In early 2007, I left one of the hottest companies in Silicon Valley. LinkedIn had millions of users, top investors and was already profitable. Our 80 person team had grown from twelve when I’d joined in 2004, and was poised to grow by another factor of ten over the next 3-4 years. I’d been leading the LinkedIn analytics team, and for a quantitative Internet guy, the data opportunities don’t get much better than LinkedIn’s.

I left LinkedIn because I had an itch to start my own company, though I wasn’t sure what that company should be. Let’s do a little entrepreneurial counseling exercise. Here was my background:

• Mathematical and Computational Science degree from Stanford
• Built PayPal’s early statistical modeling technology for fraud detection
• Started teamrankings.com to rank sports teams and help needy gamblers beat the Vegas spread

And my demographic/psychographic characteristics at the time:

• 29 year-old married male without kids
• only marginally less touchy feely than Dick Cheney

Which of the following would you have suggested I build?

• a) a socially optimized search engine
• b) a cloud-based collaborative filtering system
• c) an algorithmically personalized form of social networking
• d) a high frequency, automated stock trading system
• e) a company with the tagline “motherhood, shared and simplified”

Yeah, I wouldn’t have guessed (e) either.

How we found moms (and moms found us)

In September 2007, I launched a Facebook application with Ephraim Luft, who was my year at Stanford. Our app, Circle of Friends allowed me to create one circle for my math geek friends, another for my developer friends, and a third for my sports analytics friends (ah, diversity).

The application took off pretty quickly, attracting millions of users within a few months, and I got a crash course on scaling a web app (tip: don’t use MyISAM for tables that might become huge, unless you prefer learning MySQL DBA tricks to sleeping). Thanks in large part to our rapid growth, we secured funding in January 2008 from several top “micro cap” investors — Mike Maples, Jeff Clavier, and Naval Ravikant.

Around the time of our funding, we accidentally did something very clever. We built a small feature to allow users to upload their own (tacky clip art) circle icons. This was well-received, and we soon started to prioritize the new circles/icons that were most popular. I added some Bayesian logic to tie circles to the appropriate demographic, and we soon had a product that would upsell “Drinking Buddies” circles to 20-something males, and “special friends with heart in my life…” circles to 16-year old girls.

Still, as satisfying as it was to see hundreds of thousands of people creating a “WARNING NAUGHTY WHEN DRUNK” circle (see image on right) with 15 of their naughty-when-drunk-iest friends, we had no illusions that we’d discovered the future of human social interaction. We had some traction and some interesting data, but Circle of Friends’ usage was about as deep as a backyard kiddie pool, and revenue prospects were dim. We weren’t quite sure how to proceed, so I did what I do best: dug into the data.

I noticed that several hundred thousand people had created a “Circle of Moms”. It turned out that those circles were way more active than any others — moms shared more photos, had longer and deeper conversations, and accepted more of one another’s invitations.

That was interesting enough to us that we started to think about building out a product just for moms. We asked some questions around consumer demand, revenue opportunity, and our ability to provide real value to mothers. We got deep answers to these questions; I’ll spare the details and summarize in one word: “yes.”

We launched Circle of Moms in October 2008, and haven’t looked back since.

Moms need community, empathy, support… and some kickass algorithms

One of the best things about building Internet technology is the opportunity to do work that touches millions of lives. But a million times ten minutes of mindless game play is nothing but ten million wasted minutes. On the other hand, effectively touching the lives of millions of mothers can help to raise a great next generation of humans.

When a mom wakes up at 3 AM for the sixth straight night, needing to comfort her toddler who had been sleeping through until morning, she needs both support and information. This is hardly a new problem, but the Internet has done relatively little to improve moms’ collective support system or knowledge. Fast forward a few years, and her seven-year-old is having trouble concentrating in school and Mom is concerned he’s falling behind. By combining data and community, we can help a mom — and indirectly, her child — through these tough situations.

Being focused on moms keeps us honest as technologists. Shocking though it may be, many techies — myself included — might on occasion build something that’s cool or interesting but not especially useful. But when you have an audience that actually needs your product, it forces you to keep your eye on the ball.

To fully address these needs, we’re able to combine two great styles of Internet product development. The Google-style algorithmic approach to product development is often great for spam filters, optimized search results, and ad targeting. Facebook-style social incentives are great at encouraging users to do “work” like tagging photos, translating text into different languages, and indicating their likes and interests.

I love this balance. In our case, it means that Circle of Moms is fundamentally about community and connecting people, but we’re not just a social site. To take the next step beyond social, we aim to provide our users with good information and guide them in the right direction based on our understanding of them. Social cues encourage our users to share more information about their kids; we then use that information to improve their content experience in emails and on the site. Likewise, we derived a list of important milestones children accomplish via moms’ contributions; we then algorithmically customize the list each user sees based on what we know about them.

When we decided to build out Circle of Moms, we made a conscious decision to constrain our problem space. That makes this sort of pragmatic approach to technology a lot easier. We can focus on building a playgroup finder for your area, a baby name chooser tailored to your preferences, and a guide to the development of your child, all as separate products.

Like LinkedIn, we’re a vertical social network with an awesome data set. We discovered that at 18 weeks, moms on the East Coast are 40% more likely to give their children solid food than moms on the West Coast. We learned that San Francisco has lots of babies but very few school children. And our data tell us that conservative moms name their kids Reagan and Sarah, while liberal moms prefer Jalen and Jada.

The examples above are interesting stories, but they’re just a start. Connecting moms to knowledgable peers with shared experiences and values is a hard problem, and to get it right we need to be successful across multiple avenues. Parsing conversations for keywords and underlying meaning (which we do) is one step in that direction; getting users to self-identify in intuitive ways is another. One thing that defines us as a company is that we keep iterating in data-driven ways to create a great product and help to solve those 3 AM problems.

And fortunately, I work with an awesome group of engineers who are rapidly pushing our technology forward to build useful products for moms. Brian‘s a low-key dad and a brilliant engineer who does everything from architecting and building complex search systems to writing email copy. Chris is a gold-buying, Tetris-dominating, cloud engineer extraordinaire, who uses the latest ec2 technologies to automate everything except his daily five cups of coffee. Regina, our in-house yoga master, has an awesome mix of front-end and back-end web development skills and puts together different technologies in clever and scalable ways. Hoi Ying is a fun and talented young developer, who’s become a force for us as we beef up our back-end technology (and our Hello Kitty collection!). Emma is a sponge of new information, with a great mix of engineering, product, analytics, and marketing talents: she crunches numbers, builds features, and improves the content our users see.

Collectively, they — and the rest of our eighteen person team — have made us profitable, reached millions of moms, and created a product and community that’s helping both moms and the high school class of 2027. That’s not what I expected when I left LinkedIn four years ago, but it’s pretty darned rewarding.