People do not abandon forms. They abandon fields.
That distinction is the whole game. "Our form has a 40% completion rate" is a fact you can't act on. "Everyone stops at the phone number" is a fix. Yet almost all form advice operates at the level of the form — shorten it, split it into steps, add a progress bar — as though the form were failing uniformly along its length. It isn't. There's a place where the typing stops.
The five kinds of field that lose people
The expensive field. Cost isn't measured in keystrokes; it's measured in what the visitor believes it will bring about. A phone number is expensive because it means a sales call. A budget range is expensive because it means being qualified out. A company size is expensive for the same reason. These fields are cheap to type and costly to answer, and a form can die on a field that takes two seconds to fill.
The unexplained field. Any input whose purpose isn't obvious gets filled with a worst-case assumption. "Why do they want my date of birth?" has no good answer available to the visitor, so they supply one. A single line of microcopy beside the field — "so we can size your plan" — converts a suspicion into a reason.
The hostile field. It rejects a valid phone format. It won't accept an apartment number. It demands a postcode in a shape your country doesn't use. It clears itself when you make a mistake. The visitor isn't deciding whether to continue; they're being prevented from continuing, and they blame you rather than the field.
The premature field. Correct field, wrong moment. Asking for a mailing address before someone has built anything, or a password before they've seen value, isn't a data problem — it's a sequencing problem. The same field, later, is nearly free.
The mobile-hostile field. A numeric input that opens an alphabetic keyboard. A dropdown of two hundred options on a phone. An input positioned so the on-screen keyboard covers it. Autofill that never fires because the inputs lack `autocomplete` attributes. Each one is a physical obstacle, not a psychological one, and it explains most mobile-versus-desktop completion gaps.
Validation is a conversation, and most forms are rude
Validation is where forms lose people who were actively trying to give you their information, which makes it the most wasteful failure of all.
Validate on blur, not on submit. Telling someone at the end that something twelve fields up was wrong is asking them to re-audit their own work.
Name the fix, not the failure. "Invalid input" describes your state. "Phone numbers need 10 digits" describes theirs.
Never clear the field. Deleting what someone typed as punishment for a formatting error is the single most reliable way to end a session.
Put the message next to the field, visible without scrolling, and don't rely on colour alone to signal it.
And — a real defect I've seen live on production forms — check that each field shows its own message. A street-address input displaying "You must enter a group name" is a bug that quietly destroys trust at the exact moment someone was cooperating.
Finding your field
Look at where the typing stops, not where the page ends. If you can see which input was last touched in abandoned sessions, you have your answer directly.
If you can't, use the ordering trick. Fields fail in position, so move a suspect field to the end of the form and see whether completions rise. If they do, that field was the wall.
Compare mobile and desktop for the same form. A large gap isolates the mobile-hostile class of problem and rules out the others, because the wording and the ask were identical.
Count the fields that change what happens next. Every field that doesn't affect fulfilment, pricing, or the immediate response is a field you're charging for nothing. Delete rather than shorten: a five-field form without a phone number often outperforms a three-field form with one.
Multi-step forms don't reduce friction — they relocate it
Splitting a long form into steps can help, because each screen looks achievable and progress is visible. It can also hurt, because now there are three moments to quit instead of one, and the visitor can't see what's coming.
Multi-step works when the early steps are cheap and the expensive asks arrive after the person has invested effort and can see the end. It fails when step one contains the expensive field — you've added ceremony to the same wall.
If you split a form, show progress honestly ("Step 1 of 3"), and never let a later step surprise someone with a cost the first step implied wasn't there.
What structure can't tell you
Everything above can be reasoned about by reading your form. What can't be reasoned about is which of your fields is your wall — because that depends on your audience, your wording, and what they think each field means.
Two forms with identical structure lose people at different inputs. Defrixa's free scan reads your form's structural friction — its weight, its ordering, its clarity — and names the biggest issue. With the snippet installed, it watches which field people focus, hesitate on, error at, and leave, and confirms whether removing it brought them back.
Common questions
No. A short form containing one expensive field will lose to a longer form of cheap ones. Audit fields by what they cost the visitor to answer, not by how long they take to type.
Yes — and mark them with more than colour, so the requirement is legible to everyone. Better still, make optional fields obviously optional, or remove them.
They help when the progress is real and the end is visible. A progress bar on a form that reveals new steps as you go is worse than none, because it made a promise.
Usually keyboards, tap targets, autofill, or a validation message rendered off-screen. Test on a real device before concluding mobile visitors are less interested.