Skip to content
Microsoft Industry Blogs

It’s a truly exciting time in retail and technology plays a huge role in that. It’s not just about transactions online, either: customers have changed the way they behave even in brick & mortar stores. They have become proactive, doing a lot of pre-purchase homework. This has been a challenge for some retailers, particularly during these difficult economic times, but clearly some companies have really used this shift to advantage (think Amazon, Sephora, Nordstrom, Best Buy). The most successful companies (online or traditional) now offer customers a very custom level of services, including constructive recommendations, special discounts, and personalized service.

But this level of service takes commitment, and it isn’t possible using some of the traditional methods of retail sales. Years ago, marketing departments would perhaps publish a catalog and advertise in a newspaper, but they had no real way of knowing if or how that contributed to a customer walking through the door—and making an actual purchase. Today, thanks to the volume of information available to shoppers online combined with social media exchanges, there’s a lot of data available that can be used strategically to influence shoppers and provide the kind of service that’s expected. But how to gather and process all of this data? That’s a challenge.

Machine learning makes Big Data actionable

That’s where machine learning (a big part of predictive analytics) comes in. The right tools that include machine learning capabilities can analyze massive amounts of data—and use more interesting sources of data—and that gives a retailer an edge. From transaction volumes to number of mentions on twitter or “likes” on Facebook, these tools help retailers to segment and target customers more specifically and offer more specific guidance on what those customers want.

It’s even possible to gage sentiment these days—social media and other online conversations let a company know how (and when) people are talking about a product or service. You often even get more context than you could using only survey data or historical sales numbers. And it’s all because these conversations are happening in a way that can be tracked and analyzed. Mining this large amount of social media data is an incredible opportunity for retailers to score big with their customers.

Turning Big Data into a big advantage

Here’s an example. When I was working in Australia, I spent some time working on predicting mobile phone sales, particularly for new product launches. In the old days, you’d have to sort of guess the number of sales, using primarily historical sales data; and that’s what you would use to predict inventory and staffing needs.

Today, that scenario looks very different. We’d look at sales history of a successful product first, analyzing the sales and forecasting to create a benchmark. But then we’d also look at the conversations taking place online and through social networks. We could look at the most recent conversations about our upcoming product as well as the older products, combine that information with presales, and so on, and compare all of that against our benchmark product and begin to build a predictive model.

This provides a much better basis to come up with a solid forecast on sales, including inventory needs and service levels required. During the launch, the process continues—we can run the data through the model on day 1, day 2, and so on. We can measure energy and excitement levels across all the locations to balance inventory—how we might need to adjust, even transferring product from one store to another. So, customers are happy and we avoid disappointing them by running out of stock, long lines or delays, and so on. It also helps us have just the right amount of stock by the end of the launch—not too little and not too much, which is key to cost control and profitability.

Ultimately, machine learning helps retailers meet the demands of today’s customers while managing the business itself in a smarter way.