Usually because the form asks for more than the moment deserves. Every extra field is another decision, and decisions are where people quit. The fix is rarely a nicer looking form. It is a shorter one, with clearer labels and errors that appear where the mistake happened.
Forms are where all your marketing work either converts or evaporates. You can win the ranking, write the perfect page, and design a beautiful hero, and still lose the lead because someone hit a validation error they could not understand on a phone.
What follows is the research we actually design against, plus the rules we apply on client builds. Most of the numbers come from Baymard Institute, whose checkout research is the best public dataset on form behavior, and from the W3C's accessibility guidelines.
Fewer than you have now. Baymard Institute's ecommerce benchmark found that the average checkout in 2024 contained 11.3 form fields spread across 5.1 steps, while its own research concludes that most sites need only 8 fields in total to complete the job. The gap between those two numbers is pure friction.
The more useful finding is about what matters. Baymard states that the number of form fields in a checkout affects overall usability far more than the number of steps, and that UX performance falls as field count rises while step count has relatively minor impact. So splitting a long form into more screens does not fix it. Removing fields does.
We apply this to lead forms by asking a blunt question about every field. Will someone read this answer before the first reply? If the answer is no, the field is collecting data for a report nobody opens, and it is costing you leads to do it.
Baymard's benchmark rests on testing across more than 330 top ecommerce sites and, by its own account, over 200,000 hours of usability research. That is a stronger basis than most opinions about form length, including ours.
One column, with one narrow exception. Baymard's usability testing found that single-column layouts produced fewer skipped fields, fewer misinterpreted fields, and fewer errors overall than multi-column layouts, because a single column guides the eye in one direction and never leaves the reading order in doubt.
The failure mode with two columns is ambiguity about sequence. Baymard found that multiple columns pull attention in different directions, making it harder for people to work out which fields are required, and that users frequently skipped required fields or completed unnecessary ones. Despite this, it reports that 16% of sites in its benchmark still use extensive multi-column checkout forms.
The exception is worth keeping. Baymard accepts two or three inputs on a single line when they are highly associated and read as one thing, like a postcode and city, or a card expiry and security code. The rule is that the form stays predominantly single column.
Designers push back on this because a single column looks plainer in a mockup. It is one of the clearest cases we know where the tidier design and the better performing design are not the same design.
With a visible label sitting above the field, and keep it visible while the person types. Placeholder-only labels look clean in a mockup and then fail as soon as someone starts typing, because the instruction disappears at exactly the moment they might want to check what the field was asking for.
This is also an accessibility requirement rather than a preference. WCAG success criterion 3.3.2, Labels or Instructions, sits at Level A, the most basic conformance level, and requires that labels or instructions are provided when content requires user input. It also notes that labels or instructions may specify data formats.
Format instructions prevent a whole class of errors. If a field wants a date in a particular order, or a phone number with a country code, say so before the person guesses. An error message is a worse place to teach someone your format than a hint under the label.
Keep the label wording literal too. "Work email" is better than "Email", and "Company website" is better than "URL", because the extra word removes a decision. We cover the wider conversion picture in our notes on UX and conversion design.
Mark both, explicitly. This is the single most neglected detail in form design, and Baymard's data on it is stark. It found only 14% of sites explicitly mark both required and optional fields, a figure that stayed roughly consistent from 2016 through 2021.
The consequence is measurable. Baymard reports that when only optional fields were marked, 32% of users hit a validation error because they had not completed a required field. That error is entirely avoidable, and it happens at the worst possible moment, which is after the person has decided to give you their details.
Our approach is to keep almost every field required and label the rare optional one as optional in plain text. If a form is short enough, ambiguity mostly disappears, which is another argument for cutting fields rather than annotating them.
Next to the field that caused them, in words that say what to do. A summary at the top of the form tells someone they failed without telling them where. Inline messages tie the problem to the input, which is the only place the person can fix it.
Validate at the right time as well. Checking a field when someone leaves it is helpful. Checking it aggressively while they are still typing produces red warnings on a half-finished email address, which reads as being told off for typing too slowly.
Write errors as instructions rather than diagnoses. "Enter a date as DD/MM/YYYY" gives someone a next action. "Invalid input" gives them a puzzle. And never clear the fields they already filled in, which still happens on older builds and is the fastest way to lose a submission permanently.
Assume most of your submissions come from a phone and design for thumbs first. Use the correct input types so the right keyboard appears, enable autocomplete so the browser can fill known details, and give inputs enough height to tap without precision.
Baymard's research shows how much worse the mobile penalty is. It found that on mobile sites lacking clear field indicators, 75% of users during testing experienced severe form usability issues. A form that is merely annoying on desktop can be genuinely unusable on a phone.
The technical details are small and cheap. An email input brings up an email keyboard, a numeric input mode brings up digits, and autocomplete attributes let the browser fill name, email, and address in one tap. Skipping these is not a design decision, it is an oversight, and we cover the broader discipline in our piece on mobile-first web design.
Test on a real device with one hand. Anything you cannot complete while standing up, holding a phone, on a mediocre connection, is not finished.
Sometimes, but not to hide length. Baymard's finding that field count matters far more than step count means splitting eleven fields across four screens does not make the form easier. It just makes the length harder to judge before starting.
Where multi-step genuinely helps is when the questions branch. If answering one question changes which questions come next, steps prevent people from reading fields that do not apply to them. That is a real simplification rather than a cosmetic one.
If you do use steps, show honest progress. A visible indicator of how many steps remain lets someone decide to continue, and a progress bar that lies about how close they are to the end is worse than no progress bar at all.
Our default for lead generation is a single short form. We reach for multi-step only when the flow needs qualification logic, and we build it so a person can move backwards without losing what they typed.
Something immediate and specific. Confirm the submission landed, say what happens next, and give a rough timeframe. A generic "thank you" leaves people wondering whether it worked, and some of them will submit again or email you separately to check.
Then make sure the submission actually reaches a human. We test every form end to end before launch, including notification emails, spam filtering, and any automation wired through tools like Zapier, Make, or Airtable. A form that silently drops leads is worse than a broken page, because nothing looks wrong from the outside.
Set expectations you can meet. We tell clients to state a response window on the confirmation, and we hold ourselves to responding within 48 hours. Our practical notes on the Webflow side of this are in our guide to Webflow forms, including where submissions go and how to route them.
Be careful with spam protection. Aggressive challenges block real people alongside bots, and if you must use one, put it where it cannot stop a legitimate submission from completing.
Delete fields. Open your main form, count the fields, and remove every one that nobody reads before the first reply. Then move labels above the inputs, mark required and optional fields explicitly, and move error messages inline. Those four changes cost very little and address most of what the research above describes.
After that, test it on your own phone with one hand and fix whatever annoys you. Then check that the submission arrives, that the notification is readable, and that the confirmation message tells people what happens next.
Two of our published project results are a 230% lift in sign-ups on Ajust and a 185% lift in demo signups on Dam Secure. Outcomes like that are never one change, but the form is always on the list, because it is the last thing standing between interest and a lead. If you want us to look at yours and tell you which fields to cut, we are happy to do that. Let's talk, over at phoenix.studio.
Tell us where you want to go. We'll tell you how we'd get you there.