How to check if your CAPTCHA is losing you enquiries

Why you can’t see the loss today

When a customer fails a CAPTCHA or gives up, nothing happens: no error reaches you, no message arrives, and your analytics show nothing unusual. You only see the messages that made it. Independent studies found that people solve image CAPTCHAs correctly only 71–81% of the time, and that a large share of people abandon a task when a CAPTCHA appears [1]. The steps below show you your own version of those numbers.

Step 1: count starts against sends

You need two numbers for the same period: how many people started filling in your form, and how many sent it.

• Google Analytics 4: its enhanced measurement can record form interactions as form_start and form_submit events. Check in the real-time report that both fire when you use your own form, because some forms aren't detected automatically.
• Your developer can add the two events directly: one when someone first types in the form, one when it's sent successfully.

Sends divided by starts is your completion rate. A big gap doesn't prove the CAPTCHA is to blame (people change their minds for many reasons), but it is your starting point.

Step 2: read your CAPTCHA’s own numbers

If you use Google reCAPTCHA, its admin console reports, per day, how many visitors passed without a challenge, how many challenges were passed, how many failed, and how many sessions failed; researchers used exactly this data in their 13-month study [2]. Failed challenges are a mix of bots and real people, so you can't separate them there, but a high number is a warning sign.

Step 3: try it as your customers do

Spend ten minutes being your own customer:

• Send your form on a phone, on mobile data, not your office Wi-Fi.
• Try it in a private browsing window, where you look like a new visitor. People without Google cookies get image challenges more often [2].
• Try it with a keyboard only, and if you can, with a screen reader.
• Ask an older relative or a colleague who reads English as a second language to try it without help.

If any of these is difficult, some of your customers are having the same experience.

Step 4: run a fair comparison

Now compare two periods of the same length, such as two to four weeks each:

1. Period A: the form as it is today, with the CAPTCHA.
2. Period B: the same form with the CAPTCHA replaced by invisible protection (a hidden trap field, a time check, a signed token, link rules and a rate limit), with a spam log that records everything it blocks.

Don't simply switch protection off to test: you would be comparing your CAPTCHA with nothing, and a flood of spam would spoil the result. Keep everything else the same, including advertising, prices and the season, or run the two versions in alternating weeks to even out the differences.

Step 5: read the result

For each period, write down:

• Real enquiries per 100 visitors to the page.
• Spam that reached you.
• Attempts blocked, from the spam log in period B.

If real enquiries rose in period B while spam stayed about the same or fell, your CAPTCHA was turning customers away. In an older test on business websites, Moz found that a CAPTCHA cut spam but made 159 submissions fail that would otherwise have gone through [3]. Your numbers will be your own; that is the point of measuring.