Doug Randall: The first thing we do is we look for narrative richness. This is text that has high levels of emotion and that is not simply mundane and descriptive. Mundane and descriptive is, “I went to Starbucks yesterday.” Emotion is, “I like my green tea latte because it makes me feel good.”
We then use that to find what are the underlying claims or beliefs that are being expressed. Those component parts of the narratives are what we’re analyzing. Once we have those, we can start to look for similar claims and we can assign measurements to them. If we see a New York Times editorial about the problems of sugar, we can find all of the places where we see that claim and we can attribute a resonance factor – how often that issue was viewed.
We then use this to cluster some more claims together and to come up with a set of narratives. Each of these narratives is a set of claims or beliefs that logically fit together to explain what someone’s underlying belief is about the issue. We tend to wind up with something like 8 to 10 narratives. Once we have those, we now have an understanding of the beliefs and we have our analyst go into the data and ensure that we’ve got unique narratives. They then write those narratives up as first person descriptions of what’s happening.
Sramana Mitra: That’s a manual process?
Doug Randall: Yes. It’s enabled by technology but it’s human processed. Now we’ve got all these clusters. Humans verify the claims. That now becomes our model. Now every future piece of data that enters gets classified as belonging to one of those claims. This is how we’re all to track how the system works. That’s where machine learning kicks in because the computer is watching how humans tweak the classifications. It’s able to connect data pieces to claims to narratives.
Sramana Mitra: So what happens next once you’ve classified? What are you trying to do with that classification?
Doug Randall: What the classification allows us to do is get us a measure of resonance. We’re now able to tell how often that claim or belief is being stated or shared. It’s able to tell how often does that narrative around the scientific attributes of sugar appear versus the narrative about the need to appreciate indulgence as a part of life as opposed to chocolates having positive characteristics to it.
That’s the fundamental question that people have. How strong is each of those narratives? The next question that you have is who’s behind those narratives. When do they start to increase or decrease? What’s driving them? We’re able to use that data to understand that stuff. We can say, “We see the science narrative primarily in scientific publications that are industry professionals but we’re not seeing it moving into the mainstream. We’re seeing the indulgence narrative more common in the millennial.”
We’re able to start to take the stories and analyze them by sources and by influencers. It helps us figure out what’s happening so that we can change the system if we want to and so that we can change our strategy if that’s what we want to do.