The study that founded this entire rule reports ten form submissions in its before period and twenty-six in its after period. The two measurements were taken a year apart, on the author’s own website, with no simultaneous control.
That is not a scandal. It is an honest small case study, published as such in 2008, that the industry turned into a law of nature. The problem is what got built on top of it, and what the larger datasets say when you actually open them.
This page traces the number, shows what the biggest dataset in the field found instead, and gives a test for field count that does not require any of it.
The 120% figure, in full
It is real, it is traceable, and it is much smaller than its reputation.
The claim, verbatim. Contact form conversions increased 120% when the number of fields was reduced from 11 to 4, a 64% decrease.
Who published it. A Chicago web development firm, about its own website, in a case study released in June 2008.
The two periods. October to November 2007, then April to May 2008. Two months each, separated by a year.
The before numbers. 2,827 site visits, 184 form views, 10 form submissions, a 5.4% conversion rate.
The after numbers. 3,071 site visits, 219 form views, 26 form submissions, an 11.9% conversion rate.
What kind of study that is. A sequential before-and-after comparison, not a split test. The two windows are a year apart, so anything else that changed on the site or in the market over that year is inside the result. Site traffic itself grew 8.6% between them.
What it can support. That one firm saw more contact form submissions after shortening its form. That is a legitimate observation and the firm reported it honestly, including its raw counts, which is more than most of what followed.
What it cannot support. A general law. Sixteen additional submissions is not a finding you can build a discipline on.
Ten submissions, then twenty-six, a year apart. The percentage is arithmetically correct and evidentially thin. Source : Imaginary Landscape, form case study (2008)
What got built on top of it
The derivative claims are worse than the original, because they lost the honesty.
The per-field tables. You will find pages giving a precise conversion rate for each field count: so much at one field, so much at two, declining neatly through to ten or more. I could not trace any of them to a primary source. The pages carry no study name, no sample, no period and no bibliography.
The attribution drift. The 2008 case study is now regularly credited to a large credit reference agency. No such source exists. The chain of citation corrupted somewhere along the way and nobody checked.
The dropped context. The original firm published its raw counts, including the number 10. Every republication I found presents the 120% without them.
Who is still citing it. A major landing page software company continues to present the case as a general principle about choice and conversion, without the sample size and without noting that the two measurements were a year apart.
Why that matters more than a footnote. That company’s own benchmark report, discussed below, is the most methodologically serious document in this field. Citing a ten-conversion anecdote as a principle, while publishing a rigorous report that does not examine the question at all, is a good picture of how this subject works.
The general shape of the problem. Nobody involved is being dishonest. A small firm published a small finding openly, and an industry that wanted a rule found one.
Worth knowing before you cite it, because it gets cited on this question constantly.
What it is. The landing page conversion benchmark most of the industry treats as authoritative, covering more than 41,000 landing pages, 464 million unique visitors and 57 million conversions over a twelve-month window ending July 2024.
Its methodology, declared. Entirely observational, drawn from real platform data rather than any controlled experiment. It states that it uses medians in some places and means in others, using medians to avoid distortion by extreme values.
What it says about form fields. Nothing. There is no analysis of field count in it.
Why that matters. It is the best-documented dataset in this area by a distance, and on this specific question it is silent. Anyone citing it in support of a field-count claim has not opened it.
What it is genuinely good for. Conversion rate benchmarks by industry, with a declared sample and a declared method. Use it for that.
The largest form dataset says something else entirely
There is a study built specifically on forms, with a real sample, and its table does not support the simplification.
The sample. More than 650,000 anonymised form users, data from January 2015, mobile-responsive forms only, with disproportionate-traffic outliers removed in three sectors.
The table, by form type. Contact forms: 4 fields, 1% conversion. Order and payment: 9 fields, 9%. Donation: 7 fields, 7%. Order registration for events: 15 fields, 11%. Lead generation: 10 fields, 11%. Survey: 22 fields, 14%. Contest: 11 fields, 34%.
Read the two ends of that. The form type with the fewest fields had the worst conversion rate. The form type with the best conversion rate carried eleven fields.
What is actually driving it. What the person is filling the form in for. Someone entering a contest will give you eleven fields cheerfully. Someone contacting a supplier they do not know yet will not give you four.
The finding in that report that dwarfs field count. Splitting the same form across multiple pages moved conversion from 4.53% to 13.85%, at unchanged field count. Presentation beat quantity by a wide margin.
The report’s own conclusion. That it is not just the type of form used but the number of fields included that makes the difference. Its own table, read carefully, puts far more weight on the first half of that sentence than on the second.
Motivation dominates count. And splitting the same form across pages tripled conversion at identical field count. Source : Formstack Form Conversion Report (2015)
Two results that break the rule in opposite directions
A single practitioner, two tests, and between them they dismantle the simplification more effectively than any critique.
Test one: fewer fields, worse result. A consultant persuaded a client to let him remove three form fields. The result he reported was a 14% drop in conversion.
His explanation, in his own words. He had removed all the fields that people actually want to interact with, and left only the ones they do not.
Why that is more interesting than it sounds. It means fields are not interchangeable units of friction. Some of them are the reason the visitor is filling the form in at all. A field asking what problem you are trying to solve is not a tax on the visitor. It is often the part they came to answer.
Test two: same fields, better result. On a different test he kept all nine fields and changed only the wording of the labels. He reported +19.2%.
What the pair implies. That which fields, and how they are worded, plausibly carry more weight than the count. That claim is harder to put on a slide, which is presumably why the count version won.
The status of both. Conference anecdotes, without published sample sizes or durations, from the same source. I am using them as counterexamples to a rule, not as a rule of their own. One counterexample is enough to break a universal claim, and this is a universal claim.
Removing fields lost 14%. Rewording the same nine fields gained 19.2%. Fields are not interchangeable units of friction. Source : Aagaard, reported via CXL (2019)
A test that works without any of this literature
Since the evidence does not give you a number, use a rule that does not need one.
The named-person test. For each field, name the person who reads the answer and say what they do differently because of it. If you cannot do both, delete the field. This resolves most forms in ten minutes and it needs no statistics.
The classic failures it catches. Company size, when nobody segments on it. Phone number, when nobody calls. “How did you hear about us”, when the answer is never read and you already have the referrer. Job title, when the sales team asks again on the call anyway.
The fields it usually keeps in B2B. A work email address, because it is how you reach them. A name, because you will write to them. And one qualifying field that genuinely changes what happens next.
Why one qualifying field earns its place. A B2B form that asks nothing beyond an email produces leads a salesperson must qualify from zero. Moving one question from the call to the form is usually a net gain even if the form converts slightly worse, because you are optimising cost per meeting held, not cost per submission.
The number that actually matters. Cost per qualified opportunity. A form that converts at 12% and produces unusable leads is worse than one converting at 8% that produces workable ones. Field count arguments almost always ignore this, because submissions are easy to count and qualification is not.
And the thing to fix before the count. Labels, order, and whether the form is split across steps. On the evidence above, all three plausibly matter more than the number of boxes.
If you cannot name who reads it and what they do differently, the field is costing you submissions for nothing. Source : Method (2026)
The founding study reports 10 submissions before and 26 after, measured a year apart on the author’s own site, with no simultaneous control.
It is honest and traceable, published with its raw counts in June 2008. Every republication drops the counts and keeps the 120%.
It is now widely misattributed to a large credit reference agency, for which no source exists.
The per-field conversion tables have no primary source. No study name, no sample, no bibliography on any page carrying them.
The reference landing page benchmark does not cover field count at all, despite 41,000+ pages and 57 million conversions.
The largest form dataset points the other way. Fewest fields, worst conversion at 1%. Best conversion at 34% carried eleven fields.
Splitting the same form across pages moved conversion from 4.53% to 13.85% at unchanged field count.
A practitioner who removed three fields reported a 14% drop, and got +19.2% by changing only the labels.
Use the named-person test, then measure cost per qualified opportunity. Book a diagnostic, or see how we approach B2B websites.
Frequently asked questions
Does reducing form fields increase conversions?
Sometimes. The founding study reporting a 120% increase measured 10 submissions before and 26 after, a year apart, on the author's own site. A practitioner who removed three fields for a client reported a 14% drop.
Where does the 120% figure come from?
A case study self-published in June 2008 by a Chicago web development firm about its own contact form, reducing it from 11 fields to 4. It is real and traceable, and it is often misattributed to a large credit bureau.
How big was that study?
2,827 site visits and 10 form submissions in the first period, 3,071 visits and 26 submissions in the second. The two periods were two months each, separated by a year, with no simultaneous control.
What about the tables giving a conversion rate per field count?
They could not be traced to any primary source. The pages publishing them carry no study name, no sample size and no bibliography. Treat those numbers as unsourced.
Does the main landing page benchmark report cover this?
No. Despite covering 41,000+ landing pages and 57 million conversions, it contains no analysis of form field count, so it cannot be used to support any claim about it.
What does the largest form dataset show?
That field count is not the dominant variable. Contact forms averaged 4 fields and converted at 1%. Contest forms averaged 11 fields and converted at 34%. What people are filling the form in for matters more.
Does splitting a form across pages help?
In that dataset, substantially: 4.53% for single-page forms against 13.85% for multi-page, at the same field count. That is a larger effect than anything attributed to field count itself.
So how many fields should I use?
As many as you will actually use, and no more. The test is not the count, it is whether a named person acts differently because of each answer. If nobody reads a field, delete it.