Form abandonment analysis without expensive tools

Most owners have never seen this number, so they assume the form is fine. The form is almost never fine. People start it constantly and give up partway through, and every one of those was a person who wanted to contact you enough to begin typing.

That is the part worth sitting with. An abandoned form is not a browser. It is a buyer who tried to hire you and stopped.

What does form abandonment actually tell me?

It separates two failures that look identical in your analytics.

Failure one: nobody starts the form. That is a page problem. Your offer is unclear, the trust is missing, or the form is buried where nobody scrolls. The fix is above the form.

Failure two: plenty of people start and few finish. That is a form problem, and the fix is inside it. Different diagnosis, different work, and without event tracking they produce the same flat conversion number.

You also learn urgency. A visitor who fills three fields and abandons at the fourth was engaged. Whatever is in that fourth field is costing you real customers, not casual traffic.

Starts and completions are two different numbers. If you only track completions, you cannot tell a quiet page from a broken form.

There is a reason to measure fields rather than steps, and it comes from the deepest research available on the subject. The Baymard Institute has tracked checkout flows since 2012 and reports that the average checkout in 2024 runs 5.1 steps and 11.3 form fields, a step count that has barely moved in more than a decade. Its conclusion after that decade is that step count matters far less than the number of form fields a user has to consider. The limitation has to travel with the finding: that is ecommerce checkout research, on buyers who already have a cart, and Baymard sells UX research and benchmarking. It tells you what to instrument. It does not tell you what your quote form should look like.

How do I track it without paying for a tool?

Three events, and any standard analytics setup can carry them.

The first is form start. Fire it once when a visitor interacts with the first field. Most tag managers can do this with a form interaction trigger and no code at all.

The second is form submit. You almost certainly have this already, since it is the conversion you count.

The third is the useful one: a field-level event on the first interaction with each field. Name them plainly, like form_field_phone or form_field_budget. Now you can see how many people reached each field and how many reached the next one. The largest step down is your problem field.

If your form is on a platform that will not let you add events, there is a cruder version that still works. Put the form on its own page, then compare page views to submissions. It is less precise, but it tells you whether the form or the page is at fault, which is most of the value.

Give it a month of data before you read it. And exclude your own visits, or you will be looking at your own testing.

What does the report look like when it works?

Like this. One month, one form, six events, and the answer is visible without any statistics.

Two hundred and forty people started. Eighty-eight finished, a little over one in three. The budget field lost 62 people and the phone field lost 56. Everything else is noise by comparison.

Now price it, because that is what turns a report into a decision. Delete the budget field. If half of the 62 who stopped there carry on at the same rate as the people who got past it, that is 31 more starts reaching the last stretch, where 88 of 96 historically completed. Call it 27 additional submissions a month.

At the 2026 median cost per lead in LocaliQ and WordStream's search advertising benchmarks, $66.69 across all industries, 27 submissions is roughly $1,800 a month of lead value you were discarding at one dropdown. For a restaurant, where the same table puts median cost per lead at $30.57, it is about $800. Substitute your own cost per lead and your own counts. The arithmetic does not change.

Say plainly what that calculation is and is not. The 27 is a modelled recovery, not a measured one, because nobody has published an abandonment recovery rate for local business lead forms. What is measured is the 62. You do not need the model to act on the 62.

Which fields cause the most abandonment?

The pattern repeats across almost every local business site, and it is not random.

Phone number, when it is the only contact option offered. Some people will give you an email address and will not give you a phone number to a business they have not spoken to yet. Let them choose.

Budget or price range. This one is brutal, because the person often genuinely does not know. Asking a homeowner to estimate a project budget before you have described anything is asking them to guess and be wrong in public. The better move is to publish a range yourself, and putting pricing on the site answers the question they were stuck on.

Address fields, particularly on mobile, where they break into several boxes that autofill badly.

Dropdowns with no matching option. If a visitor's situation is not in your list, they stop. A short open text box beats a long dropdown almost every time.

And anything that reads as intrusive without explanation. Asking how they heard about you is fine at the end and expensive at the start.

Where does this measurement mislead you?

In four specific ways, and knowing them is the difference between reading the report and being fooled by it.

Autofill collapses the funnel. A browser filling name, email and phone in one action can fire three field events in the same instant, which makes the top of your funnel look flawless and pushes all the apparent loss onto the fields autofill cannot guess. That is usually the budget or detail field, which is genuinely the problem, so the distortion happens to point the right way. Do not assume it always will.

Tabbing is not typing. A first-interaction event fires on focus. Someone tabbing through to see how long the form is will register on every field and complete none, which inflates your reach numbers and deflates your apparent per-field loss.

Conditional fields break the sequence. If a dropdown reveals two more fields, your step-down table is comparing populations that never saw the same form. Instrument the branches separately or you will misread the whole thing.

Small numbers still bite. Forty starts in a month will not produce a readable step-down table, and no amount of event naming fixes that. Nor can you resolve it by testing. What a split test needs is set by the metric's variance and by how small a change you want to detect, not by a visitor count, which is how Kohavi, Deng, Longbotham and Xu put it in their KDD 2014 paper. The standard two proportion version is 16 × p × (1 - p) ÷ d², where the 16 is 2(1.96 + 0.84)² rounded, for 95% confidence and 80% power. A form converting at 3% needs about 12,900 sessions per variant (derived) to detect a 20% improvement. If a tool vendor quotes you 10,000 monthly visitors instead, that figure traces to reference [29] of the same paper, a QuickSprout blog post from January 2013. Below that, watch sessions instead and treat what you see as evidence rather than statistics.

The measurement is still worth having at low volume, with one adjustment: read it as a list of observed failures rather than as a rate. One person visibly stuck on a field is information. A percentage built on nine people is not.

What do I do once I know where they quit?

In order of how cheap the fix is.

Delete the field, if you can live without it. Fastest fix available and it costs nothing.

Make it optional, if you want the answer but do not need it. Half of people will still answer, and the other half will finish the form instead of leaving.

Rewrite the label, if the field is necessary. Most abandonment on a required field is confusion rather than refusal. "Approximate square footage, a rough guess is fine" performs differently from "Square footage." Wording carries a lot of weight here, which is part of the case for taking the website copy seriously rather than treating it as filler.

Move it later, if the field is intake rather than contact. Ask on the phone or in your reply, once the person is already talking to you.

And give people who abandon another route. A tappable phone number beside the form catches the ones who were never going to finish typing, which is why click to call still matters more than most site owners assume. Sterling Sky's 2026 field data documents call buttons becoming scarcer on Google's own mobile surfaces, including AI powered local packs that carry no call button at all, which makes the one on your page the reliable path rather than the redundant one.

What to do this week

Add the form start event. That is a fifteen minute job in a tag manager and it is the single measurement that changes the conversation.

Then add field-level events on the three fields you suspect most. You do not need all of them on day one.

While you wait for data, fill in your own form on your phone and count the taps from first touch to submitted. Anything over about a dozen deserves a second look.

At the end of the month, build the six line table above with your own counts and find the largest step down. Fix that one field. Then leave the form alone for a month and look again, because changing five things at once means you will never know which one mattered.

If a large share of abandoners never come back, that is worth knowing before you buy the tools people will suggest next, whether that is a chat widget on the page or something more expensive.

Questions about your form drop-off?

Email me at eric@seod.com with the URL of the page your form sits on. I will send back the exact list of events to add, named and in order, so you can hand it to whoever manages your tag manager and have it running the same day.

If you already have events firing and cannot read the report, send me a screenshot of the numbers and I will tell you which field is costing you the most.

Otherwise there is more on conversion here.