The formatting reduced the error rate on the credit card number field by 15%.
Our data showed that the employer and occupation fields had one of the highest error rates of all the fields. When we looked at the values people submitted we saw things like "none of your business" so we assumed that the reason the error rate was so high was because people were uncomfortable providing this information. However we learned through user interviews that retired people and students were unsure what to put in the fields. To solve this problem we added one lin below the fields that read "If you are retired, please enter 'retired' in both fields."
The one line hint reduced the error rate on the employer field by 63% and the occupation field by 58%.
While spending so much time trying to reduce error rates on our donate form we noticed a latent affect. After one of our users made a donation they were taken to an upsell page which asked them to save their payment information. We noticed that when we reduced the error rate on the donation form, we got up to a 7% increase in conversions on the upsell page, without touching the upsell page at all. This is one of the data points that justified spending so much time improving the user experience of our forms.
The qualitative data that user testing provides made a big difference both to our users' experience and to our conversion rates. We found out that a lot of the assumptions we made about our donate forms were wrong and in hindsight I'm glad we challenged our assumptions. Reducing user frustration with your products is a good thing to do for your users, but it can also help achieve business goals. User testing is worth every minute of your time.
Comments here: http://news.ycombinator.com/item?id=5133760