Thought Leaders in Big Data: Omer Artun, CEO of AgilOne (Part 4)

Sramana Mitra: What e-commerce systems do they work with?

Omer Artun: This doesn’t matter to us. It is a rest API.  Any e-commerce system can pull the data out of it. If you think about the day in a life of a marketer, we are trying to do other things than just starting up recommendations or cluster IDs.

SM: I am talking more about a personalized store. That is one of the ways to use this data. The other way to do the data is to run different kinds of promotions whether these are email marketing, coupons and discounts, offers, catalogs, bullet marketing pieces, and so on. From an organizational point of view, it is marketers and merchandisers that use your technology, right?

OA: Yes. Sales ops do as well. For B2B cases we see the salespeople as a channel as well, just like web, email, and catalog. The salespeople also need the information.

SM: How is this system architected? Is something sitting in the cloud in your data center and the processing is happening in a public cloud, or are you being asked to put this on private clouds for various customers? How does the deployment work?

OA: It is a private cloud environment with multi-tenancy, but the multi-tenancy doesn’t mix customer data together. Basically, it is all in the cloud.

SM: Would you talk a bit about specific customers and what kind of impact you are having on their business?

OA: Sure. But I think I haven’t described a solution to you yet. We went down the clustering path. Let me pull back up and tell you what we do.  As we get data from different sources, this data is usually not in a great form to do any sort of analysis on it. So, we bring this data from different sources – we have a very powerful innovation engine that brings us data on a near real-time basis and cleanses data like names, phone numbers, addresses and geocodes it, etc.

All of this is done automatically on our platform. That is the foundation for adding intelligence to the data. Aggregation algorithms work based on that cleansed data to add intelligence to it like clustering, propensity predictions, and so forth. This is what we call a smart data hub. It exposes that to a set of features that we have on the application layers, which then can be used by the marketer. The user experience is designed around solving a problem for the marketer. How does it start? When I come in to the office in the morning, I go into the application and see a bunch of alerts the system is showing me. It might say that I used to acquire 1,000 customers from XYZ affiliate, and that has dropped to 500 over the past two weeks. This is a significant drop that is going to cost me a certain amount of customers over the next certain number of days, and I should do something about it. Another example of an alert might be: I have this many high-value customers that are about to lapse. You should be able to understand why they are lapsing. Once it gives you this alert, it prompts you to take action on it. The next question a marketer would ask is: why? Why is this thing happening? So the next two applications we have are the metrics application, where you can look at your traditional DI in order to see who these customers are and when was the last time they bought, their value, what is their affiliate, the way they had been trending over time, etc.

I can also go into our pathway application, which is the interworking of the black boxes we have. If I have a turn problem, for example, it factorizes that problem and tells me why people are turning. It is all about preparing the marketer, coming up with ideas about the situation, whether it is good or bad. Once you are equipped with that, you can consider your options. You can take all your high-value customers who are about to lapse and find out which of them live within a 20-mile radius of a store, or which of them have outstanding quotes, so your salespeople can call them up and offer special deals. You basically want to take that intelligence and act on it. This is where our T-60 profile and action applications come into play, where you can look at individual customers and understand their behavior. This action application is a campaign automation tool that allows you to disseminate this information to various marketing channels like email, SMS, direct mail, websites, Salesforce.com, etc.

One of our customers is PetCareRx. They are a private company so I can’t talk about their revenue, but they are a sizable e-commerce operation. I believe they sell pet medicines. Last year introduced different product lines like pet food, which they didn’t sell before. Their average order value (AOV) immediately dropped – they started selling a lot of pet food products. Our system immediately picked that up and said: you have a huge drop in your AOV, your product mix is changing, and you are acquiring far too many customers. This immediately created an alert. They went to the DI tool and started understanding that the new customers they acquired are actually coming from food categories, and they are buying stuff at a much lower order value than their original buyers. This way they started to understand that they built a new business model on top of an existing business model, and that the AOV drop was because of this. The system then automatically picked those customers up and put them in a different cluster. Now they can start treating those people differently because the detention rate of the pet food-buying people and the regularity with which they buy is higher. This way they got a much better handle on their customers than they would have had if they had just ran a bunch of reports, where they would have had to face a much higher effort by hiring a bunch of analysts to figure out what was going on.

This segment is part 4 in the series : Thought Leaders in Big Data: Omer Artun, CEO of AgilOne
1 2 3 4 5 6