How to A/B test a lead form — and how many visitors you need first
What to test on a lead form, how many views each version needs (a table by completion rate), why two versions beat four, why checking daily crowns false winners, and how Heyflow, involve.me, Jotform and ClarNow handle A/B tests.
An A/B test of a lead form is simple to describe: half of your visitors see version A, half see version B, and you keep the one more people finish. The hard part is not the setup. It is knowing how many visitors the test needs before it can tell you anything, and not stopping it the day B happens to look good.
This guide covers what to test, how much traffic each kind of test needs (with a table you can read your own numbers from), how to read the result without fooling yourself, and how the tools that have A/B testing built in compare.
The short answer
- Test one big change. Fewer questions, a different first question, or the email asked later. Button wording rarely moves a lead form enough to measure.
- Work out the traffic first. A form that 20% of visitors finish needs about 770 views per version to detect a 30% improvement. If you don't get that in a couple of months, test a bigger change.
- Two versions, 50/50. Every extra version needs more traffic for the same answer: four versions need about 2.7 times the visitors of two.
- Decide once, at the end. Checking every day and stopping at the first "95% confident" crowned a winner in 29% of our simulated tests where A and B were identical.
- Then check lead quality. A shorter form often gets more leads; make sure they are leads you want before you keep it.
What to test on a lead form first
A lead form has fewer moving parts than a landing page, and that helps: there are only a few changes big enough to show up in a small sample. Roughly in the order we would try them:
- Fewer questions. Three questions against five, or dropping the phone number. This is the change most likely to move completion a lot, and also the one most likely to change who gets through, so it is the one where lead quality matters most.
- The first question. An easy, low-stakes opener ("What are you looking for?") against a qualifying one ("What is your budget?"). Many people decide whether to continue on the first screen.
- Where the email goes. First, to catch everyone who starts, or last, after the visitor has already put in some effort.
- The welcome screen. With one, or straight into the first question; a promise of what the visitor gets against none.
- The ending. Usually tested for what happens next (a booked call, a download) rather than for completion, which the ending cannot change.
Change one of these at a time. If B has fewer questions and a new theme and another opener, a win tells you B is better, not why, and you cannot reuse what you learned on the next form.
How much traffic you need
Two numbers decide it: the share of visitors who finish the form today, and the smallest improvement worth detecting. The table gives the views each version needs to detect that improvement with 95% confidence and an 80% chance of catching it if it is real, the usual defaults. Double it for the whole test.
| Completion rate today | To detect +20% | To detect +30% | To detect +50% |
|---|---|---|---|
| 10% | 3,841 | 1,774 | 686 |
| 20% | 1,683 | 772 | 294 |
| 30% | 963 | 437 | 163 |
| 40% | 604 | 270 | 97 |
The improvements are relative: +30% on a 20% completion rate means 20% to 26%. Read it with your own numbers. A form with 1,000 views a month and a 20% completion rate needs 2 × 772 = 1,544 views to detect +30%, which is about seven weeks. To detect +20% it needs 3,366 views, more than three months. That is why the first test should be a big change: small ones need traffic most lead forms never get.
Tools set a floor before they show any verdict at all: ClarNow waits for 100 views and 10 completed responses per version, Heyflow for 10 submits per version. That floor is when a verdict becomes possible, not a sample size to stop at.
Why A and B, not A, B, C and D
Testing four versions sounds like learning four times as much. With a fixed amount of traffic it works the other way. Each version gets a quarter of the visitors instead of half, and with three comparisons against A, each needs a stricter threshold, or one of them will look like a winner by chance.
Put together, four versions need about 2.7 times as many visitors as two to detect the same improvement with the same confidence. For the form above (20% completion, detecting +30%) that is 4,116 views instead of 1,544: about four months instead of seven weeks. For most lead forms, two versions run one after the other answer more questions than four at once.
Reading the confidence without fooling yourself
When ClarNow says "B converts 35.5% better than A, with 96.4% confidence", the confidence comes from a standard two-proportion test on completions per view. It means that, if A and B were really the same, a gap this big would come up by chance less than 5% of the time.
That 5% only holds if you look once. Every check is another chance for noise to cross the line, and the numbers swing a lot early on. We simulated it: 4,000 tests where A and B were the same form, both completed by 20% of visitors, 2,000 views each, judged by ClarNow's own rules.
| How the result was read | False winners |
|---|---|
| Once, at 2,000 views per version | 5.6% |
| After every 50 views, stopping at the first 95% | 28.9% |
| After every 10 views, stopping at the first 95% | 34.5% |
So: pick the number of views from the table above before you start, run whole weeks, and decide when you reach it. Watching the numbers in between is fine; acting on them is what breaks the test. If you stop at the planned point and there is no clear difference, that is an answer too: the change did not matter much, so keep the version you prefer and test something bigger.
Run a lead form A/B test in ClarNow, step by step
1. Create variant B
Open the form in the builder and click Create variant B above the questions. B starts as a copy. The A | B switch shows which one you are editing; with B selected, every change goes to B and A stays as it is. B has its own questions, choices, pages, logic, welcome screen, endings and theme. The link, embed code, notifications and integrations belong to the form and work for both.
2. Make the one change, then publish B
Remove the two questions, or move the email, then click Publish with B selected. Publish publishes B only.
3. Start the test
Go to Share → A/B Test and click Start test. Each new visitor gets A or B at random, 50/50. Someone who leaves halfway and comes back to finish sees the version they started.
4. Wait for the number you planned
The tab counts views, started and completed responses and conversion for each version, with one line underneath: Not enough data yet, No clear difference yet, or which version converts better and with what confidence. It never stops a test or picks a winner by itself.
5. Check the leads, then pick
Before you keep B, open Results and filter by Variant: are B's leads as good as A's? Then click Stop test and choose Use B, which publishes B's content under the same link, or Keep A. Every response stays, marked with its version.
The A/B testing guide has screenshots of each step; the product page has the overview.
How Heyflow, involve.me and Jotform do it
Three tools that lead form buyers often compare. Heyflow and involve.me split traffic themselves; Jotform does not. All of it read from each vendor's help center and pricing page on 17 November 2026, Heyflow's prices on 10 November 2026.
| ClarNow | Heyflow | involve.me | Jotform | |
|---|---|---|---|---|
| Built in | Yes | Yes | Yes | No: clone the form and use an outside testing tool |
| Where B lives | Inside the same form | Inside the same flow | A second funnel; a test joins the two | A second form |
| Link to share | The form's own link and embed | The flow's own link | The test's link; a funnel's own link shows only that funnel | Each form embedded on its own page |
| Traffic split | 50/50, fixed | 50/50 by default, adjustable | Random; no split setting described | Set in the outside tool |
| Statistics shown | Confidence (two-proportion test); a leader only at 95%, after 100 views and 10 completions per version | Probability of Being Best (Bayesian); "won" at 90% or more, shown after 10 submits per version | Raw numbers (visits, starts, submissions, completion rate); no significance described | None built in; a separate A/B test calculator |
| Choosing the winner | By hand: Use B or Keep A | By hand: Finish A/B test, the other version is discarded | By hand | By hand |
| Plan with A/B testing | Pro, €24/mo billed yearly (€30 month to month) | Insights bundle: on Scale, €239/mo billed yearly (€299 month to month), or €40/mo extra on a smaller plan | Scale, $139/mo billed yearly ($199 month to month) | — |
A few things the table can't show. Heyflow's 90% Probability of Being Best and ClarNow's 95% confidence are different measures, so the thresholds are not comparable number for number. Both are stricter than eyeballing two conversion rates, which is what involve.me's results leave you to do: its help article says not to call a winner too early on low traffic and leaves the call to you. Heyflow also lets you send, say, 80% of traffic to the baseline, which limits the damage of a risky variant at the cost of a longer test.
Jotform's guide to A/B testing web forms (last updated 22 April 2026) clones the form, embeds each copy on its own page and measures with Google Website Optimizer, a tool Google retired in 2012. Its A/B test calculator gives a p-value for numbers you collect yourself. It can be done, but it is a setup project, not a feature.
What ClarNow's A/B testing won't do
- Only two versions, only 50/50. You can't test three versions or send 10% of traffic to a risky B. Heyflow lets you set the split.
- One goal: completed responses per view. It does not compare lead quality, scores, a particular ending or revenue. Use the Variant filter in Results and judge those yourself.
- No protection against peeking. The confidence is shown live and is only valid when read once at the planned end, as above. It does not adjust for repeated looks.
- Edits during a test mix the results. The builder warns you, but a change published mid-test means some visitors saw the old B and some the new one. Stop, edit and start a new run instead.
- Pro plan and up. Free and Start can't create B. If you downgrade during a test, it pauses and everyone sees A.
Questions people ask
How many visitors do I need to A/B test a lead form?
It depends on how many visitors finish the form today and how big a difference you want to detect. A form that 20% of visitors complete needs about 770 views per version to detect a 30% relative improvement (20% to 26%) at 95% confidence and 80% power, so about 1,540 views in total. A 50% improvement needs about 300 per version; a 20% improvement about 1,700.
How long should a form A/B test run?
Until each version has the number of views you planned before you started, and for whole weeks, so that weekdays and weekends are both in it. At 1,000 form views a month, a test that needs 1,540 views takes about seven weeks. Stopping the first day the result looks good is the most common way to get a false winner.
What should I A/B test first on a lead form?
The change most likely to move completion by a lot: the number of questions, the first question, or where you ask for the email address. Small wording changes need far more traffic to show a difference, and most lead forms never get it.
Can I test more than two versions of a form?
Mathematically yes, but each extra version splits the same traffic further. To detect the same improvement with the same confidence, four versions need about 2.7 times as many visitors as two. ClarNow only lets you test A against B for that reason.
Which form builders have built-in A/B testing?
As of 17 November 2026: Heyflow (Insights bundle, on Scale or as an add-on), involve.me (Scale plan) and ClarNow (Pro plan) split traffic between two versions themselves. Jotform has no built-in A/B test; its own guide clones the form and uses an external testing tool.
Do I need a second form or a second link to A/B test in ClarNow?
No. Variant B lives inside the same form, so the link, the embed code, notifications and integrations stay the same. ClarNow decides which version each visitor sees and marks every response with its version.
Ready to try it? Create a ClarNow account, build your lead form, and add variant B on the Pro plan. The A/B testing guide walks through every step, and the lead form builder comparison covers the rest of what these tools do.
ClarNow details checked against the product on 17 November 2026. Heyflow: help center articles "A/B Testing with Heyflow" and "A/B Testing and Calculating the Winner", read 17 November 2026; prices from heyflow.com/pricing, read 10 November 2026. involve.me: help article "A/B Testing" and involve.me/pricing, read 17 November 2026. Jotform: "How to A/B Test Web Forms" on jotform.com/blog and the A/B test calculator page, read 17 November 2026. Found a mistake? Tell us and we will correct it and date the correction.