Descriptive, Predictive & Prescriptive: Understand & Optimize Performance
How Analytics Around a Core Consumer Demand Variable Help Businesses Understand & Optimize Performance
Harvard Business Review (HRB) published a useful overview on how advanced analytics “can help companies solve a host of management problems, including those related to marketing, sales, and supply chain operations, which can lead to a sustainable competitive advantage”.
The article focuses on three different categories of analytics and what each approach can help businesses: Descriptive “business intelligence” to understand what happened, Predictive “predictive engines” to forecast what will happen, and Prescriptive “decision automation” to answer what should be done next. As companies increasingly look to incorporate artificial intelligence (AI) and machine learning (ML) capabilities into their everyday operations, the authors lay out how the different analytics approaches best fit in the spectrum ranging from human-driven decisions to automated AI-directed actions.
Planalytics has worked with hundreds of retailers, restaurants, consumer goods, and service companies to provide demand analytics that isolate and quantify the impacts of the weather. In providing Weather-Driven Demand (WDD) analytics to clients over many years, the ability to precisely calculate how the weather – which is unmatched in its significance and volatility as a daily external influencer of consumer purchasing decisions – is applicable in all three (descriptive, predictive, and prescriptive) analytics environments.
For the consumer-centric clients Planalytics partners with, WDD analytics offer both understanding and actionability around a core determinant of constantly changing demand for what can be thousands of a businesses’ products across thousands of locations on a day-to-day basis.
The HRB article describes descriptive analytics as aggregated observations often referred to as business intelligence. “Descriptive analytics is about making sense of the past to inform the future” and these insights tend to be “coarse in nature, and they require the nontrivial step of extrapolating past trends and projecting them into the future.”
Weather-Driven Demand analytics do fit in the descriptive bucket as the WDD models that determine the weather-to-sales relationships start with a company’s historical transactions data. One critical application of these descriptive analytics is the ability to quantify the weather impacts that are embedded in past sales performance and cleansing the history to create a planning baseline with the weather bias removed.
WDDs are also useful descriptive analytics on an ongoing basis. The ability to calculate the weather impacts across the business enables companies to evaluate last week’s (or month’s or quarter’s) sales from a weather-adjusted or weather-neutral perspective, providing a clearer read on performance that can inform reactions and decisions. For example, a retailer that understands that strong weather “headwinds” have slowed sales the last two weeks may resist implementing heavy promotions or price markdowns thinking that the lower demand levels are permanent.
One key advantage WDDs offer over typical descriptive analytics is the fact that the metrics are accessible at a very granular level (e.g., product by day by location) that can also be aggregated.
The article moves on to discuss predictive analytics and the “limited view of the future” this approach can provide. The authors describe predictive analytics as suitable for more frequent, partially automated decisions with application areas such as demand planning and promotion management.
Calling out structural limitations, the article notes that “…predicting individual input variables can be highly complicated: Weather, competition, and supplier performance, for example, require their own prediction models,” and that “There are also limits to the number of input variables that can be modeled and the level of granularity that can be achieved.”
By having Planalytics provide the specialized model for weather impacts, retailers and other businesses can address one of the more complicated parts of the demand picture. These WDD metrics become a key element for managers to consider when using predictive analytics to identify when and where additional staff or product may be needed to meet increased demand or how it may make to adjust digital advertising spend based on how responsive consumers in specific locations are likely to be due to the weather they will be experiencing.
Lastly, the article focuses on the opportunities of machine-driven prescriptive analytics. “Well-designed prescriptive models can deliver greater financial rewards and better business performance” but are often quite challenging and costly to set up.
One must match the tool to the job and AI/ML-driven prescriptive analytics are likely to disappoint in areas like strategic planning where “the initial definition of the question is actually more important than the formation of accurate answers. But when it comes to optimization of prices, inventories, or marketing investments, analytics offers companies substantial opportunities because accurate answers will better serve their customers’ needs.”
Weather-Driven Demand metrics are perfectly suited as an input into a company’s larger prescriptive analytics framework, especially in situations where a high volume of decisions need to be made frequently at the product/day/store level. For example, a grocery chain or mass merchant that needs daily demand adjustments for thousands of products across all its stores can feed WDD calculations into their demand forecasts (AI/ML-based platforms or packaged software solutions) to optimize store-level inventory replenishment. Financial benefits include increased sales from fewer out-of-stocks, lower inventory costs, and the ability to grow profit by 2-6%.
Read HRB’s “Analytics for Marketers” article for more detail and explore Planalytics’ website to learn more about Weather-Driven Demand analytics.