Click to learn more about author Georges Bory.

Setting the right price for each product is a critical function for all retailers. And over time, retailers have identified a go-to method of setting up a pricing strategy, which relies on factors such as market price research and adapting prices up and down based on inventory volume and target margins.  Problem is, this approach doesn’t work on occasions where there is either limited –or no –previous data available.

Luckily, there is a work around: data science.

Even with the deluge of data that is available today, it is still relatively common that retailers encounter an instance – such as the debut of a unique product – where no historical data exists. Data science offers a possible solution to this problem. Through the precise, automated analysis of large data sets, it is possible to calculate the best price for a product without relying on the traditional points of comparison. Here is how:

Classify Products and Anticipate Customer Expectations

Data science brings to the table tools capable of analyzing very large datasets much faster than before. This makes all the difference in the precision of the analysis and the prediction. Whereas before you could only identify nebulous trends or misleading correlations, it is now possible to work with volumes of data large enough to uncover marketing insights that are precise, reliable, and immediately actionable for price optimization.

One example consists in leveraging checkout receipts – a treasure trove of data that all retailers possess – to automatically assign each product a score depending on how often it is purchased together with other products. By applying a classification algorithm, we can differentiate between “driver” products (purchased sometimes alone, but often together with other products); “complementary” products (always purchased together with the same few products) and “independent” products (almost always purchased alone). This classification allows you to apply price rules automatically in order, for instance, to make baskets as a whole as attractive as possible or encourage cross-purchases.

Combined with similar analyses performed, for example, on the behavior of your e-commerce clients or on the evolution of stock levels, it is possible to create an accurate, autonomous set of pricing rules that can be relied on whenever competitor prices are not available.

Manage Large Assortments

This data science approach is also particularly useful for vendors who sell hundreds of thousands of different products and do not have the luxury to match each manually with its equivalent at competitors. The goal, therefore, is to have a set of classification algorithms and pricing rules that can be applied automatically but nevertheless remain relevant and precise enough to recommend the right price at the right time.

These situations become more and more common in retail with the development of marketplaces. Marketplaces are spaces where retailers often have only partial information about the products sold and no way to compare the offering with what competitors sell or even to their own sales history. In spite of that, the host of the marketplace has a vested interest in being able to recommend the best price for any item, one that will satisfy both the vendor and the buyer.

Data science now offers predictive models to calculate such recommendation instead of relying solely on the vendor’s intuition. At the heart of the competition between retailers today is a race to be the first to take control of those tools and leverage all their potential.

In the end, retailers don’t need a vast set of historical data to create effective pricing strategies. Instead, all they need to do is invest in the proper data science tools that can help them plug any gaps they fear might exist, and even uncover valuable insights they did not even know were there.