Executive Industry Consultant, SAS Global Retail/CPG Industry Practice
Charles Chase, executive industry consultant for SAS Global Retail/CPG Industry Practice
Shipping ports are overwhelmed. Truck drivers are in short supply. The White House considers the global supply chain crisis a top priority for the foreseeable future. And retailers large and small continue to struggle with having enough goods available to satisfy shopper demand.
This added pressure on an already fragile supply chain means retailers with advanced demand planning and forecasting skills have a competitive advantage. Using those skills to create an agile supply chain saves on inventory carrying costs and waste, issues particularly critical in the grocery industry. With the right set of demand planning skills and processes, a retailer can hone its forecast accuracy to a razor sharpness that also helps manage working capital and drives sales and revenue.
Start by Harnessing and Processing Traditional and “Unstructured” Data
The amount of data the world creates and consumes is staggering, and more than 90% of that data has been created in the last two years. Much of this is due to the rise of connected devices and unstructured data. Unstructured data doesn’t fit neatly into fields and records — data from video, text, sensors and telematics. It’s referred to collectively (and understatedly) as Big Data.
Supply chain networks are alive with data signals streaming from sources like supplier warehouses, internet-connected vehicle sensors and point-of-sale terminals. We have the technology to collect this data in near-real time to create a complete digital picture of the supply chain. Being able to see this digital picture can help you take meaningful action.
Fortunately, computer processing power is keeping pace with modern demand planning needs. That processing power allows you to apply analytics to produce a model of that digital picture so you can interrogate, manipulate, and derive “what-if” scenarios using friendly interfaces that don’t require a data scientist to understand.
Consider consumption-based forecasting and planning, a modern forecasting technique that models the effects of the marketing mix on retail demand then links it back to the effects on shipments in other channels in the supply chain. Using consumption data (point-of-sale and/or syndicated scanner data) can tell you how your promotion of fresh baby spinach affects consumer demand, allowing companies to react to shifting consumer demand patterns to alert suppliers, distributors, wholesalers to take actions in nearly real-time. The model may point to promoting baby kale instead because it’s more readily available from your suppliers.
Approaching the problem this way offers a compelling opportunity to wring out more supply chain efficiency. But this information on its own is meaningless unless the processes, practices and methodologies of demand forecasting evolve to keep pace with the shifting supply chain landscape.
Embedded Supply Chain Analytics Can Help Grocers Become More Agile
The most important innovation in demand forecasting isn’t even technological. It is the integration of forecasting as an embedded analytics function that permeates the marketing and sales and operations planning processes to drive demand generation. Executive sponsorship and leadership play a powerful role in coordinating the collaboration of these cross-functional teams to create regular briefs that value data and analytics over intuition and detail the strategies and tactics that validate forecast assumptions. Because of the current landscape, effective demand planning may have been compressed from a quarterly function to a biweekly function simply to keep pace with changing demand.
Demand planning is an iterative process. The loop must be closed with metrics that measure planning process performance. That isn’t new, but methodologies must be improved to drive value out of forecasting.
Stop Forecast Value Leaks With Modern Methodology
Typically, the planning process begins with a naive forecast — something comparatively simple, for example, the last known value or a year-over-year figure. This is followed by a statistical forecast, which is subject to overrides from analyst and sales functions to reach a consensus forecast. The consensus forecast is measured against the actual results, using a mean absolute percent error (MAPE) to determine forecast accuracy.
But a MAPE can only tell the size of the error, not isolate the reason for the error or determine whether the forecasting process is good enough. Enter forecast value added (FVA) analysis, or lean forecasting. FVA measures the percentage error at each touchpoint in the forecasting process to determine which is adding value (in this case, accuracy and its related economic benefit). If the consensus forecast has a higher MAPE than the naive forecast, value is leaking out somewhere. FVA is a method used to find those leaks. And segmenting brands by their value to the company provides a laser focus on where the critical leaks are — a 10% MAPE on a line of minimal value can be less significant than a 5% MAPE on a more valuable brand.
These are challenging times for supply chain managers, and accurate demand planning and forecasting are more critical than ever to retail success. The right data tools, processes and methodologies can be used to help drive that success.