Form abandonment: how to keep the leads who leave halfway

What form abandonment is

Form abandonment is when someone starts filling in a form and leaves without sending it. It is different from a bounce, where the visitor never touches the form. A lead form left halfway is a person who read enough to start typing, then stopped, and on most sites they leave no trace at all.

For a business that lives on inquiries, that is the most expensive kind of lost visitor. They found you, they wanted something, and they were one step from telling you who they are. This guide covers why it happens, how to reduce it, and how to keep the lead when it happens anyway, on a form you built yourself, without breaking privacy law or annoying anyone.

How often it happens

There is no single reliable “form abandonment rate”, and two of the most quoted figures do not measure what they are usually quoted for.

• Benchmarks from form analytics. Zuko, which measures forms for its customers, reports that about 55.5% of people who start a form on desktop complete it, and 47.5% on mobile, across 93 million form sessions^[3]. Contact forms are the hardest: only about 9% of people who see one send it^[4]. This is vendor data, without a published weighting or period, but it is the largest public dataset.
• The “81%” figure comes from a 2018 survey of about 500 US adults by The Manifest, in which 81% said they had abandoned an online form recently^[1][2]. It describes how many people have ever left a form, not what share of form starts are abandoned, and it is self-reported and eight years old.
• The “70%” figure is the online shopping cart abandonment rate (70.22%, an average of 50 studies by Baymard Institute)^[6]. It is about carts, not forms.

The useful number is your own: how many people start each form, and how many send it. If roughly half of your starters leave, you are close to the benchmark, and every one of them was a person who wanted something from you.

Why people leave

People rarely leave a form for one dramatic reason. Usually it is a small cost too many at the wrong moment:

• Too much to fill in. Every field is a question the visitor has to answer before they get anything back.
• Fields that feel like a trap. Some fields cost more than others. Across 1,362 forms, Zuko found the highest abandonment on password fields (10.5% on average), then email (6.4%), phone (6.3%) and name (5.3%)^[5]. In the same company’s data the number of fields has almost no effect on completion^[4]: it is which fields you ask for, and how, that loses people.
• No idea what happens next. “Submit” says nothing about whether a person replies, when, or whether a sales call follows.
• Errors that punish them. Validation that fires only at the end, wipes what they typed, or rejects a perfectly good phone number format.
• The phone. Tiny fields, the wrong keyboard for numbers, and a form that scrolls under the keyboard.
• Long or multi-step forms with no sense of progress. A wizard that never says how many steps are left feels endless.
• Life. A call, a meeting, a second tab with a competitor. Some people mean to come back and never do.

The first six you can fix. The last one is why recovery matters even on a perfect form.

Fix the form first

Recovering people who leave is the second line of defense. The first is a form fewer people want to leave.

• Cut every field you would not act on today. If you won’t use the company size to decide how to reply, don’t ask for it yet.
• Make phone optional, and say why you would call. “For a quick call about your quote, if you prefer” earns more numbers than a required field.
• Ask for contact details first in multi-step forms. Name and email on step one means even someone who leaves at step three can be answered.
• Say what happens after Send, next to the button: “A person replies within one business day. No newsletter.”
• Validate as people type, kindly, and never clear the form on an error.
• Use the right input types and autocomplete (type="email", type="tel", autocomplete="email") so phones show the right keyboard and fill fields in one tap.
• Show progress in multi-step forms: “Step 2 of 3” beats a progress bar with no numbers.

Measure where they leave

You can’t fix what you can’t see. Start with two numbers per form: how many people start it, and how many send it. Google Analytics 4 records both when enhanced measurement’s form interactions are on (form_start and form_submit)^[19], which is enough to know whether you have a problem. For multi-step forms, send one event per step, so you know which step loses people.

Field-level analytics tools go further and show which field people stop at. Use them when the start-to-send gap is large and you can’t tell why; they are overkill for a three-field contact form.

Recover the ones who still leave

Even a good form loses people. Recovery means saving enough of what they typed, before they send it, to reply to them. The table shows the difference on one visitor who fills in step one of a quote form and then closes the tab.

What to save, and what not to

• Save contact fields only: name, email, phone. They are enough to reply, and they are what the visitor expects you to have.
• Save when a field is left, never on each keystroke. Recording keystrokes is what the privacy lawsuits below are about, and it captures typos and things people deleted.
• Never save free text before submit. People draft, reconsider and delete. Half a message is not something they chose to send you.
• Never store passwords, card numbers or identity numbers, not even on submit.
• Delete what you don’t use. If a left-early lead is not followed up, it should not live forever.

Three ways to do it

Write it yourself. Listen for focusout on contact fields and send them to your own endpoint with keepalive, so the request survives the tab closing. Then you need a session id, a job that marks sessions as left early after a quiet period, de-duplication when the visitor does submit, and a notification. It is a day of work done well, and a source of quiet bugs done quickly.

// Save contact fields when the visitor leaves each one: never keystrokes,
// never free text, never anything sensitive.
const CONTACT = /e-?mail|phone|tel|^name$|first_?name|last_?name/i;
let sessionId = null;

form.addEventListener('focusout', (e) => {
  const el = e.target;
  if (!CONTACT.test(el.name) || !el.value.trim() || el.type === 'password') return;
  const fields = {};
  for (const [k, v] of new FormData(form)) if (CONTACT.test(k)) fields[k] = v;
  // keepalive lets the request finish even if the tab is closing
  fetch('/api/partial', {
    method: 'POST',
    keepalive: true,
    headers: { 'content-type': 'application/json' },
    body: JSON.stringify({ sessionId, fields, step: currentStep })
  }).then((r) => r.json()).then((r) => (sessionId ??= r.sessionId)).catch(() => {});
});

Use a form analytics tool. Several products add a script to any page and capture what visitors type into forms. Check what they capture (fields, keystrokes, session replays) and whether they may use the data themselves, against the privacy section below, before you add one: scripts that send typed emails to third parties before submit were found on thousands of sites^[16].

Use a form backend that does it. If your form already posts to a form backend, partial capture can live in the same place as your submitted leads, alerts and integrations. This is what we built OllieLeads partial submissions for: contact fields only, saved as each field is left, with a ready-made notice for your visitors, on any form you or your coding agent built.

Follow up the right way

Speed matters, though less dramatically than the internet claims. In a 2011 Harvard Business Review audit, researchers sent a test web lead to 2,241 US companies: 37% replied within an hour, 24% took longer than a day, and 23% never replied^[7]. In the same article, companies that contacted a lead within an hour were nearly seven times as likely to have a meaningful conversation with a decision maker as those that waited even an hour longer^[7].

The famous “reply in five minutes or your odds drop a hundredfold” comes from a 2007 vendor study of six companies that did not measure sales^[8], and a 2014 analysis of 20.7 million car-dealer leads found that dealers replying in 16 to 20 minutes closed more often than those replying in under 15^[9]. The honest reading: reply the same business day, ideally within an hour. For someone who left a form halfway, a short pause before you write is fine.

A good follow-up to someone who left your form halfway is:

• One message, from a person, about the thing they started asking for.
• Easy to answer: ask for the one detail you need instead of sending them back to the form.
• Easy to ignore: say plainly that you won’t chase them.
• Never a newsletter sign-up. Starting a quote form is not consent to marketing.
• Not a cold call, unless they gave you their number and the law where they live allows it.
• In the EU, only with consent. A regulator in Germany treats a reminder email to someone who left an order unfinished as advertising; ask when they give their email, for example with a checkbox (see below).

Subject: Your question about the kitchen project

Hi Mara,

You started our quote form earlier today and didn't get to the end.
No problem: if you still want a price, just reply with the rough
size of the kitchen and when you'd like to start, and I'll come back
to you within a day.

If you've sorted it out already, ignore this. You won't hear from us again.

Simon, Alpenglow Web

Privacy and the law

Saving what someone typed before they send it is the most legally sensitive part of form recovery. No court or regulator has said that it is lawful or unlawful as such. What has mattered so far is who captures the data, whether the visitor agreed before capture started, and what they typed.

United States

• California wiretap lawsuits (CIPA). A wave of class actions has targeted scripts that record what visitors type or do on a page. In Javier v. Assurance IQ the Ninth Circuit said consent must come before recording; agreeing to a privacy policy afterwards did not count^[10]. In Torres v. Prudential (2025), about a quote form whose vendor recorded clicks and keystrokes from the first interaction, the defendants won on the evidence, but the court found a real dispute over whether the vendor was an outside eavesdropper, because its staff could replay sessions and its terms allowed it to use the data^[11]. California’s SB 690 (2026) curbs one kind of these claims from 2027 and leaves the main wiretap section unchanged^[13].
• Federal appeals court, May 2026. In In re BPS Direct, the Third Circuit described code that that “captures inputs that a user types into text fields even if the user does not click ‘submit’”, and held that shoppers whose payment details were captured could sue, while people who only browsed could not^[12].
• Calls and texts. Under the TCPA, marketing calls with an autodialer, prerecorded or artificial voice, and marketing texts need prior express written consent, signed by the person^[14]. A form that was never sent contains no such agreement, so do not call or text left-early leads that way.

European Union, UK and Switzerland

• Collecting unsent input needs consent. The European Data Protection Board says information a user types into a page and that is collected before it is sent falls under the ePrivacy rules on accessing a device (paragraph 62), the same rules as cookies^[15]. Consent is the safe basis for visitors in the EU. The UK has the same rule in its own law (PECR), so treat UK visitors the same way.
• It happens a lot without anyone asking. A 2021 crawl of the top 100,000 websites found email addresses sent to tracking and marketing companies before the form was submitted, without consent, on 1,844 sites in the EU and 2,950 in the US^[16].
• Reminder emails can be advertising. The data protection authority of North Rhine-Westphalia treated emails sent after an online order was left unfinished as advertising that needs the person’s consent^[17].
• Switzerland asks sites to inform visitors at or before collection and to make unexpected uses prominent; consent is generally not required, except for sensitive data^[18].

How to reduce the risk

• Tell people before you save, in a short notice next to the form, not only in a privacy policy.
• Ask EU, UK and Swiss visitors first: save nothing before submit until they agree, for example as a purpose of its own in your cookie banner, or with an unticked checkbox by the email field.
• Save contact fields only, when each field is left. No keystrokes, no session replay, no free text, and never card numbers, passwords, health data or ID numbers.
• Keep the data with you or a processor that cannot use it for itself, with staff access restricted and logged.
• Treat left-early leads as not consented to marketing: one personal reply about their request, no automated calls or texts, and in the EU no follow-up email without consent.
• Delete what you don’t follow up, on a short schedule, and on request.

Checklist

• Every field earns its place; phone is optional and explained.
• Multi-step forms ask for contact details first and show “Step 2 of 3”.
• The button area says what happens after Send.
• Validation is kind and never clears the form.
• You know how many people start and how many send each form.
• Contact fields are saved when left, never keystrokes, never free text, never sensitive data.
• A short notice next to the form says what is saved before Send.
• Visitors in the EU, UK and Switzerland agree before anything is saved early.
• People who leave get one helpful message, quickly, and never a newsletter.
• Left-early leads you don’t follow up are deleted on a schedule.