Big data shows how algorithmic pricing means the first price usually isn’t the best price.
Image: Flickr/Mike Mozart
Started as a bookseller in 1994, Amazon has quickly come to dominate the online retail landscape and has at this point fundamentally changed the way we shop. One of the main reasons for the retail giant's meteoric success is pure convenience: You can now buy everything from dildos to groceries in Amazon's marketplace, and comparing a product's price at two different retailers only requires a mouse click, virtually guaranteeing you'll get the best possible deal. In fact, you don't even have to make the price comparisons yourself—Amazon's sophisticated algorithms will find the best deal and shove it right under your nose for you. All you have to do is click 'Add to Cart.'
Yet according to new research coming out of Northeastern University, the sellers selected by Amazon's algorithms aren't offering you the best deals in the majority of cases.
The study was specifically looking at prices featured in the 'buy box,' which is featured on every product page on Amazon. For a product being sold by multiple retailers, the retailer in the buy box is the kind of like the default seller for that product—to buy from another retailer you have to navigate to a different page where all the different prices for the product are listed side by side. Since nobody these days has the extra 30-seconds to look at a pre-assembled list of price comparisons, being the default retailer in the buy box is very important. In fact, 82 percent of sales on Amazon occur through the buy box.
Given the importance of being featured in the buy box, it makes sense that retailers will do just about anything to make that happen. This has led to the rise of algorithmic, or algo, pricing on Amazon, where retailers use algorithms bought from vendors like Sellery to adjust a product's price in real time in reaction to changes in competitors' prices and the seller's inventory levels. This can result in the price of a product automatically adjusting itself hundreds of times a day.
"If you want to offer products on Amazon and compete with the big boys for the top selling product, you pretty much have to adopt algo yourself," said Christo Wilson, an assistant professor of computer science at Northeastern university and the lead author of the study. "Otherwise you don't have a prayer."
As the Northeastern researchers found out, Amazon is significantly more likely to feature sellers using algo pricing in the buy box, even though 60 percent of retailers using algo pricing are offering prices that are higher than those who are not. Although Amazon's algorithm for determining which seller is featured in the buy box is unknown, the Northeastern researchers found it to be some combination of competitive pricing and positive feedback on the seller's page.
"There's a feedback loop," Wilson told me. "If a company uses algorithms then that gives them an advantage over their competitors because they can move quickly and get more sales. And then they start getting more feedback on those sales and that is a positive signal that they should get the buy box again in the future. I don't think Amazon cares if people use algorithms to sell—they make money either way."
To arrive at these results, the researchers spent four months collecting data on 1641 of the most popular products on Amazon. Every 25 minutes, the researchers would collect data on the top 20 sellers of each product, such as prices and ratings to see how these variables change over time. To determine which vendors were using algorithmic pricing, the researchers tracked the target price of each product (generally the lowest price or the price offered by Amazon). They then did a correlative analysis to determine which vendors were altering their own prices based on changes in the price of the target product.
Overall, they were able to determine that 500 vendors were very likely using algorithmic pricing. Although it's estimated that only 2-10 percent of Amazon vendors are using algorithmic pricing, algo merchants represent a nearly third of the best-selling products on the site.
Algo pricing is affordable for even the smallest vendors on Amazon and according to the Northeastern study are generally more successful and "win" the buy box more often than non-algorithmic sellers.
Since algo selling offers such a competitive advantage to retailers, Wilson and his colleagues envision "an arms race that may terminate with all serious sellers adopting automation." When algo pricing mechanisms are pitted against each other in Amazon's marketplace, the results can get weird. Case in point is when the price of a used textbook was accidentally ratcheted up to $23-million when two pricing algorithms were left to their own devices. Other times the results are more nefarious, like when retailers used algorithms to fix prices on Amazon.
The lesson here is this: as its name suggests, Amazon is a marketplace. It would be unwise to accept the price of the first vendor promising you the best deal in an IRL marketplace without shopping around first. The same applies online.
So if you want the best deal on Amazon, it looks like you're going to have to make that extra effort to click the mouse one more time.