives is spam
17.0% of submissions are spam. Every lead number above is counted after that filter, so the conversion rates are real enquiries rather than bots.
That share is worth knowing before you add a form, because it sets your expectation for the inbox rather than for the statistics. One in six of the things that ping you will be nothing. On a page collecting four real enquiries, that is not a problem. On one collecting thirty, it is a habit you develop.
Every number here comes from projects published through one host, so it describes what happens on small, mostly single-page sites written by an agent. It is not a survey of the web.
It also says nothing about cause. We can see that a form has labels and that it collects, and we cannot show from this that the first produced the second.
What we would take from our own numbers
If you are publishing a page and wondering whether to put a form on it, the data supports three plain conclusions.
A form is worth it when a single enquiry pays for the page. Given a median of four, that is the bar to check against, not a percentage.
Four fields is enough. The median form asks for four things and converts at about one in nine views, and nothing in this data suggests that asking for more helps.
And send yourself one submission after publishing. The one in five with traffic and nothing is the group where a five-second check would separate "quiet page" from "field the agent never declared", and only one of those is worth your afternoon.
Publish the page, read what comes in
Your agent declares the form and publishes the page. The enquiries land in your panel.