SAN FRANCSICO — Everyone is trying to personalize advertisements based on the user data they have. However, this approach has a limit. Online shoe and clothing retailer Zappos shows that bringing in third-party data like weather forecasts can open up a different world of insights.

At GrowthBeat 2014 today, Zappos head of demand generation Lisa Archambault talked about her company’s collaboration with programmatic media buying company ChoiceStream in its weather-based customer advertising campaigns.

“We have been thinking about weather for years. We know people care,” said Archambault. “We want to make sure if it’s gonna rain in your neighborhood, how can we stay personal with you and show you something that’s relevant for you.” Archambault said that Zappos approached a few companies and that ChoiceStream was the first that said yes.

Eric Bosco, the chief executive at ChoiceStream, said that there are three things to take into consideration in such a campaign. One is about the consumers. Who they are? Are they men or women? Younger, older, stylish, or conservative folk? The second is what shoes or clothes from Zappos appeal to these consumers. Third is what’s the weather in the customers’ neighborhood. ChoiceStream used a weather forecast API for that. “All three of them have to come all at once in real time,” said Bosco.

AI Weekly

The must-read newsletter for AI and Big Data industry written by Khari Johnson, Kyle Wiggers, and Seth Colaner.

Included with VentureBeat Insider and VentureBeat VIP memberships.

The result was fantastic, according to Archambault. But she didn’t give exact figures.

And in fact, Zappos is a data-centric company. As orders come online and into the call center, its analytics team deals with the data and pushes reports out in real time, according to Archambault. She added that Zappos’ analytics team not only looks at data after the fact but also participates in conceiving tests.

In terms of what kind of advertising technology partners Zappos prefer, Archambault said that the offering of “multiple level we can pull in real time” is important as well as the capability to measure “negative sentiment.”