That runs counter to the prevailing notion in Web design that you should do everything you can to “make it really easy,” noted Josh Elman, Twitter’s product lead in charge of user retention, in a talk at the Web 2.0 Expo in San Francisco today.

“If people are willing to do a little work, and you can give them context, it will show them the value they can get out of the service,” Elman said. That 29 percent figure refers to users completing the sign-up process, something many people used to start and then drop out of.

Elman, whose goal is to make Twitter users stick around, said his team has identified an “aha moment” when a casual user turns into an “active user.” That moment happens when users follow 30 accounts, and when one-third of the people they follow also follow them back.

“Once we get them to a certain amount of visits per month, there’s a 90 percent chance they will do the same thing the next month, so everything we do is to try to get to that number,” Elman said. He did not disclose what the number was.

Elman’s team considers four types of users: New users (signed up yesterday), active users (used the service in the last month), inactive users (last used the service one month ago) and resurrected users (first use in over one month).

It’s interesting to hear how focused Twitter is on making sure that users are active, given that the company still publicly discloses only its total registered user number.

Elman, in fact, said that a big part of his technique is “user accounting,” understanding changes in active users just like credits and debits to a bank account. He also builds models to try to understand how a curious person becomes a casual user becomes a core user.

Without getting into specifics, Elman said, “what I’ve seen in our business is once negative growth happens it doesn’t happen slowly.”

Elman gave two specific examples of user experiences Twitter changed to increase retention: emails and captchas.

• If a user signs up and doesn’t engage with the service, Twitter sends them an email. It tested an email with all sorts of interesting content to entice them to use the service, as well as a simpler, friendly email with a single blue button that said “Hey, come back to Twitter.”

  The simple email turned out to be much more successful, with an 8-9 percent click-through reactivation rate. And what’s been even more successful is sending emails to users when other people follow them, Elman said.

• Part of the Twitter sign-up process is a captcha for users to prove that they are not bots. But 30-40 percent of the time, users got the captcha wrong; in Japan the error rate was as much as 70 percent. A new Twitter engineer came up with the idea that the first time someone signed up from an IP address, there would be no captcha. Only if another account was created from that IP address would a captcha appear.

  That increased the rate of signup 4.5-5 percent, with no noticeable increases in spammers creating accounts. Elman used this as an example of where novel thinking makes sense, rather than looking at measurements to make incremental changes.

CAPTCHA, for those of you just joining us, stands for Completely Automated Public Turing test to tell Computers and Humans Apart. Essentially, it’s a challenge-response test used to distinguish between humans and spam-spewing robots. What’s interesting about reCAPTCHA’s implementation is that it’s used for digitizing books.

“Since computers have trouble reading squiggly words like these, CAPTCHAs are designed to allow humans in but prevent malicious programs from scalping tickets or obtain millions of email accounts for spamming,” Google explains in a post to the company blog. “But there’s a twist–the words in many of the CAPTCHAs provided by reCAPTCHA come from scanned archival newspapers and old books. Computers find it hard to recognize these words because the ink and paper have degraded over time, but by typing them in as a CAPTCHA, crowds teach computers to read the scanned text.”

An ingenious idea, crowdsourcing book transcriptions in this way. An effective one too: reCAPTCHA boasts 99.5 percent accuracy at the word level.

Little wonder, then, that Google (GOOG) has acquired it. The company can clearly put reCaptcha’s technology to good use, not just as a security measure, but as a means of improving its own massive book-scanning project.

CAPTCHA, for those of you just joining us, stands for Completely Automated Public Turing test to tell Computers and Humans Apart. Essentially, it’s a challenge-response test used to distinguish between humans and spam-spewing robots. What’s interesting about reCAPTCHA’s implementation is that it’s used for digitizing books.

“Since computers have trouble reading squiggly words like these, CAPTCHAs are designed to allow humans in but prevent malicious programs from scalping tickets or obtain millions of email accounts for spamming,” Google explains in a post to the company blog. “But there’s a twist–the words in many of the CAPTCHAs provided by reCAPTCHA come from scanned archival newspapers and old books. Computers find it hard to recognize these words because the ink and paper have degraded over time, but by typing them in as a CAPTCHA, crowds teach computers to read the scanned text.”

An ingenious idea, crowdsourcing book transcriptions in this way. An effective one too: reCAPTCHA boasts 99.5 percent accuracy at the word level.

Little wonder, then, that Google (GOOG) has acquired it. The company can clearly put reCaptcha’s technology to good use, not just as a security measure, but as a means of improving its own massive book-scanning project.