(guest post from our partner Ahlers – learn more on Ahlers’ website)
Demand planning goes further than just looking at historical data. Historical data is a good starting point, but a proper plan includes data from external sources, like market intelligence, weather patterns, and geopolitical or economic developments. Combining information from multiple sources increases the accuracy of the planning and lengthens the horizon. Demand planning goes further than looking at general levels but provides valuable insight at the SKU level.
When you need to make strategic decisions on future demand, you need to look further. Demand modeling becomes a necessity. The more accurate you can forecast your demand, the easier it is to make sure your supply chain design is future-proof. Outcomes of your demand models give you actionable data on make-or-buy decisions and can be used to determine the right inventory levels, keeping them as low as possible without disappointing customers.
Combining knowledge and experience with the right tools
Due to the increased digitalization of supply, there is a growing amount of data available for modeling and planning. Great news! But it also creates the need for more advanced models powered by machine learning—powerful digital tools, combined with data from multiple sources and traditional demand planning. Our partner Ahlers says:
“We use machine learning to create demand models with Tangent Works. The Tangent Works engine enables us to predict the right demand drivers and identify the parameters that have the most impact on demand. The tool can find correlations between the many variables that impact demand.”
4 Steps to Come to a Demand Planning
Step 1: Collecting data
Historical data is always a good starting point. The whole organization comes into play here, as input is needed from marketing, sales, operations, purchasing, and logistics. External data is added as well.
Step 2: Creating scenarios
Next up is the creation of scenarios. Together with your subject matter experts, our data analytics experts determine the drivers for demand and create the scenarios that will run.
Step 3: Running the scenarios and evaluating the outcomes
By running different scenarios, we can use machine learning to determine the most important variables that drive your demand patterns. Running worst case, “normal case”, and best-case scenarios gives an overview of possible outcomes.
Step 4: Making strategic and operational decisions based on the outcome
When all scenarios are run and the data is analyzed, we use the outcomes to update operational plans, change inventory levels, or even redesign the whole supply chain.
Cases: demand in an assembly-to-order environment
A large international manufacturer of radiology equipment found their lead times increasing and inventory costs rising. The manufacturer operates with a make-to-order strategy. From the moment a hospital ordered a radiology machine until it was delivered, the total lead time was 90 days. Once the order was in, the manufacturer started ordering parts and assembling the device.
Ahlers was asked to build a demand planning model of this assemble-to-order strategy and run it through different scenarios to see the exact impact on their supply chain. Ahlers investigated different scenarios on inventory levels to see how those would change when sub-assemblies were pre-built and put in stock. Also, how would these sub-assemblies be distributed globally?
First, the historical data for the last 3 years of sales were analyzed. Then a first rough demand model was created. Workshops with the manufacturer’s sales experts were organized to discuss potential forecasting models and these were updated and combined with historical data. Different scenarios were run: historical demand, double demand, and extreme demand. Outcomes were studied and stock levels for all parts and subassemblies were determined.
When the manufacturer operated under the build-to-order model, a lot of WIP stock was needed, as large quantities of partially assembled machines were waiting for other parts during production. With assembly-to-order, the manufacturer is able to ship out machines to customers more than four times faster, with lower WIP stock levels than before. Total savings on total working capital were 11 percent and the lead time went from 90 to 21 days.