Thought Leaders in Big Data: Dmitri Williams, Professor USC and CEO of Ninja Metrics (Part 4)

Dmitri Williams: I have no idea if I impacted sales because I have 10,000 followers. I don’t know if I have generated $1 or $3 million. I’ve got no clue. It’s all an assumption. You’re really looking at a proxy for action. Any marketer with their head on straight would rather have proof. What we do is much more difficult but it results in proof. We are not a business-to-consumer technology. We are a business-to-business technology, which means that we have the buying data. We will connect the social graph to get that social whale phenomenon measured. We will connect it to the actual purchases or actual buying of tickets. We will link those two sources of data together – the social and the behavioral. As a result, our scores aren’t scores. They’re in the units that the company cares about.

If this was, for example, a streaming video technology, the units of social value could be in dollars of subscriptions retained, created, or rippled out. In other words, it’s transparent, provable, and repeatable. If I say that you’re going to watch 20 shows next month and that you watching them is going to cause your friends to watch another 50 episodes in total across your graph, that’s a claim that I can now measure. Then next month, I can compare the prediction versus the actual and I could say, “My model was 62% accurate.” Therefore, it’s not a proxy. It’s not a “take my word for it”. It’s science and it’s proof. When you do that, all of a sudden people pay a lot more attention. The business model can be based much more on performance rather than on promise.

Sramana Mitra: Now let’s talk about the process of being able to do this. You are working with purchased data or behavioral logs. Those are one of the data sources that you’re working with?

Dmitri Williams: Yes.

Sramana Mitra: Other data sources, I imagine, would be social media data sources.

Dmitri Williams: It depends on the sector. We could start in video games where the sourcing for a social graph is different. If I go to an online shoe e-commerce retailer, that’s not a social hub. The people shopping there are very rarely doing things with each other on the site. There’s no evidence that shopper A is connected to shopper B. For them, how do I have a graph? I don’t unless e-commerce companies decide to use social sign-in as part of their registration path. When we go and talk to an e-commerce company or a social sign-in enabler, these are all companies who have a ton of social graph data but don’t really know what to do with it. They simply think, “I’ve got more information on these people because I’ve got their Facebook profile.” What they don’t realize is they have their friends list, which means they can construct a social graph.

We use and automated API system and, on their behalf, take these tokens, create a graph, and that’s the one half that we need. Now if I’m doing this with video games, you don’t even need that step because that’s an inherently social activity. You just check the session data.

From this, you can tell that there are some sectors where it’s very easy for us to walk in and help and other sectors where they’re five to 10 years from us being able to do anything. You could also see the appeal of trying to tie communication graph data along with these things. In some sectors, it’s early days. In others, we have everything we need.

This segment is part 4 in the series : Thought Leaders in Big Data: Dmitri Williams, Professor USC and CEO of Ninja Metrics
1 2 3 4 5 6 7