You are here:  Home  >  Uncategorized  >  Current Article

Machine Learning May Be Online Marketplace Sales Magic

By   /  June 12, 2013  /  No Comments

When it comes to selling in online marketplaces, it’s all too often a race to the bottom. There are plenty of rules-based tools out there to help sellers on Amazon or eBay beat competitors’ lowest prices by the proverbial penny, but winning the deal in that way sacrifices flexibility and profits.

That’s Feedvisor’s take on the topic, and it’s why the company recently came out with a machine-learning based approach to setting prices for Amazon Marketplace sellers. It says it’s the only re-pricing solution on Amazon to leverage the technology in Version 2 of its Algo-Pricing software, which aims at helping sellers win buyers without necessarily driving them into pricing wars.

“We came up with the idea of looking at artificial intelligence because we realized that trends — such as a specific item becoming popular very quickly or sellers running out of stock — play a very, very important role in predicting upcoming price changes,” says Feedvisor director of marketing Shmuli Goldberg. “The market itself is constantly changing. Just looking at a snapshot of what is going on right now is extremely useful, and it’s what we’ve been doing until now. But if you have just a bit of historical context and the ability to predict upcoming trends, you can do things that other re-pricing software hasn’t thought of, like raising the price of a product you see is just about to become popular.”

The Amazon Buy Box, where multiple variables – not just the lowest price – inform which third-party seller will be featured, help Feedvisor with context and direction. Feedvisor, an Amazon technology partner that uses the online giant’s API for all its information, gets the goods not only of who’s selling something cheapest but competitors’ ratings, shipping times and costs, fulfillment methods, and so on, to set a price. Hundreds of millions of data points are involved in the process. With the machine learning element onboard, “every time we re-price an item, and we do it over 100 times a day on most items, we use machine learning to gauge how effective that re-pricing was [for instance, did it encourage or discourage sales], and this learning takes that into account for all future re-pricings of that same item.”

On a basic level, it constantly gets better and smarter about what the market will bear, based on what it has seen – and based on what it is starting to see. For example, during the Christmas season, if it sees a particular children’s toy is in massive demand and that, during its hundred checks a day, it is starting to sell at a higher cost, it will start to increase prices for its customers who offer the same product. Or, if it sees that competitors are running out of stock, even for just a half an hour, it will instantly raise prices for its customers to take advantage of the fact that they’re among the few who still have the item available.

With machine learning, it can learn from such data that it observes. “It is constantly watching and improving itself and taking advantage of any changes or fluctuations within the market itself,” says Goldberg. In contrast, he says, rules-based systems can get very complex; while they can support as many rules as are wanted, they are still confined by their rules and can’t react to new market conditions.

For its customers, the result may be that they’ll increase profits with perhaps a tiny reduction in volume when prices get higher, or realize a big increase in sales volume when prices are lowered by a very small amount. Goldberg says that the machine learning capabilities were the work of a team of Feedvisor mathematicians who are basing a lot of what they do on the same principles that apply to the technology’s use in stock markets. The addition of “the ability to notice a pickup on trends, to account for historical data, and any other fluctuations, and more importantly the ability to constantly improve itself — almost to second-guess itself to guess at how well a certain change will perform, then check how right or wrong it was — for every single product it re-prices lets it develop its own set of perfect conditions,” whether that’s being more or less aggressive at any particular time.

The solution is cloud-based, so Feedvisor essentially can run an unlimited number of items through its system. The company is bringing users onto the new platform at a controlled pace. And, says Goldberg, Amazon may only be a starting point. “The algorithm itself can be applied to any controlled marketplace,” he says, and to that end it’s working on building relationships with sites like eBay and others in that vein.

About the author

Jennifer Zaino is a New York-based freelance writer specializing in business and technology journalism. She has been an executive editor at leading technology publications, including InformationWeek, where she spearheaded an award-winning news section, and Network Computing, where she helped develop online content strategies including review exclusives and analyst reports. Her freelance credentials include being a regular contributor of original content to The Semantic Web Blog; acting as a contributing writer to RFID Journal; and serving as executive editor at the Smart Architect Smart Enterprise Exchange group. Her work also has appeared in publications and on web sites including EdTech (K-12 and Higher Ed), Ingram Micro Channel Advisor, The CMO Site, and Federal Computer Week.

You might also like...

Smart Data Webinar: Advances in Natural Language Processing II – NL Generation

Read More →