The power of social data

from influx insights:

Interesting move by Walmart

September 14, 2011

Yesterday, Wal-Mart made an acquisition of a company One Riot.
It’s a purchase that could easily escape most people’s radar, since the company is tiny and focused in a very specialized area.

On paper, the company is in the mobile space and provides ad targeting technology, but if you dig deeper by simply reading their blog posts- you get a good idea of why Wal-Mart purchased the company.
Here’s an extract from one of the company’s blog posts.

“Based on publicaly available social data that we license via our friends at Gnip, we’ve generated user interest maps for just north of 75MM active Twitter users. This number gets bigger every day as we suck in more Tweets from the licensed hose. Meanwhile, our demographic coverage is 80% accurate for about 80% of the Twitter population. These stats improve with every new release (approx every 3 weeks).


How to do we determine this information? We start at two points. The first is to create language models that breakdown how users tweet. The second is to study follow-graphs that reveal who users are influenced by. By fusing these two approaches, we can find discriminating characteristics that identify a specific user as belonging to a certain demographic group or a certain interest-category group.


For example, it turns out that Hispanic 25yr old males tend to tweet “#LMAOOOO” (emphasizing the “O”) more predominantly than other groups. Meanwhile 40-something females tend to follow soppy singer @joshgroban at a statistically significant higher rate than males of any age. Niche influencers such as@CrankyOvary provide very strong signals for the interests and demographics of their followers. Meanwhile, broader-based influencers such as @barackobama offer few discriminatory clues about their followers – as many users in many demographic groups follow the US President, to the point where any signal is drowned out by noise. Our system spots literally tens of millions of similar subtle patterns. These patterns then “compete” inside an artificial intelligence system that results in an accurate rendering of “who you are” according to your Twitter activity. It’s good stuff – lots of large scale Machine Learning, Natural Language Processing, Statistical Analysis and Classification Techniques worthy of subsequent blog posts, all sat on top of a humming Hadoop cluster.”


Wal-Mart isn’t buying a team and a sophisticated database of consumer behavior that will help them to understand the dynamics of the social geography that surrounds their stores. This could help enable them to do obvious hings like reach out to local influencers, but perhaps more relevant would be re-configuring their store merchandise around local needs.
It also reminds us all what a rich source of information social media data can be- if you have the talent and bandwidth to analyze, it’s incredibly powerful source in to help us who consumers are and what they think and do.
This is an example of the big data that will fundamentally change and disrupt not just the media business, but marketing as we know it.