Framing items as a question without a strong action verb is not as compelling, I think. The username policy is not a suggest tool, but does in fact make recommendations about what your username should be, will help people choose one.

That copy change is a great idea. “Help me choose” is much more inviting, and if people get errors (duplicate username, etc.) it’s appropriate for that case too.

Could the “policy” be rebranded as “How to pick a username” or “Help me choose”? Not only does that seem in line with the newer, friendlier look, but it might get more people to read it/sign up.

As outlined in the Meta page about our data collection and analysis, we’re going to measure the impact of removing that text by getting A) how many users from control and test groups were blocked B) filtering those blocks by the log comment, to look for the rationales that are related to username policy violations. In comparing the two groups, we’re essentially looking for what percentage of usernames need to be blocked.

As for cost: I think it’s pretty clear the user experience of the signup process is broken and ugly for those who have to try and navigate it. Fixing that is priority number one, and doing nothing is not an option in my view. That definitely doesn’t mean that whether we implement a particular design is not checked by whether it creates an untenable and unnecessary workload for the community. Ultimately, the data will hint at whether the messages are really preventing a significant number of people from registering with bad usernames, or if they just depress signups overall.

We haven’t yet added any smarter validation to the username field, but we’ve got a basic spec up on MediaWiki.org. First step is just making the fields validate on the client side, so that we can deliver _any_ kind of message before people submit the form.

A lot of the text on the current account creation page has been implemented as a means of stopping (or trying to reduce) bad account creations. For example, the username box explains in great detail various prohibited usernames (domain names, e-mail addresses, etc.).

Will you be measuring the effect of these UI changes on the number of bad account creations? If so, how?

And more importantly: what level of cost was considered acceptable before choosing to experiment in this way? That is, it’s completely possible (and perhaps even likely) to see a marked increase in the number of blocks issued for username violations. This creates a hostile atmosphere for new users and wastes volunteer administrator time. What level of cost was considered acceptable before you embarked on this experiment?

Finally, are there any other (possible) issues that you’ve anticipated or attempted to mitigate before starting this experiment? Is the username field now “smarter”?