"The form doesn't work." That is the sentence I hear most often when a client calls me about a conversion problem. In most cases the form works perfectly well. It sends exactly what it has been asked to send. The problem is what it is being asked.
One quote request form I took over had eleven fields: title, first name, last name, company, job title, email, phone, postcode, town, estimated budget, message. Nobody in the company remembered deciding on that. The fields had piled up, one per department, over three years.
A form does not collect information. It asks for trust, one field at a time. And every field carries a price nobody bothers to work out.
1. The only calculation that matters
A field is a trade-off. It earns you information and it costs you enquiries. Until you put both sides on the scales, you are arguing about taste.
Baymard Institute has been measuring this on checkout flows for more than ten years, and its conclusion fits in one line: what weighs on the experience is not the number of steps, it is the number of fields. Its 2024 benchmark gives an average of 11.3 fields, when most sites could complete an order with eight, according to its study on minimising checkout form fields.
A contact form is not a checkout. The mechanics are identical though: with every extra field, a share of your visitors recalculates what writing to you is going to cost them.
My method is simple and slightly brutal. Start from zero fields, and add one only when you can say what you will do with it in the hour after the enquiry lands. Not "it might come in handy". What you actually do with it, when you read the message. A field you use once a quarter costs you all year.
2. The fields I remove first
Always the same ones, in this order.
Title. It exists to personalise an acknowledgement email. That is a steep price for a "Ms" that will be wrong one time in twenty anyway.
Separate first and last name. People think of their name as one block. In Baymard's testing, 42% of participants typed their full name into the "First name" field at least once, and 89% of the sites in its benchmark still keep two separate fields. A single "Full name" field removes the problem and gives you the same thing.
Postcode and town. You get them from the signature on the reply, from the company registry, or you ask on the call. Unless you bill for travel, they have no place in a first contact.
Job title. It only earns its keep if you route enquiries automatically in your CRM. If you read every enquiry by hand, the message tells you who is writing better than any dropdown.
One exception, and I hold on to it: the "How did you hear about us?" field. I keep it, optional, last. Now that measurement tools have lost part of their visibility, it is the only self-declared signal that survives tracking blocks, and it recovers channels that analytics files under "direct". I went through that measurement gap in what GA4 can no longer see. Optional, it costs almost nothing. Required, it would scare people off.
3. The phone field is the most expensive on the form
A phone number is not a piece of information. It is permission to call. Those are not the same thing at all, and visitors know it perfectly well.
Baymard reports that 39% of sites require it with no explanation whatsoever, and observes the result in testing: some participants enter a fake number, others leave. Its recommendation is to explain plainly why the phone field is required, or to make it optional.
I go one step further. A vague line such as "so we can serve you better" reassures nobody, because it says nothing. What works is a sentence that commits you: "I call you back within one working day, never for cold sales." You state the deadline and you close the door on everything else. If you cannot honour that sentence, drop the field.
4. The message field is the only one that really sorts
This is the field teams want to cut "to shorten the form". It is precisely the one to keep.
A form with no free-text box gives you interchangeable enquiries. Ten identical lines land in your inbox and you have no idea which to open first. With three hand-written sentences you know in ten seconds who has a project and who is collecting quotes.
On the lead generation programme I ran for DecisionIA, a training organisation, acquisition cost on Meta came down to €3.73, against €7.71 cost per action on Google. Those numbers only meant something once the first sign-ups had been checked by hand: only a fraction matched the target. The form was not at fault, it was doing its job. But with no material to sort on, cost per lead stayed a flattering number with no use. I have drawn the same lesson everywhere since: a form with no sorting field hands you an attractive cost per lead and a real cost per customer you will discover three months later.
The label does half the work. "Message" invites "Hello, I would like some information". "Describe your project in a few lines: what you sell, what is blocking you today" invites an answer you can act on. Same field, entirely different yield.
5. Execution counts as much as content
You can pick the right fields and still lose enquiries on the build. Four things I check every time.
A label above the field, permanently visible. Replacing the label with text inside the field saves space and loses submissions: the hint vanishes on the first keystroke, exactly when the user needs it. Nielsen Norman Group documents this in detail in Placeholders in Form Fields Are Harmful. The label goes above and is tied to the input with the for attribute, as described in the W3C Labeling Controls tutorial.
The autocomplete attribute. Setting autocomplete="email", autocomplete="tel", autocomplete="name" lets the browser fill the form in one gesture. It is also an accessibility criterion, covered by Web Content Accessibility Guidelines (WCAG) technique H98. Two minutes of markup, fields filled without effort.
Errors next to the offending field, never at the top of the page. And above all, never wipe the form. Losing your message when you click "Send" means losing the visitor with it.
One column. Two fields side by side break the vertical reading order and behave badly on mobile, where most of the traffic sits. I mapped the mobile leak points in the mobile funnel.
| Field | What it costs you | What it earns you | Where I land |
|---|---|---|---|
| Title | Friction for no gain | A polite salutation | Removed |
| Full name | Almost nothing | Essential | Kept, as a single field |
| Nothing, it is expected | The reply channel | Kept, required | |
| Phone | The most expensive on the form | A shorter sales cycle | Optional, or justified in one line |
| Company | Moderate in B2B, heavy in B2C | A start of qualification | Kept in B2B only |
| Free-text message | Effort asked of the visitor | Sorting, prioritising, tone | Kept, required, precise label |
| Budget | High, it gets in the way before the conversation | A crude filter | Removed from first contact |
| How you heard of us | Low if optional | The only self-declared channel signal | Kept, optional, last |
6. Required, optional, or not at all
The compliance question and the conversion question meet here, which does not happen often. The French data protection authority reminds organisations to collect only data that is adequate, relevant and necessary for the stated purpose. In other words: if you can reach your objective without that piece of data, do not ask for it.
It is the same test as in section one, written by the regulator. A minimised form is more compliant and it converts better. Take the win, they are rare.
Two implementation rules I apply everywhere. First, mark explicitly what is required and what is optional, rather than leaving the visitor to work it out. Baymard finds that only 14% of sites in its benchmark mark both. Second, the newsletter opt-in stays separate from the form submission, unticked, with its own wording. A contact captured by surprise costs more in unsubscribes than it earns in opens.
7. What happens after the click
The form does not stop at the button. Three details I see neglected almost everywhere.
The button label states the action, not the mechanism: "Get my audit", not "Submit". The confirmation page gives a stated deadline, because doubt starts the second the page changes. And the conversion event fires on the actual server response, not on the click. Otherwise you are counting submissions that failed, optimising campaigns on thin air, and you will never know. That is where conversion work meets measurement, and it is almost always where the chain breaks.
To find the field that stalls people on your own site, session recordings are the fastest tool. You watch the cursor enter a field, hesitate, back out. I described the method in heatmaps and session recordings.
In the crucible: what is left once the fields are gone
The natural reflex in front of a form that converts badly is to add. One field to qualify better, one for the sales team, one for reporting. Every addition starts from a good intention and melts the volume a little further.
The work runs the other way. You remove, you remove again, and you look at what settles at the bottom. Usually: a name, an email, a reason for writing. Three fields that weigh more than the original eleven, because they ask only for what the visitor is willing to give at that stage. The rest comes out in the conversation that follows, and the conversation only starts if the form gets sent.