5 Ways Data Science Drives Ecommerce Revenue

Analyzing data to improve retail sales is not new. What is new is the volume of data and sophisticated machine-learning algorithms to analyze it.

Data science is the processing and analysis of large datasets — structured or unstructured. Tweets from customers and prospects is an example of unstructured data.

Most of the data-science algorithms have been used for years; many pre-date the first computer. What has led the growth of the field (and all the hype) is that for the first time in history, companies are sitting on massive amounts of data that humans cannot process without advanced computer methods.

Here are five data-science uses to drive ecommerce revenue.

Data Science for Ecommerce

Market basket analysis is not a new concept. Retailers have been doing it for years. The idea is if a customer buys one item, she is likely to buy a related product. For example, a customer who purchases a toothbrush presumably will also need toothpaste.

Enterprise retailers have traditionally acquired expensive reports from research companies such as Nielsen or NPD that contained this insight. Retailers would then know which products to place next to each other in a physical store to increase sales. Now, you can use the purchase history of your online customers to recommend similar products in the checkout process.

Price optimization. Historically retailers have set prices using a few data points, such as profit margin, cost of goods sold, competitors’ pricing, and manufacturer’s suggested retail price. Today, merchants can increase and decrease prices based on many more factors, such as seasonality, demand, customer location, and frequency of purchase. The variables that merchants can use greatly depends on the availability of data.

Promotions. Most marketers gauge the performance of promotions by comparing the results to previous campaigns, A/B testing, and accessing the overall impact on sales. With machine learning, marketers can go further by customizing promotions at the item and customer level. For example, if customer A typically buys once a year on Black Friday, a merchant can send a promotion to that customer on that day. Conversely, a customer who has purchased only when an item is 10 percent off could respond to an instant, on-site, 10-percent off coupon.

Recommendations. Amazon and Netflix, as examples, have sophisticated recommendation algorithms. They suggest products based on each customer’s purchase and search history. Not all merchants can recommend products for each customer in this manner. But they can use a similar process by recommending common upsells and cross-sells from popular items.

Product visualization. Image analysis is fairly new. But companies are increasingly utilizing product visualization to understand what customers find attractive. For example, is a white background better than pink? Does having a close-up photo of a product’s texture make it more salable? Does a human model help sell the product? What about the height of the model? The above questions can be answered by coding each picture on each product. Data science can take it a step further to find the optimal combination of model, texture, photo quantity, light, and other variables — all to make the product more appealing.