Search
Please complete your details.
Sep 27, 2018 @ 01:09 by Alexander Voss

How Prescriptive Analytics can help modern retailers

Prescriptive Analytics is likely to mature in the near future and yield high potential for any organization, especially for retailers.
This article simplifies the term prescriptive analytics and clarifies, how it differentiates from other, more established types of data analytics while illustrating the potential behind it.

The potential of prescriptive analytics for modern retail

Prescriptive analytics is right at the “Peak of Inflated Expectation” in the Gartner’s Hype Cycle for Data Science and Machine Learning 2018, alongside deep learning and machine learning. Gartner’s Hype Cycle assesses a technology’s maturity, in which the “Peak of Inflated Expectation” describes a stage that is characterized by over-enthusiasm resulting more often in project failures than successes (Gartner 2018).  Even though this doesn’t sound overly promising, we believe that prescriptive analytics is something to keep an eye on as it is likely to mature in the near future and prospectively yield high potential for any organization, especially for retailers.

Before we explain this hypothesis in detail, it makes sense to clarify what prescriptive analytics exactly is and how it differentiates from other, more established types of data analytics: According to Forrester, prescriptive analytics is “any combination of analytics, math, experiments, simulation, and/or artificial intelligence used to improve the effectiveness of decisions made by humans or by decision logic embedded in applications” (Forrester 2017).
In short: Prescriptive analytics can be leveraged to assist or automate decision making and, above all, it has the potential of improving the quality of decisions, opposed to the ones of humans. In this regard, prescriptive analytics classifies as the fourth dimension in Gartner’s analytics capability framework illustrated in figure 1, differing fundamentally from established analytics capabilities, such as descriptive, diagnostic or predictive analytics, in terms of the human involvement in the decision-making process.


Figure 1: Analytics capability framework (Gartner, 2016).

As depicted in figure 1, the output of prescriptive analytics could, on the one hand, be a set of possible next best actions for what should be done to achieve a desired outcome, such as profit maximization for a current scenario (decision support). This would leave the final decision about what to do next to the human being. On the other hand, however, it could also be used to automatically trigger a process based on the next best action (decision automation), such as automatically replenishing stocks or e-mailing a lapsing customer.

Prescriptive analytics assists a user to refocus on executing instead of analyzing

The strength of prescriptive analytics comes into play either when the complexity of a given scenario exceeds human comprehension, or when a myriad of tasks and hundreds of micro decisions are to be made day in, day out – with respective actions to be executed. A very tangible example is the following scenario:

Based on the current sales forecast and stock levels, a retailer is in danger of running out of stock for a certain product faster than previously anticipated, which would result in the inability to meet existing customer demand. The question here is: What is the best course of action to ideally act on the current situation?

What is the best course of action to ideally act on the current situation?

  • Increase prices for the remaining items in stock to restrain demand? But what is the optimal price based on the customers price elasticity and the respective product prices of the direct competitors?
  • Reduce or reallocate promotions and campaigns to restrain demand? But how should media budgets be reallocated ideally?
  • Change to a faster but more expensive shipping method to replenish before running out of stock? But what is the maximum shipping price to replenish profitably?

Given that there is not only one product listed in the assortment but probably several thousands of products, one would not know where to start taking action. And this is where prescriptive analytics steps in: The technology shines in complex, intertwined scenarios like the above where it outperforms human cognition in terms of speed and accuracy. Prescriptive analytics promises regaining focus for the user towards execution rather than analysis and consequently achieve operational excellence to ideally serve customers.

Figure 2: The promise of prescriptive analytics (dgroup).

How to get started

How should retailers leverage prescriptive analytics? It depends on the retailer's current analytics maturity. This can be determined by answering three key questions:

  • Is a clear data analytics strategy (backed by top-management) defined and implemented, which exceeds reporting and descriptive purposes?
  • Is a performant IT infrastructure in place that allows collecting and processing of vast amounts of data? 
  • Are suitable human resources available, capable of building and deploying prescriptive models (especially data architects, data engineers, data analysts and data scientists)?

If all the answers to the questions above are yes, then prescriptive analytics is something which is worthy of further exploration. In recent years, a growing number of vendors have emerged offering prescriptive analytics software that addresses a wide variety of use cases and can often be tailored to a company's specific needs. Unfortunately, most retailers haven't yet gotten to that point of maturity. Typical constraints are siloed data and the absence of data analytics talent (McKinsey 2016). The good news in that case is, that recently a few, specialized prescriptive analytics vendors entered the playing field, providing "plug & play" prescriptive analytics solutions with a clear focus on business users. The biggest advantage of these vendors is their fast setup time. While this sounds great, one has to keep in mind that retailers who implement these solutions are generally limited to the use cases the vendor offers. In summary, prescriptive analytics is very likely to significantly change the way a retailer is taking decisions over time, especially improving in terms of speed and quality. The technology can be expected to mature quickly, and retailers should not wait too long to deal with this topic but lay the internal basis to be ready to apply prescriptive analytics solutions in due time.

Martin Boes is a Manager at dgroup. He recently designed the overall data analytics strategy for a leading German retailer which included an extensive screening of the prescriptive analytics vendor landscape.

Lennart Scheufler is a digital transformation analyst consulting companies on retail analytics digital innovation as well as organizational change.


About Work Services News & Views Join Contact
©  2018 dgroup