SM: How do you get the context?
JJ: You have to emulate the brain. If you can emulate sensors and the neocortext, then we can behave just like humans. If you have human input on one hand and human behaviors on the other, then you can interpret human behaviors. You can give people what they want.
SM: Give me a use case of your solution.
JJ: Let’s think about Urban Outfitters’ e-commerce position. They sell dresses, jeans, and cameras. Users can browse, but they are not connected to their peers. If people can’t find the product they are looking for in three clicks, 95% of them will abandon the site.
Urban Outfitters has hundreds of thousands products to sell. That is a great long-tail store. We know that millions of people come to the site. They actually have a lot of common interests. There are people who care about shoes, but that is a pretty big, generic peer group. Men’s shoes and women’s shoes would then have different peer groups. Some customers want what everyone else has, others want something unique. Your behavior, based on browsing and search activity, will quickly tell you what kind of user you are. We will quickly hook you up with like-minded peer groups.
SM: How does this connection occur?
JJ: The peer group is already formed. It is clustered underneath. When we find seven people who have similar characteristics based on a particular product context, it is not noise. That is a small crowd. You as a shopper may go away, but your shopping spirit does not. Your engagement stays behind. When the next person comes that has a similar behavior, we are better positioned to show them what they want. That is how peers convince each other to buy things, automatically, without even knowing that they are doing it.
SM: In your example, what have you seen in terms of metrics?
JJ: A huge increase, a triple-digit percentage increase.
SM: What about Netflix? Have you tested with them? They seem to be looking for a good recommendation system.
JJ: We have been dancing with them for a while. The $1 million test that they have is more of a marketing thing. If you have an algorithm that can increase by 10%, it is worth more than $1 million. The competition is specifically constrained by their data set. In our case we would need a completely different set of data, and we would need to be on their site live.
SM: Where are you seeing adoption in the industry?
JJ: We have 250 different customers, mainly in a lot of different places. About 30% to 40% are e-commerce customers. We have people such as Motorola, TI, Juniper, NetApps, Campbell Soup’s recipe sites, DisneyFamily.com, and other customers. We have a lot of marketing B2B sites such as Intuit and QuickBooks. Sales support is also big because post-sales support is also part of the pre-sales. I won’t download the latest copy of TurboTax unless I know it solves my tax problem. Before Baynote went onto their site, the click-through rate was 15%. We went live two years ago, and went worldwide live in eight days, and the exact same metrics we were tracking jumped from 15% to 35% the first day alone. We are a learning system, so over time it went higher and higher and hit 70% to 80% click-through.
They decided not to stop there and went ahead and made their entire navigation community driven. Rather than re-design website navigation every three months, why not let it be dynamic every hour? If you go to their support navigation the links and manual buttons are context based. Before April 15th and after April 15th you will see tremendous flips. Before it is all about filing taxes, after it is all about when refunds will arrive.
This segment is part 6 in the series : Simulating The Brain: Baynote CEO Jack Jia
1 2 3 4 5 6 7