Robin Mellstrand

So he basically ended up

He had, he's done one sale in the company's history and he talked like a company called bubbler, which is basically the local copy of Netflix.

If you would say back then they don't exist anymore.

And he pitched them saying that, yeah, I can be the exact same recommendation.

And you know, Netflix has, because back then it was all, everything was in white papers was a big, big bus around that six or seven years ago.

Okay.

So you said 2012, 2011, I think.

And he said, yeah, I can build the same recommendation as they had.

And it did, and obviously worked really well on Netflix data, public data.

Problem is that all of these kind of algorithms are built around you having massive amounts of user data.

So they have their global company and they have lots of behavior data, basically.

So it worked perfectly for them.

But when we implemented it at Butler, everything fell apart because they're a regional company with not as much data as everyone else.

And then all the theory behind this algorithm

fell together basically so we gave him the product saying this is how you do it use it when you get bigger and then afterwards since we didn't have a new product for our mathematician he sat down and thought all right is there any way else i can understand how products fit together without knowing anything about the users so we applied new type of mathematical research in this particular problem and found a way to do that long story short it turned out to be quite a horrible recommendation engine still but

beginning to really interesting search engine.

Yeah, right now we are, have about a billion a month.

So we really did a bit of trial and error a couple of years and we launched it middle of 2013.

And then we changed it on the company.

We sold off the customer stock for peanuts because, yeah.