81% of People Abandon Forms. Here's How I Get Them Back Without Rebuilding Anything.

Most people who open your form will not finish it. In The Manifest's survey, 81% said they had abandoned at least one online form recently. Across the industry the number that gets quoted is around 67% of starts never reaching submit.

I work on chatform, a form builder, so I see these numbers every day across other people's forms. That's my stake, and I'll say when I'm talking about my own product. I used an AI assistant to help outline this and wrote and checked the rest myself.

What follows is the order I'd fix things in, starting with changes that cost nothing and work in any form tool, including Google Forms. The product comes last, and you won't need it for most of this.

First, find the question where they leave

Abandonment is not a single number. It's a cliff at one question. Nine times out of ten it's a field asking for more than the person wants to give at that point: a phone number, a budget, a home address for a digital product.

Most builders show a drop-off per question somewhere in their analytics. If yours doesn't, split the form into two pages and compare submissions on page one against page two. That's a crude funnel and it's enough to find the cliff.

Once you know the question, you have three options: make it optional, move it later, or delete it.

Make the phone number optional

I know sales wants it. A required phone field is the single most common reason people quit a form. Make it optional and ask for it again in the thank-you email. You'll lose a few numbers and gain a lot of completed forms, and a completed form with an email is a lead you can still call once they reply.

Say how long it takes, and show progress

In the Loqate survey, people said they often don't know whether finishing a form is worth the effort. Tell them: "Eight questions, about three minutes." Multi-step forms without a progress indicator abandon about 20% more than forms that show where you are. Both fixes are a sentence and a checkbox in most tools.

Put the easy questions first

Ask for the name before the budget. Ask for the kind of project before the email. People who've answered three easy questions have invested something, and they finish more often than people who hit a wall at question one. This is the same reason a chat feels lighter than a page of fields: one small ask at a time.

Keep what they've already typed

This is the one most people skip, and it's where the real money is. If someone answers five of eight questions and closes the tab, those five answers are gone in most form tools unless you turn on partial saving. Gravity Forms calls it partial entries, HubSpot users ask for it in the community forum every month, and the newer builders have it as a toggle.

Turn it on. A half-finished response with a name and an email is still a lead. It's also how you find out which question is the cliff, because you can see exactly where each person stopped.

Then email them a way back

Here's what almost nobody does. The person who left at question six was interested enough to answer five. A reminder that lands them on question six, with the first five already filled in, is the cheapest lead you'll ever recover.

This isn't a hunch. Sauermann and Roach's 2013 study in Research Policy on survey reminders found that varying the timing and wording of follow-ups raised response rates, and that the people who come back after a reminder are not the same people who would have answered anyway.

How to do it with no special tool: collect the email as question one or two, turn on partial saving, and once a day export the partials and send a short plain-text email with the form link. It takes ten minutes a day. It works.

Fix the answers, not just the count

Completion is only half the problem. The other half is what's in the completed forms. Ask "what went wrong with the last tool you tried" on a page of fields and you'll get "it was bad", "n/a", and "pricing". Those are rows you throw away.

Xiao and colleagues published a study in ACM Transactions on Computer-Human Interaction in 2020 comparing a standard web survey with a chat-based one that asked a follow-up when an open answer was thin. The chat version got longer, more specific answers to the open questions. People will say more when something asks them to.

If your tool can't follow up, the manual version is to write the follow-up into the question itself: "What went wrong, and what were you trying to do when it happened?" Two-part questions get two-part answers.

What I built when the manual version got old

Doing the export-and-email loop by hand every morning is fine for a month. Then it's a chore, and the follow-up questions you wrote into the form get skipped by the people in a hurry anyway.

So this is the part I work on. chatform runs the form as a conversation. When a free-text answer is too thin to use, it asks one more question in plain words and records the real answer. Choices and scales are matched exactly and never sent to a model. When someone leaves, it saves every answer they gave and emails them up to three times, four hours, then a day, then three days later, with a link that opens on the question they stopped at. If they've finished in the meantime, the reminder is dropped.

Full disclosure: I'm on the team that builds it. The free plan has unlimited responses; the follow-up emails and partial responses are on the paid plan. If the manual loop above is already working for you, you don't need it. If you'd rather not run the loop by hand, the chatform home page has a live form you can try before signing up.

The order, one more time

1. Find the question where people leave.
2. Make the phone number optional.
3. Say how long it takes and show progress.
4. Put the easy questions first.
5. Save partial answers.
6. Email a way back to the exact question.
7. Ask a follow-up when the answer is thin.

Do the first four this afternoon. They're free in every tool. Then look at your completion rate next week and decide whether the last three are worth automating.

AI disclosure: I used an AI assistant to help outline and structure this post. The facts and prices were checked by hand on 20 October 2026.

Originally published on Medium.