At Zynga, Analytics is About Impact

Anyone can play Zynga’s FarmVille and other games, but very few get to analyze the data that runs off.

Ken Rudin, VP of analytics and platform technologies, is always on the lookout for new talent, but he’s picky. What does he look for in an applicant?

Anyone can play Zynga’s FarmVille and other games, but very few get to analyze the data that runs off.

Ken Rudin, VP of analytics and platform technologies, is always on the lookout for new talent, but he’s picky. What does he look for in an applicant?

“Knowing what questions to ask,” he says. He might present an applicant with a problem his staff is puzzling through and asks what approaches to take. Say the company has invested in a some game, and even though it did well at first, growth has flattened out. What would the applicant ask? A good answer is, “It flattened out compared to what?” A not so good answer might involve degrees of standard deviation.

Contrary to the widespread assumption, most data analysts in business don’t need much knowledge of statistics. Far more important is knowledge of business. That’s the second qualification Ken looks for.

He wants business aptitude. “Ninety percent of what we do is making a business recommendation,” he says, to answer simple questions like whether to invest more or not?

What kind of analyst rises to the top at Zynga? It’s someone who can persuade someone else to change course. “In our company, if you have brilliant insight, and you did great research, and no one changes,” he says, “you get zero credit.”

People are busy. They have barely enough time to read all their email, much less spend precious half hour reading an analyst’s report telling them what to do. So Ken’s team routinely makes “sales calls.”

One recent insight took several such visits to the game team. In Zynga’s games, players send gifts to others. But if you send 10 gifts, for example, and only five of your recipients open it and only one sends you a gift, you might feel frustrated. The analytics team figured out how to help players judge who was more likely to respond.

Once an insight has been sold, it needs followup to stay sold. The game team might try it out for a week, decide it’s not working, and rip it out. To make it work, the analytics team may have to adjust the models, or whatever it takes.

“Analytics is not about insights,” he says, “Analytics is about impact. If no one changes behavior, there’s no impact.”