Somebody opened your form, answered the first question, and left. They wanted something badly enough to start, and something before the last screen changed their mind. That is form abandonment, and unlike most conversion problems it leaves evidence: the exact screen where each person stopped.
This guide works through it in the order you would fix it: the arithmetic, the seven causes behind most abandoned forms, how to tell them apart in your data, and a fix for each. One ground rule: statistics about form abandonment rarely measure the same thing, so none are quoted here. Every number on this page is arithmetic or a labeled illustration.
What form abandonment is, and what it is not
Form abandonment is when someone starts a form and leaves before submitting it. The word doing the work is starts. A visitor who loads the page and never answers anything has not abandoned your form. They declined it, which is a different problem with a different fix.
- Declined: loaded the form, never answered question one. That points at the first screen, the promise that sent them there, or load time.
- Abandoned: answered at least one question, left before the end. That points at a specific question, or at the length of the whole.
- Completed: reached the end and submitted. The only group that becomes a response, a contact, or a row in a connected spreadsheet.
The middle group is usually the more valuable one to work on, because these people have already shown intent. Someone who typed their name and email wanted to finish. Something you asked, or the way you asked it, stopped them.
How to measure form abandonment rate
You need three counts and one list. The counts are views, starts and submissions. The list is how many people reached each question. Everything else is division.
| Measure | What it tells you | How to get it |
|---|---|---|
| Views | How many people loaded the form | Counted |
| Starts | How many answered the first question | Counted |
| Submissions | How many reached the end and sent it | Counted |
| Start rate | How well the first screen works | Starts ÷ views |
| Completion rate | How well everything after it works | Submissions ÷ starts |
| Form abandonment rate | The share of starters you lost | (Starts − submissions) ÷ starts |
| Question drop-off | Where exactly you lost them | Left at a question ÷ reached it |
An illustrative example: a registration form gets 1,610 views in a month, 820 visitors answer the first question, and 347 submit. The start rate is 51%, the completion rate is 42%, and the form abandonment rate is 58%: 473 people began registering and did not finish.
Some tools calculate abandonment from views rather than starts, which blends decliners with quitters and produces a bigger, less useful number. Pick one definition, write it down, and never compare figures built on different ones. Completion rate and abandonment rate always add up to 100%, so track whichever your team prefers. The form conversion rate guide covers the view-to-submission side.
There is no honest universal answer: a two-question signup and a twelve-question application should not share a target. Compare the form with itself, month over month, and treat any question that loses half the people who reach it as this week's problem.
Why users abandon forms: seven causes, ranked
Ordered roughly by how much damage one instance can do. Most abandoned forms have one dominant cause, not seven small ones, so one change usually buys most of the improvement.
1. Too many fields
Every question is a small decision, and a form with fourteen of them offers fourteen chances to decide the rest is not worth it. Length rarely shows up as one dramatic drop. It shows up as a bleed of a few percent at every step, adding up to a completion rate nobody can pin on a single question.
The fix is the delete test. Imagine a question's answers vanished from every response. If nothing anyone does would change, the question goes. The survivors can often move to a follow-up email once you have the submission.
A 14-question demo request asking for job title, company size, industry, annual revenue and how you heard about us → five questions: name, work email, company, what you want to see, and what time works. The rest gets asked on the call.
2. A sensitive question asked too early
Phone number, budget, home address, date of birth. People rarely quit on these because they are hard. They quit because at question two they have no reason yet to trust you with the answer. The same question at question six, after the form has explained itself, costs far less.
This is the easiest cause to spot: one red row near the top of the funnel and a normal form after it. Move the question later, add a line of help text saying what the answer is for, or ask whether you need it at all.
Question 2: "What's your phone number?" → question 5: "What's the best number to reach you? We only use it to text you if the venue changes."
3. Unclear progress
People commit to a form based on a guess about its length. With no indicator, every new screen raises the same question: how many more? A progress bar that jumps or stalls is worse, because it teaches the reader the form will not tell them the truth.
This shows up as drop-off that climbs in the back half, where patience runs out. Use an indicator that moves one step per question, or say it plainly on the first screen: six questions, about two minutes.
"Question 7" with no total and no bar → a bar that fills one sixth per screen, and a first screen that says "Six quick questions."
4. Required fields that feel optional
Some questions feel optional whatever your form insists: how did you hear about us, company website, a second phone number. Mark them required and you build a wall where the reader thought they were nearly done. Some leave. Others type "n/a" to get past it, which is abandonment you never see, recorded as data you cannot use.
The tell is a spike at one mid-form question plus junk answers in your responses table. Make it optional, add an answer for people who have none ("No website yet"), or delete it.
A required "Company website" on a form that many freelancers fill out → optional, with help text: "Skip this if you don't have one yet."
5. A bad mobile experience
A form that works on a laptop can be miserable on a phone: tiny tap targets, a letter keyboard for a phone number, a dropdown of two hundred countries, a next button below the fold. None of it is visible from a desk, which is why it survives so long.
Compare completion rate by device. If the form completes well on desktop and poorly on mobile, the questions are probably fine and the inputs are the problem. Use the field type that opens the right keyboard, and test on a real phone, one-handed.
A phone number in a plain text box and a row of small radio dots → a phone field with a country flag and dial code, and full-width options that take one tap each.
6. Surprise asks
A form makes a promise before the first question: download the checklist, get a quote, save a seat. Every question is judged against it. A work email to send a checklist feels fair. A phone number or a budget for the same checklist feels like a trap.
In the funnel it looks like a sensitive question, a single spike, but moving it later helps little if it still does not match the promise. Cut it, or change the promise: "Get a tailored quote" justifies a budget question in a way "See pricing" never will.
"Download the checklist" followed by "What's your budget for this project?" → the budget question removed, and a separate quote request form for the people who actually want a price.
7. Slow loading
The last cause strikes before question one, so it hides in the start rate rather than the funnel. A form embedded in a heavy page, or carrying an uncompressed background image, loses people who never saw a question.
If views are healthy and starts are low, test the first screen on a phone over cellular data, not office Wi-Fi. Compress images, and check what the host page loads first. The how to embed a form guide covers the embed side.
A 4 MB background photo behind question one → the same photo compressed to a few hundred kilobytes, and the first question on screen before the reader has time to reconsider.
How to find the exact question people quit on
Your abandonment rate tells you there is a leak. It does not tell you where. For that you need drop-off per question: of the people who reached each question, how many left there instead of answering it.
In Upform this is the Question Completion Funnel on the Results tab of every form, part of the built-in form analytics. Every question gets a row with a count and a drop-off figure, and any question that loses half or more of the people from the step before it is flagged in red. Here is the illustrative registration form from earlier:
Overview
347 total submissions
Views
1,610
Starts
820
Completion rate
42%
Avg. time
1m 06s
Question Completion Funnel
The email question loses 23 people, about 3%, which is normal: some people were only looking. Then the address question loses 426 people, more than half of everyone who reached it, and the form behaves normally afterwards. That is not a length problem. It is one question, and the list above says which kind: a sensitive ask, placed early, that most registrants had no reason to answer.
- Wait for a real sample. A funnel built from a dozen starts is noise.
- Split by device. Compare completion rate on mobile and desktop before you touch a question. A big gap points at the inputs, not the wording.
- Change one thing at a time, then compare the next period with the last one of the same length.
Once you have a red row, the numbers around it narrow down the cause:
| What you see | Likely cause | First fix to try |
|---|---|---|
| Healthy views, low starts | Slow loading or a weak first screen | Test question one on a phone over cellular data |
| One red row near the top | A sensitive question asked too early | Move it later and say what it is for |
| A mid-form spike plus junk answers | A required field that feels optional | Make it optional or add a way out |
| A spike at an unexpected question | A surprise ask | Cut it or change the promise |
| A few percent lost at every step | Too many fields | Run the delete test |
| Drop-off rising toward the end | Unclear progress | Show honest progress or state the length |
| Mobile completion rate well below desktop | A bad mobile experience | Fix field types and tap targets |
Why one question per screen reduces form abandonment
A long form shown as one page asks for a decision before anyone has typed a thing. The reader scrolls, counts the fields, estimates the effort and decides whether it is worth it. Plenty of the declines in your start rate are that decision.
A conversational form never shows the wall. It asks one question, takes one answer and moves on. The first answer is cheap, so more people start. Each answer is a small commitment that makes the next one easier. And every question becomes its own step in your data, so a problem field cannot hide among ten others.
It is not free: one question per screen adds a click per question, which a two-field signup does not need. It earns its keep when questions need thought, when there are more than three or four, and when some only apply to some people. The multi-step forms guide covers where that line falls.
4 →
Want the printed program mailed to you?
Say yes and we'll ask for an address next. Press Y or N.
Use conditional logic to skip what does not apply
The most reliable way to shorten a form without losing information is to stop asking people questions that do not apply to them. Someone attending online does not need a dietary question. A customer who rated you ten out of ten does not need the "what went wrong?" follow-up.
That is the job of conditional logic. In Upform each rule joins one or more conditions with AND or OR, using 20 operators, and takes one of four actions: show a question, hide it, jump to another question, or end the form. The registration form needs two rules:
Conditional Logic
2 rules configured
Rule Summary
When Printed program equals "Yes" → Show field: Your home address
Logic changes how you read the funnel, too. A question hidden from most people collects far fewer answers than the one before it. That is the rule working, not people leaving.
A worked example: one fix, before and after
Here is the whole loop on the illustrative registration form. The funnel pointed at the address question, and the cause list said a sensitive ask, placed early. It turned out the address existed for a printed program that about one registrant in seven wanted. So the address moved behind the yes-or-no question above, the dietary question moved behind "How will you attend?", and the next 30 days looked like this:
| Before | After | |
|---|---|---|
| Starts | 820 | 790 |
| Saw the address question | 797, everyone who got past email | 104, only people who asked for the program |
| Left at the address question | 426 | 9 |
| Submissions | 347 | 706 |
| Completion rate | 42% | 89% |
| Form abandonment rate | 58% | 11% |
Illustrative numbers from a sample event registration form over two 30-day periods. They show the method, not a result to expect.
Starts barely moved, because the first screen did not change: the whole gain came from inside the form. And the form now collects 95 addresses instead of 347, which is the point. Each one belongs to someone who wants something mailed, and the other 611 registrants never had to decide whether to hand one over.
Not every fix is this clean. Often the question is one you need from everyone, and the answer is to move it later and explain it. The loop stays the same: find the red row, name the cause, change one thing, compare.
A checklist to reduce form abandonment
Run this against any live form. Each line takes minutes to check, and each one maps to a cause above.
- Every question passes the delete test: somebody reads the answer and does something with it.
- The first question is quick to answer and asks for nothing personal.
- Phone, budget, address and other sensitive asks come after the form has explained itself, each with a line saying what the answer is for.
- Nothing is asked that the promise on the first screen does not justify.
- Questions that apply to some people sit behind logic, so everyone else never sees them.
- Required means genuinely required. Everything else is optional, or offers an answer for people who do not have one.
- Progress moves one honest step per question, or the length is stated on the first screen.
- Every question uses the field type that opens the right keyboard.
- Someone on your team has completed the form on a real phone, one-handed, over cellular data, and the first question appeared quickly.
- You know last month's completion rate, and which question lost the most people.
The form design best practices guide covers the layout habits behind several of these, from labels to error messages. Start with the red row, though. One question fixed this week beats a redesign planned for next quarter.