Impossible to see, the future is.
– Yoda, Star Wars

Why Do We Forecast Demand?

Supply chains are similar to living organisms making a multitude of daily decisions. It is an endless stream: how much, what, and where to buy, source, deliver, and store. To make appropriate decisions and weigh the pros and cons of potential outcomes, we need both qualitative and quantitative insights. In supply chains, these decisions ultimately depend on expected revenues and underlying costs (fixed and variable). At the core of these decisions lies the question of how much demand you can expect in the future. The better you can estimate it, the better your decisions.

In short, supply chains are about making decisions. And demand planners are there to provide meaningful, actionable information to support these. Better forecasting will allow your supply chain to face streamlined operations, fewer shortages, less useless inventory, and more sales. Ultimately, fewer costs and more profits. In some cases—for example, when supply is constrained—better forecasting will result in a competitive advantage as you can better prepare for the future. We can picture a demand planner as a sailor on a boat with a spyglass. You want to bring relevant information about what’s on the horizon to your comrades on the ship—leaving them to decide what’s best to do.

This book will take you on a journey towards demand planning excellence. The objective is not to work harder but to work smarter. We will aim for efficiency, efficacy, and focus. As you progress through the book, you will be able to implement new best practices in your demand planning process while challenging the status quo. Data, metrics, process, models, people: no stone will be left unturned. These best practices will allow you to improve the quality of your forecasts to deliver more value to your supply chain and support your colleagues to make better, more informed decisions.

Figure 1 Demand planning excellence: efficacy and efficiency

Throughout the book, I will explain these best practices step by step, highlighting how they will lead you to efficacy (more useful forecasts) and efficiency (reducing your team’s workload while making the most out of their available time). Each chapter will be another opportunity for you to improve your forecasting process and make your forecasts more useful for your supply chain.

The best practices, tips, and tricks I share in this book are anchored in my experience advising supply chains around the globe. Even if these best practices are widely applicable, the particular implementation will change from company to company—mostly depending on your business drivers, your team’s maturity, and the data you have access to. Moreover, I will highlight many widespread bad practices throughout the book, explaining how they are inconsistent with process excellence. Practitioners should quit them.

Five Steps to Demand Planning Excellence

Demand forecasts are used to support supply chain decisions (how much to order, produce, or move). I define demand planning excellence as a combination of efficacy (your forecasts help decision-making) and efficiency (you spend as little time as possible working on these forecasts).

In this book, I will show you how to set up your demand planning process using an original 5-step framework (Figure 2). Forecasts should:

• Be done at the right aggregation level and on the right time horizon
• Leverage appropriate data
• Monitored using relevant metrics
• Rely on appropriate models to generate a forecast baseline
• Be enriched by an efficient review process.

Figure 2 Step Framework for demand planning excellence

As you use this framework—and all its underlying concepts and best practices—you will be able to set up a tailor-made process aiming for demand planning excellence. I like to use this framework to kick off and organize my forecasting projects. So should you.

Before jumping into these best practices, let’s take the time to outline our journey through the five steps and the various questions we will discuss in this book. Note that the chapters follow a path to simplify your learning journey—They do not always strictly stick to the chronology of the demand planning excellence framework.

Objective. What do you need to forecast?

A forecast is a piece of information that various teams in the supply chain will use to make smart decisions. When discussing demand planning improvement with clients, I always start with why. Why do you forecast demand? What decisions are you supporting with this forecast? This should be your starting point too. (Answering these two main questions will also answer a third one: Who will use this forecast?) Knowing the types of decisions your supply chain needs to make (e.g. how much to produce, where to deploy inventory, whether to open or close plants) is the first step of the 5-step framework. Based on these decisions, we will assess the relevant material and temporal aggregation levels as well as the forecasting horizon. We will also discuss how different teams might have to use different forecasting models and processes.

Data. What data do you need to support your forecasting model and process?

The most critical data to collect is unconstrained demand rather than constrained sales. Moreover, you will also need to assess what external drivers impact your demand (such as promotions, pricing, or product launches) and start collecting this data as well.

Metrics. How to evaluate forecasting quality?

First, you will need to select relevant metrics to assess your forecasting quality. We will discuss accuracy and bias. You will learn how to assess if your forecasting model and process are achieving satisfactory accuracy thanks to benchmarks. Furthermore, we will refine our metrics to cope with broad product portfolios.

Baseline Model. How to create an accurate, automated forecast baseline?

You do not want your demand planning team to generate (and review) every single forecast by hand. Instead, to reduce human work to a minimum, you want to use a forecasting model as powerful as possible to generate a forecast baseline. This model should leverage a wide range of insights (such as promotions and other demand drivers). To create this baseline, you can use time series models, predictive models, or machine learning.

Review Process. How to review the baseline forecast, and who should do it?

Once the baseline forecast is generated by your forecasting engine, the enrichment phase can begin: Various teams will review the forecast and suggest modifications. These suggestions should improve the baseline forecast as they bring human expertise and insights to which the model doesn’t have access. The cornerstone of this review process should be the forecast value-added (FVA) framework. It promotes ownership and accountability by tracking each team’s modifications to the baseline forecast and measuring how much they improved (or worsened) it. Using this framework, you will achieve unprecedented levels of efficiency and efficacy. On the other side, if you overlook it, be ready to face influence wars and inefficiencies. Finally, we will also discuss how to reduce your teams’ workload by focusing their work on the most critical products and those for which they are the most likely to add value.

Let’s recap with three examples:

• Short-term forecast. Let’s imagine you need to decide what to ship to your stores every week. The forecast could be updated every week, with a horizon of a few weeks forward. The granularity would be SKU per store. As you need to populate the forecast every week, the time to review it will be limited. Henceforth, only a few demand planners should validate it and focus on only the products where they are the most likely to add value. Machine learning models should typically be preferred here as they can leverage different granular insights (promotions, prices, shortages, weather) and allow demand planning staff to prioritize other forecasts that require more qualitative insights.
• Mid-term forecast. You want to assess what to produce in the coming months. This is your typical S&OP forecast where you need to gather inputs from many stakeholders (sales, finance, marketing, planners, clients, suppliers). The forecast can be generated (and its accuracy measured) at a global level per SKU and once per month using value-weighted metrics. You will have to track forecast value added to ensure everyone contributes to better forecasts while avoiding inherent functional bias.
• Long-term forecast. You need to set the budget for the upcoming year. This is a long-term, aggregated forecast (most likely done at a value/revenue level per brand/segmented). To create various scenarios (based on pricing, marketing, or new product introduction), you will want to use a causal model where the weight of inputs can be set and discussed. You will also have to follow best practices to leverage your team’s insights and avoid intentional and cognitive biases.

Who this book is for

This book was written for anyone who wants to improve their demand planning process. In particular, this book will help the following roles: demand planners, S&OP managers, supply chain leaders, and data scientists working on supply chain projects.

As a demand planner, you have many insights about your industry, products, and clients. You know your business. But you might face an inefficient demand planning process. Repetitive tasks—like manually filling up excel files every month—slow you down and keep you away from more value-adding tasks. Discussions, negotiations, and political alignments between teams (such as sales and finance) might erode your overall forecasting accuracy as it diverts you from focusing on what drives business value.

What will you learn in this book?

This book is packed with knowledge of the best practices in demand forecasting; readers can expect to learn the following:

• How to leverage tools and analytics to focus your work where you will have the most impact.
• How to use a forecasting model to create an accurate forecast baseline.
• How to manage stakeholders (sales, marketing, production, finance) and leverage their inputs.

As an S&OP manager or supply chain leader, you manage a team of professionals working on the demand planning process. You want to be sure that your demand forecast helps the other departments (sales, purchasing, manufacturing, logistics) to make the right decisions. You need tools to assess whether the overall forecasting process is efficient and effective. You want your teams to focus on the most critical products. Moreover, you need insightful metrics to track your process quality (and forecast accuracy). In the end, you need to ensure that your forecasting process is done in the most efficient way and adds value to the supply chain. To do this, you will learn the following:

• The appropriate forecasting granularity and horizon to use when forecasting demand.
• How to select appropriate forecasting metrics to track the quality of your demand planning process.
• How to use benchmarks to assess the efficacy and efficiency of your demand planning process.
• How to segment your products to focus the work of your team where they will add the most value.
• When multiple teams review a forecast, how to promote ownership and accountability using the Forecast Value Added framework.

As a data scientist working on forecasting models, you need a dataset, a clear business objective (metrics, granularity, horizon), and a set of metrics to optimize. Unfortunately, data scientists often kickstart projects by jumping into creating models rather than spending time understanding the business requirements. This is what this book is about. Data scientists can expect to learn the following:

• To identify the business requirements when forecasting demand.
• Which data to feed to your model.
• Which metric(s) to use when assessing the quality of your model?
• Which demand drivers you could use in your model.

Summary

• There are five steps to demand planning excellence: Identify your objective; decide what data you need; understand how to evaluate the quality of your forecasting; create an accurate baseline model; understand how to review the baseline forecast.
• Objective: Always start by asking, why you need to forecast demand; what decisions are you supporting with this forecast?
• Data: Collect data on unconstrained demand, as well as data on external drivers that impact your demand.
• Metrics: Accuracy and bias must be considered.
• Baseline Model: Use a forecasting model as powerful as possible to generate a forecast baseline and reduce the need for manual human efforts
• Review process: Understand where to prioritize efforts, and what sort of forecast will provide the most value; short-term, mid-term, or long-term forecasting?

Learn more about the book here.