Vanguard has been working with customers large and small for 25 years, and our team has run into plenty of misconceptions about forecasting; from misplaced “treatments” to too much or too little historical data. Our mission is to provide the best forecasting and planning systems to our customers, and sometimes myths need to be busted.
Myth #1: Use All the Data
We are often met with teams eager to share all of their customer, product, and sales data dating as far back as their transactional systems reach. While we much prefer to start with too much than too little, we need to prioritize the data we use in a modeling system. Typically, there is a subset of more recent and relevant data that serves as a sufficient and capable model of current business and the current market conditions. We all know that business is constantly changing, so building a business model that reflects the current landscape is a better practice than loading all of it in.
A Practical Guide to Supply Chain Planning Maturity
Myth #2: Complex and Cool Always Rule
Think of some of the highly complex, and even cool, demand sensing techniques you may have heard about. The fundamental assumption is that by using highly complex models, you may have the opportunity to shape demand (or other outcomes) in the near term. While we would love to show off some of the more complex models in Vanguard Predictive Planning, we find that unless you have an operating model to support this type of reaction power (perhaps a vertically integrated manufacturer who owns their retail stores), your ability to sense or shape end-consumer demand is limited.
In these (and most) cases, demand is sell-in, not sell-through. You can shape sell-through via pull strategies, but the effects of a pull campaign are generally well beyond the demand sensing horizon. Very often we live in a world where there is not enough data to support the more complex analysis, and simple outperforms complex. A key, though, is to have a modeling system that can handle the full spectrum, allowing planners to start with simple models and build in more complexity as needed.
Myth #3: One for All, One Forever
Another misconception is that there is a singular perfect forecasting model that will not only work for all products, but once it is working will serve a product (or product set) indefinitely. The truth is, models tend to age over time, since the environment is constantly changing.
Furthermore, if you find a model that works phenomenally for one product, you still need to evaluate the models for other products, and continue re-evaluating all of your models over time. The importance of finding a forecasting system that can automate the evalutation and re-evaluation process is paramount here; otherwise, your planners will have to constantly revisit and revise models across potentially thousands of SKUs on a consistent basis.
Solutions like Vanguard Predictive Planning for Demand Planning will self-evaluate models at the SKU level every time new data comes in. Using Artificial Intelligence, Vanguard automatically assigns the forecasting model that will produce the least error in the near-term. To scale planning activities, a system that dynamically evaluates forecast methods is absolutely necessary.
Impact of Advanced Analytics Forecasting