How to reduce form abandonment: what the research shows

To reduce form abandonment, first measure which question people leave on, then cut what they see at the start: fewer fields, one question at a time, no sign-in or phone number before they've committed. Let them ask what a question means. Keep the answers of anyone who stops, and remind them with wording that changes each time.

That's the whole method. The rest of this post is why each step works, where it comes from, and how it goes wrong.

One warning first. Search this topic and you'll find confident numbers everywhere: "conversational forms convert 40% better", "inline validation cuts abandonment by 22%". Follow most of them back two or three links and they end at a vendor's marketing page with no study behind it. (We traced the most popular ones on form statistics, traced to their sources.) I've kept every claim here to a primary source or to data from a real form, and said so when the evidence is thinner than I'd like.

1. Find the question people leave on

You can't fix abandonment in general. You can fix question seven.

Most form tools report one completion rate for the whole form. That number tells you that you have a problem, not where. What you want is drop-off per question: of the people who saw this question, how many answered it and carried on? One question nearly always stands out, and it's rarely the one you'd guess.

Separate two kinds of leaving while you're at it: people who never answered anything, and people who answered some questions and stopped. They have different causes.

We saw this clearly on a university hackathon registration that ran on chatform this September. 129 students started a registration and 55 finished, 43%. But roughly three in four sessions ended before a single answer, on a form that asked people to sign in before question one. (Sessions include reloads and repeat visits, so treat that as a direction, not a rate.) The biggest loss wasn't in the form at all. It was the door.

Where this goes wrong: optimising the wording of question nine when most people never reached question one.

2. Cut what they see before they start

People decide whether to fill in a form by looking at it. A page of twenty fields announces twenty fields.

Baymard Institute's shopper survey found that 17% of US online shoppers have abandoned an order because the checkout was too long or too complicated. (You'll still see 22% quoted everywhere; that's Baymard's older survey round.) That's for checkouts, where people already want the thing. An enquiry form has less goodwill to spend.

Two ways to spend less of it:

• Delete fields. For every field, ask what decision its answer changes. If the honest answer is "none, but it's nice to have", remove it or ask it later, by email, to the people who became customers.
• Show one question at a time. The total is the same, but the decision to continue is made in small pieces instead of once, at the top of the page. This is the core idea behind conversational forms, and the reason even the simplest one-question-per-screen builders help.

Where this goes wrong: hiding length rather than reducing it. A one-at-a-time form with forty questions still has forty questions, and people notice around question twelve. Keep a progress indicator so nobody feels tricked.

3. Move the expensive questions later

Some questions cost the respondent more than others: a phone number, a budget, a sign-in, an upload. Asked first, they're a toll at the entrance. Asked after a few easy questions, they land on someone who has already invested something.

There's decent evidence that a task which feels already begun gets finished more often. In a well-known field experiment, Nunes and Drèze gave car-wash customers a loyalty card needing eight stamps, or one needing ten with two already stamped. The work left was identical. The second group was about twice as likely to finish. A form that opens with an easy question someone actually wants to answer ("what are you looking for help with?") borrows the same effect.

If a gate genuinely has to come first, as sign-in did on that hackathon form (it stopped duplicate registrations), say why on the screen before it. "Sign in so each team can register only once" is a reason. A bare Google button is a toll.

Where this goes wrong: asking for contact details last and losing the people who abandon, whom you then can't contact at all. The compromise is to ask for an email early, with a reason, and everything costly after it.

4. Let people ask what a question means

A form can't explain itself. When a question is ambiguous, the respondent either guesses, which gives you a bad answer, or leaves.

Survey researchers have tested this for decades. Schober and Conrad found that letting an interviewer clarify what a question means sharply reduced errors in the answers, and the effect held in a later household telephone survey, at the cost of longer interviews.

On a web form, the cheap version is help text under the question. The better version is letting people ask. In chatform (on Pro), the respondent can type a question mid-form ("do I need to bring my own laptop?"), get an answer from notes you've written, and carry on from the same question.

Where this goes wrong: help text nobody reads because every field has it. Put it only on the questions people actually get wrong. Your answers will show you which.

5. Don't punish a short answer

Required fields with validation errors are the most common way a form tells somebody they've failed. Mostly it gets you a worse answer, typed only to make the error go away.

The research on open-ended answers points the other way: ask again, don't reject. In a 600-person experiment, Xiao and colleagues found that a chatbot which probed thin answers got responses that were significantly more informative and specific than the same questions as a web survey. Kim, Lee and Gweon found people answering by chat satisficed less, meaning they were less likely to pick whatever answer ended the question fastest.

In practice: if someone types "n/a" into "what's your budget?", a follow-up ("a rough range is fine, even under $1,000 or over") gets you a usable answer where an error message gets you "1".

Where this goes wrong: following up on everything. A clear choice answer needs no follow-up, and a form that interrogates every reply is slower than the one it replaced.

6. Keep what they gave you, and go back for them

Somebody who answered six questions and left is the warmest lead your form produced. Most tools throw that away, or keep the partial answers but never contact the person.

Two things matter if you do follow up.

Vary the reminders. Sauermann and Roach ran a web-survey experiment with over 24,000 people and found that reminders which changed their wording raised the odds of a response by over 30% against reminders that repeated themselves. Three copies of the same email is the weakest version of this. chatform's reminders are written differently each time and link straight back to the question the person stopped on.

Measure it honestly. On the hackathon form, 98 students who left partway got at least one reminder, and 24 finished afterwards. It's tempting to call that 24 recoveries. It isn't, quite: the form had no holdout group, and some of those students were always going to come back before the deadline. The only way to know what reminders earned is to hold a share of people back and send them nothing, then compare. Any recovery rate quoted without that comparison, including every one you'll see in vendor marketing, is counting people who might have returned anyway.

Where this goes wrong: reminding people who already finished. Check the response status at the moment each reminder is due, not when it was scheduled.

7. Know when a plain form is the answer

Everything above assumes your form is long enough, or ambiguous enough, for abandonment to be a real problem. A newsletter signup isn't. A three-field contact form with obvious fields isn't. For those, a plain form on one page is faster, and a conversation just adds clicks.

A rule of thumb that has held up for us: if any question could reasonably be answered "it depends", or you've ever had to email someone back to ask what they meant, the form is a candidate for the fixes above. If not, shorten it and leave it alone.

Where to start tomorrow

1. Turn on per-question drop-off and find the worst question.
2. Delete or defer every field whose answer doesn't change a decision.
3. Move sign-in, phone and budget questions after the first easy one.
4. Make sure partial answers are saved, and set up reminders that change their wording.

If you'd rather start from a form that already does most of this, the free form templates show every question and branch before you sign up, and the research behind conversational forms has every study above with its DOI.