A/B test form fields and placement

Start with the form that gets the most visits and leads to the most revenue. For most B2B sites that is the demo or contact form, not the newsletter box. A test on a form with 50 visits a week will never finish.

Write down its current numbers first: weekly visitors, submissions, and the share of submissions that become a booked meeting. That is your baseline. If you cannot get these three numbers, fix the tracking before you test anything. Google Analytics with form submit events is enough.

Do this sum before you build anything. A rule of thumb helps here. If your form converts at 3 per cent and you want to detect a rise to 4 per cent, you need roughly 5,000 visitors on each version.

If your form sees 400 visitors a week, that test takes more than six months. In that case, do not run a classic A/B test. Make one bigger change, such as cutting the form from seven fields to three, and compare four weeks before with four weeks after. It is less precise, but it is honest about what you can know.

A test without a hypothesis is just a change. Use this template:

"If we [change], then [metric] will [rise or fall] by [amount], because [reason]."

For example: "If we remove the phone number field, submissions will rise by 15 per cent, because buyers do not want a sales call yet." The reason matters. When the test finishes, it tells you what you learned even if the variant loses.

Three kinds of change are worth testing, in this order:

• Field count: seven fields against four, or four against two.
• Form format: one long form against a two-step form where the email comes first.
• Placement: form in the first screen, form after the proof, or form in a side panel that follows the scroll.

Leave button colour and microcopy until later. They rarely move the number enough to detect. The principle of changing one thing at a time is covered in Test the capture mechanics, do not redesign them on a hunch, so this chapter stays on how to run the test.

Use a testing tool such as VWO, or the form tests built into HubSpot or Unbounce. Split traffic 50 and 50 at random. Check that both versions load at the same speed and that the thank-you event fires on both. A broken tracking tag on one version is the most common reason a test lies.

Also write down the stop date and the minimum number of visitors before you start. Put both in your test log.

Run the test for at least two full weeks, even if you hit your visitor target sooner. Traffic on Monday differs from traffic on Saturday. Do not look at the result every day, and do not stop when one version is ahead. Early leads often vanish.

Do not launch an ad campaign, a newsletter or a pricing change in the middle of the test. Anything that changes who arrives also changes the result.

A short form nearly always gets more submissions. The real question is whether those extra people turn into meetings. Look at the booked-meeting rate for each version, not only completions.

If the short form gets 20 per cent more submissions but the same number of meetings, you have added noise for your sales team. Keep the version that produces more qualified conversations. Where the form leads to a booking page, check Meeting setup too.

Use one sheet with these columns: date, form, hypothesis, versions, visitors, result, decision, what we learned. Roll the winner out to your other forms, and record the losers too. In a year, this log is worth more than any list of best practices.

• Testing a form with too little traffic.
• Stopping the test as soon as one version looks ahead.
• Changing several things at once.
• Measuring submissions and ignoring meetings.
• Never writing the result down.

After three months you should have at least two finished tests in the log, each with a clear decision. The form you test should convert better than the baseline, and the booked-meeting rate should be flat or up. If completions rise and meetings fall, the test taught you something, and you should say so in the log.