the best price of a given item, it could be much more efficient for buyers and sellers alike.
That’s where artificial intelligence comes into play. Marot and Botvinik looked into techniques involving game theory, neural networks, machine learning, and other sophisticated ways of determining prices given market conditions. They settled on a technique known as reinforcement learning, whereby the software continuously updates the price of an item in order to maximize a long-term “reward”—in this case, the expected revenue to the seller, as well as cost savings to the buyer. (You might think it would be either one or the other, but it’s not—there is an equilibrium point at which both can benefit.)
This technique, Marot says, “doesn’t require a lot of calculations. It scales well.” Also, it avoids some pitfalls of other machine-learning approaches that might fail to maximize revenues because they get stuck in a bad loop and keep refining the price even though it’s way off. “We think it’s better to be approximately right than precisely wrong,” he quips.
Valu Valu is opening its beta trials on a very specific market segment—used video games, which Marot says is a $2 billion annual market. As Marot explains, he wants to start with standardized products, like Xbox and PlayStation games, that have lots of active consumers. He is aiming for the sweet spot between extremely high-selling products that everyone buys and the long tail of rare, collectors’ items that are best-suited for online auctions. “We need to build traction and prove ourselves,” Marot says. “We want to go from small guys and move up to businesses.”
The goal is to bring a 5 percent increase in revenue to businesses across the board. So Valu Valu won’t tailor-make an analysis for Nordstrom or American Airlines, say, but it will offer its services to smaller players who want a modest bump in sales. From there, depending on how successful the company’s pricing engine is, the aim could be nothing short of reinventing how everyone makes purchases on the Web. “The potential for improving commerce is gigantic,” Marot says.
