predictive analytics

Predictive Analytics For Retail: What Lies Ahead?

Pinterest LinkedIn Tumblr

Although I’m not certain that they all were “Robert Wendland originals,” my late father had many witticisms to which I credit him. With impeccable timing, a simple, pithy phrase would be spoken that was not only appropriate for the moment but also stuck with me for a lifetime. One in particular that I reference often suggests that “You can’t drive forward if you spend your time only looking in the rear-view mirror.”

The rate of change occurring in our lives and across virtually every industry is unprecedented. Oddly, the smartphone was first introduced less than a decade-and-a-half ago, and yet its ubiquity makes it unfathomable to think of going through life without one. The same could be said of self-check-in at airports. (Remember the days of paper tickets, travel agents and full-service?) And, in retail, where our company Hamacher Resource Group works across the retail supply chain, everything I remember from my childhood seems like ancient history.

I am personally fascinated by predictive analytics and the ability to combine data elements to create an approximate idea of what the future may hold. Although data modeling has been used quite extensively in certain industries (e.g., weather predictions based on specific historic models and key indicators, insurance estimates and actuarial tables, etc.), the concept of applying similar science to retail holds unrealized potential. Given technology advancements and data-mining tools, potential future predictions — in other words, answering the question “What lies ahead?” — seems far more attainable than ever before. Considering big data already captured within the retail sector, it becomes mind-boggling.

Virtually all industries have pools of information that could be used for predictive purposes. This reminds me of a presentation I sat through nearly three decades ago by a then-IBM executive who suggested that the future will be owned by those who have the keys to data and the intelligence that it generates. And, as recently as this month’s National Retail Federation event, the power of data was repeatedly emphasized.

According to a recent blog post from McKinsey, “winning decisions are increasingly driven by analytics more than instinct, experience, or merchant ‘art’; what succeeded in the past is now a poor predictor of the future, and analytics is helping to inform and unlock new pockets of growth.”

So, what types of data do we have available in the retail class of trade? Here’s a partial list:

• Point-of-sale transaction-level data

• Retail pricing strategies

• Consumer-based loyalty information (shopper insights)

• Physical store size and demographic characteristics

• Department sizes and product assortment (planogram data)

• Store navigation intelligence (traffic flow)

• Social media metrics

• Competitive intelligence

• Seasonality and other external trends (e.g., weather)

If one takes the time to imagine connecting discrete data elements and begins creatively combining them in unique ways, amazing predictions can be formed. Whether informing promotional campaigns, personalizing customer offers, improving employee training, building customer retention or simply forecasting demand and optimizing assortment and stocking levels, using big data and some gut instinct can conjure up interesting insights. Here are a couple on my mind.

Imagine aligning store navigation intelligence, shopper loyalty data and neighborhood demographics to consider how to attract others within the area by enhancing navigation and department placement to match their needs. Once again, continuing to do the same thing time and time again may not be positioning the retailer to best capture additional sales.

As an example, say a grocery store was initially designed to attract stay-at-home moms with two kids. The store was arranged to cater to her needs, but the neighborhood was now comprised of older adults with limited mobility and fixed incomes. How could the data predict their navigation pattern, category preferences and better cater to their overall shopping occasion? The hypothesis I would look to prove is whether shelves should be lower and aisles wider, whether certain categories (e.g., sugary cereals and baby diapers) should be downsized and how to rearrange the checkout to be less confined and more staff-centric. Predictive analytics could be used to model the potential result of such changes and allow the retailer to assess whether such an investment would be justified by the return.

Another scenario fueled by predictive analytics could look at the success ratio of certain product launches within a retail operation. Examining planogram and assortment data alongside point-of-sale transaction details and customer loyalty intelligence, future predictions can be created to fuel decisions about placement, promotion and timing.

Today a delicate balance exists between staid and proven stock keeping units (SKUs) and potential innovators and disrupters. Developing analytic modeling that can better predict performance could greatly improve buying decisions and bolster the performance of the category. This would certainly help in honing the process of new item evaluation and potentially reduce the number of new products shelved that do not perform to the expectations of the retailer.

In essence, only looking in the rear-view mirror and continually employing the same process will merely produce the same results. On the other hand, using learnings from the past and combining new, enriched data elements could generate a true breakthrough that drives new results.