We've got plenty of argument that CAPTCHA can easily be beaten, and that they introduce usability problems to users. People say that because they can be broken, they shouldn't be used.
So if you've chosen to deploy CAPTCHAs on your forms, do you keep track of the number of users who fail (and who are therefore either robots or insufficiently advanced humans)? What rate of failures occur? Is the likelihood of failure high enough to justify the effort (and cost to legitimate users) of deploying the CAPTCHA?