Hypothesis
Removing non-essential fields from the lead form will raise qualified leads per visitor, because effort falls without losing the information sales needs to qualify. Both halves of that sentence are testable. The first half predicts that more people finish the form. The second half predicts that the leads they produce are still good enough for sales to work, which is the part teams forget.
The evidence usually comes from two directions. Analytics shows visitors starting the form and abandoning it at a particular field, and sales confirms that several fields are never read. The forms guide covers why field count, field order and optional labels affect completion; usability research such as the Nielsen Norman Group articles describes the same effect in terms of interaction cost. Write down which fields the challenger removes and why each is non-essential, in the experiment hypothesis template, and get sales to agree in writing before launch. That agreement is what stops a later argument about whether the shorter form "lost" vital information.
Metrics and guardrails
The primary metric is qualified leads per visitor: leads from each arm that reach the qualified stage, divided by visitors to the form page in that arm. The definition of "qualified" must be the one sales already uses, and the lead stages guide explains how marketing and sales stages should be defined so both sides count the same thing. Submissions per visitor, the raw CVRShare of visits or clicks that convert., is still worth reporting, but only as context.
The guardrails are lead-to-opportunity rate, which guards against leads that pass a quick qualification but rarely become pipeline, and junk submission rate, which guards against spam, duplicates and fake details that a shorter form can attract. If the business tracks cost per qualified leadSpend divided by only the leads that pass the business's qualification criteria, such as budget, need and fit; always higher than cost per lead., compare it as well; a cheaper CPACost of each conversion. on raw leads that hides a dearer cost per qualified lead is the classic false win in form tests.
Design
Randomise at visitor level and keep the assignment across sessions, because many B2B visitors return several times before submitting. The control is the current form; the challenger is the same form with the non-essential fields removed. Keep the layout, button text, privacy notice and thank-you page identical. Make sure both arms write to the CRM with a field that records which variant the lead saw, or you will not be able to join leads back to the test.
| Design element | Choice | Reason |
|---|---|---|
| Unit of randomisation | Visitor, held across sessions | B2B visitors often return before submitting |
| Variants | Current form versus shortened form | Isolates field count |
| Variant tagging | Hidden field passed to the CRM | Lets you join leads to outcomes later |
| Enrichment | Same enrichment applied to both arms | Prevents one arm looking better through data fill |
| Run period | Whole weekly cycles until the planned sample is reached | Covers weekday and weekend patterns |
| Read point | After the qualification window has closed | Lets lead quality mature |
The test therefore has two clocks. The run period collects visitors and leads. The qualification window, after the run ends, gives sales time to review them. Agree its length with sales in advance, based on how quickly they normally reach a decision on a new lead.
Sample size reasoning
Because the primary metric is qualified leads rather than submissions, the event you are counting is rarer, and rarer events need more visitors to detect the same relative change. Start from the current qualified-lead rate per visitor, decide the smallest improvement that would justify changing the form, and work out the visitors needed with the sample size and significance guide. The conversion rate calculator helps you derive the baseline rate from your own numbers.
If the answer is far beyond your traffic, you have two honest options. Test a more radical cut, which gives a larger expected effect, or accept that this test can only detect large changes and say so in the hypothesis document. What you should not do is switch the primary metric to raw submissions halfway through because it reaches significance sooner.
Reading the results
Wait for the qualification window to close, then read in order. First, qualified leads per visitor: did the shortened form produce more, beyond what noise allows? Second, lead-to-opportunity rate: are the extra qualified leads turning into pipeline at a similar rate? Third, junk submissions: has spam or fake detail risen enough to cost sales time?
Three outcomes are common. More qualified leads with stable guardrails is a clear win; roll out and remove the fields from the CRM requirements too. More submissions but flat qualified leads means the removed fields were doing a filtering job; keep the long form, or move those questions to a second step after submission. Fewer qualified leads is rare but happens when a removed field helped routing; check whether leads went to the wrong team.
A software firm cuts its demo form
Fieldnote, an invented field-service software company, has a demo request form with eleven fields. Sales confirms it only uses six. The team tests a six-field version for four weeks, then waits three further weeks for qualification. Submissions per visitor rise from 2.2% to 3.4%, but the read that matters is qualified leads per visitor, which moves from 0.9% to 1.2%. Lead-to-opportunity rate is roughly 40% in both arms. Junk submissions rise slightly, mostly fake company names, so the team adds enrichment to flag them and rolls out the shorter form.
Pitfalls
Counting raw leads is the first and largest pitfall. It makes nearly every shorter form look like a winner, and the cost appears months later as a sales team buried in leads it cannot use. Keep the primary metric fixed from the start. Counting attempts as submissions is a related trap in both arms, and the tutorial on tracking forms with Tag Manager records a lead only after the server accepts it.
CRM lag is the second. Leads from the last days of the test have had less time to be qualified than leads from the first days, and if one arm generated more leads late in the period it will look worse at an early read. Waiting for the qualification window to close for every lead fixes this. Enrichment gaps are the third: if the shortened form relies on an enrichment tool to fill company size or industry, check that the tool is filling those fields reliably, or the shorter arm's leads will be routed and scored on missing data. Finally, watch for sales behaviour changing because reps can tell which form a lead came from.
To size the extra enquiries a shorter form might bring, and to pair them with lead quality, follow the lead generation walk-through for the conversion rate calculator.
Before testing length, check that length is the problem; the page on form abandonment that is too high shows how to tell field count from validation, mobile input, privacy and silent-failure causes.
On LinkedIn, where the form opens pre-filled inside the platform, the diagnosis of a Lead Gen Form that members open but rarely submit separates field count from the offer, the typed work email field, consent wording and gating before a length test is worth running.
Common mistakes
Form tests go wrong in predictable ways, nearly all of them about what gets counted and when. The general rules for running a clean split live in the A/B testing guide; the landing page headline test pairs well with this one, and the experiment library routes other questions.
- Declaring victory on submissions. More form completions is the expected side effect, not the result.
- Reading before qualification. Fresh leads look promising until sales reviews them.
- Not tagging the variant in the CRM. Without it, the test cannot be joined to pipeline outcomes.
- Removing fields sales relies on. Ask which fields are read, and get the answer in writing.
- Enriching one arm only. Data fill must be the same on both sides or quality comparisons are unfair.
- Changing the thank-you page too. A second change muddies the comparison of the form itself.
