Sramana Mitra: Let’s do more use cases. I’d like to understand how your product is being used in other contexts.
Harry Glaser: Another of my favorite customer is Crisis Text Line. They are first responders. Folks who are in personal crisis can text a trained professional who will go through that moment of personal crisis and get them to a better place. One of the things with Crisis Text is they use machine learning in order to suggest responses to their first responders. These responders are literally saving lives with Crisis Text Line.
If you can suggest the right response to the right responder, you can have a meaningful impact there. They are training machine learning models in the product that suggest better responses to their first responders so those first responders can save lives. They are able to look into their dataset of what has worked in the past and train different models that will help the first responders choose better responses in the future. They’re able to do that in the product so they can have a meaningful impact.
Sramana Mitra: I can see the data angle. What is the Big Data angle? Can you differentiate between data and Big Data?
Harry Glaser: Some of that is just what’s the size of the dataset we’re working on. Another customer is New Relic. New Relic is a public tech company with several thousand employees. They make server monitoring software. They have a marketing-driven business. They do a lot of different marketing programs that will drive traffic to their website. Some of that will convert into leads for the sales team. The sales team then works those leads and closes them.
They get many millions of hits per day. They are doing analysis on which marketing programs are converting into most traffic. Which traffic then converts into leads? You might have a marketing program, for example, that delivers a lot of traffic but much of that doesn’t actually convert into leads.
Then you may also have leads that are better or worse at closing into customers. You need to be able to operate on this very large dataset in almost real-time. They’ll want to make day-to-day and week-to-week decisions on which traffic sources are the best and where to double down.
Also, they’ll want to be able to run experiments with scaling these different channels. Just because a marketing channel has really profitable characteristics doesn’t mean that it will scale up well. You want to be able to tag those sources of traffic, try to scale them up, and see what the scaling properties are.
Based on that, you can actually build different models of what the future state of the business will look like if you invest in different marketing programs. In order to do that, you need to be able to crunch that size of data. You need to be able to run analysis on that data.