Price Optimization Helps Electronics Retailer Raise Revenue by 16%

By Nikolay Savin

In any difficult situation, cut the price. That’s the basic rule omnichannel consumer electronics retailers use when setting prices. And copy the moves of your competitors. That’s rule No. 2. If a retailer suddenly changes the price for a particular product for a specific reason — e.g., lowering the price to liquidate excess inventory — the whole market mimics the move within two or three hours without even analyzing why.

Mature companies use dynamic pricing, where a rule-based algorithm always maintains the price at the same level or lower than that of competitors A, B or C by x percent. It goes on and on, making pricing wars nonstop, particularly during a low season. Thus, retailers dig a hole of bottom-level prices, and it usually takes a long time (up to three months in some cases) to come out from it and resume generating profit.

Furthermore, operational expenses within the industry are relentlessly growing. Therefore, retailers are looking for a solution to increase operational efficiency, set prices independently from their competitors, and boost revenue.

Foxtrot, a major Eastern European consumer electronics retailer, wanted to craft independent prices and maximize revenue without losing margin.

“We were looking for an algorithm-based solution which could scale and consider all the factors which influence prices and which we usually neglect," said Tatiana Moiseenko, commercial director at Foxtrot. "We realized clearly that our managers wouldn't be able to handle all of these variables, including all our historical data spanning 24 years, and offer optimal prices in real time. We could put together a whole department of data scientists instead, but it would take a long time and cost a lot. So, we preferred a technological solution.”

The artificial intelligence-powered price optimization software Foxtrot chose managed to boost revenue by 16 percent, raise the number of transactions by 13.6 percent, and maintain a 98.5 percent margin level (as compared to 53 percent in the control group where the retailer’s managers set the prices) during a one-month pilot.

The self-learning algorithm factored in all nonlinear interconnections, or pricing and nonpricing parameters, when recommending a price: e.g., the price elasticity, the effect of price changes of product groups and SKUs on other product groups and SKUs, a grace period, seasonality, and customer behavior. Also, it took into account the reserve price, the stock, the average market price, as well as the role and positioning of the SKU, which are usually part of expert-based pricing.

Introducing machine learning algorithms wasn't easy for Foxtrot, though. “The company’s managers were not entirely ready to rely on seemingly counterintuitive pricing recommendations, as the algorithm would suggest setting prices by 2.75 percent higher compared to the rest of the market for some products. We were hearing, ‘I understand everything, but I'm just afraid’ or ‘I’ve been in retail for 20 years. What about you?,’ said Vladimir Kuchkanov, data scientist at Competera.

Besides, Foxtrot noted that “the preparation and processing of the data took slightly longer than expected.” The retailer needed first to collect macroeconomic, historical, sales and Google Analytics data in a single and relevant format.

Foxtrot “is planning to scale up the solution across key product categories, which will help it set optimal prices and increase the predictability of pricing decisions.”

Nikolay Savin is the head of product at Competera, a price optimization software for retailers.