Brick-and-mortar stores have taken a tremendous hit in the last ten years, thanks in large part to online megamalls like Amazon.com. But we’ve become so used to the idea that the best deals are found online that we may not even check to see if that’s true. A new study, presented at the 25th Inter­na­tional World Wide Web Conference in April, suggests that Amazon’s featured sellers often charge more than their competitors.

This stealth de-bargaining is the result of two algorithms colliding. First, there’s algorithmic pricing, in which a third-party (3P) seller uses one of several formulas to determine how to price each item. Some sellers may opt to undercut the competition, charging less than the current lowest price, but others may choose to charge more, banking on the fact that you won’t notice or care. 

Next, there’s Amazon’s internal algorithm, the one that determines which seller’s product will be featured in the “Add to Cart” box, also known as the Buy Box. The algorithm does take an item’s price into account, but it also considers a seller’s ranking, positive reviews, and participating in Amazon Prime.

Image Credit: Wilson, Le, and Mislove, 2016.

Researchers at Northeastern University wondered how common algorithmic pricing was, and if it was helping or hurting shoppers. Over the course of two extended site “crawls,” the researchers captured pricing and 3P seller information on 1641 of Amazon’s best-selling products. 

Their results suggest that algorithmic pricing, while not common, gives sellers a significant advantage. Buyers? Not so much. “Amazon has a rel­a­tively low number of algo sellers—from 2 to 10 per­cent,” lead author and computer scientist Christo Wilson said in a press statement. “But they cover almost a third of the best-selling products offered by out­side mer­chants, so the impact is large.”

These high-profile sellers don’t always rise to the top by offering the most stable prices. Wilson and his colleagues found that the prices of 333 of the 1000 products they tracked changed once a day. Prices for 170 of those prod­ucts changed more than three times per day, and 50 changed more than eight times per day. “The prices of products offered by algo sellers were almost 10 times more volatile than prod­ucts with no algo sellers,” Wilson says. 

Nor are featured 3P sellers’ prices necessarily better than those of the competition. “We found that 60 per­cent of sellers using algo­rithmic pricing have prices that are higher than the lowest price for a given product,” Wilson says. “Now, 70 per­cent of the time they only raise the price by $1, but there are many cases where the price increase is on the order of $20 to $60. So you really have to take that extra step and click through to the list of all sellers for a given product if you want to find the lowest price.”