Harvard Prof: Marketers Need to Step Up Their Predictive Abilities
Statistics expert Edo Airoldi says data must be paired with predictive analytics before marketers can truly forecast customer behavior.
Airoldi in his Nanigans team picture.
Marketers need to get more clinical to fully realize the benefits of their investments in data and technology, says Harvard statistics professor Edo Airoldi, one of the developers of the Nanigans ad automation platform that allows companies to do their media buying in-house. Brands that use social graph marketing to track people's social media behavior and inform their future campaigns can have some luck in predicting their future behavior, he says, but luck it may well be.
“It's making predictions based on correlations and not causal effects. Drug companies could never get away with that. They have to undergo clinical trials,” Airoldi says. “For instance, a baby products marketer can observe a huge stork migration and assume that a lot of babies will be born. But one day the stork becomes extinct and there are still babies. It's the same thing with the circle of correlations on Facebook. You can use them to predict forward, but it's problematic.”
Two years ago Airoldi teamed with brothers Ian and Jeremi Karnell to found Offergraph, a company that provides marketers with a predictive analytics tool to inform future campaigns by accounting for cause and effect. Offergraph was acquired today by social media data company Polygraph Media, and Airoldi, who will serve as the merged entity's technical adviser, thinks that such combinations of data and predictive analytics platforms will make machine learning accessible to all marketers.
“We're at the early stages of helping marketers become truly data-driven,” Airoldi says. “There are a lot of tools out there that help marketers do things—like text mining—to analyze social media feedback, but these tools are immature relative to what is possible.”