What is Demand Forecasting?
The two biggest challenges that can cut into your profit when it comes to inventory planning is overstocking while underselling and understocking while overselling. Demand Forecasting models help combat this by evaluating fluctuations in demand and making an educated estimation on the number of future sales. This allows you to make informed decisions when ordering and reordering products so you maintain a profitable balance of stocked inventory.
How does Flywheel’s Demand Forecast model work?
Flywheel’s Demand Forecasting model looks at an array of variables when calculating predictions for future demand. In addition to factoring in current sales trends and historic demand, we also factor in variables such as product marketplace, category, seasonal effects, and product price.
Since the Demand Forecast calculations rely heavily on the availability of historical product data, our model is trained on multiple years of network-level data. This enables you to capitalize on trends at the market, category, and product-level to see more accurate demand forecasts for products that are not only seasoned products but also products that are new-to-market.
Within the Products Page > Inventory Metrics you see the Demand Forecasting metrics in Flywheel:
3 Month Forecast
Weeks of Supply
What is the 3 Month Forecast?
The 3-month forecast shows the projected customer demand for a specific SKU over the next 3 month period. This forecast is an estimate of how many units you will need to have in stock to meet customer demand, allowing you to make informed decisions when planning your inventory purchases.
The demand forecast model’s estimate represents the median 50th percentile of future demand; 50% of the time estimated demand will fall above this value, and 50% of the time it will fall below this value.
Important Note: The 3-month forecast does not factor in current inventory levels when making predictions. We recommend filtering to look at your products that do not have excess as a state when making reordering decisions.
What is Weeks of Supply?
Weeks of supply represents the estimated number of weeks your current inventory will last based on the projected demand for the product. It is calculated by comparing the inventory on hand to the average week's demand while factoring in your current sales velocity.
Look at weeks of supply as your projected sales actuals for each week. For example, if your total fulfillment center (FC) inventory quantity is 16 units and weeks of supply is 4, you’re projected to sell 4 units each week on average.
Weeks of supply is used to derive the ‘inventory status’ or the health of your inventory on hand, such as low stock or excess stock. You can also use this metric to gain insights on when to next replenish your inventory.
Weeks of Supply and Inventory Status:
Flywheel assigns the inventory status value for a SKU based on standard weeks of supply thresholds. The aim is to always keep your SKUs at the healthy threshold to ensure you’re always able to meet customer demand but are not overstocked.
Out of Stock = 0 weeks of supply
Low Stock = less than 4 weeks of supply
Healthy = between 4-13 weeks of supply
Excess = greater than 13 weeks of supply
Tip: Consider the type of product you sell and it’s average sell-through rate when applying this data to your business. Use methods such as the ABC Analysis to prioritize your products and understand which ones need the most attention.
Next: Applying Demand Forecasts to your Business
We’re using all available data to make our best prediction, due to the number of variables, this is our best guess or estimate for future demand. The AI predictive capabilities are influenced by the quality of the underlying data. The more data your products have, the more accurate the forecast will be.
Demand Forecasts could be impacted by changes made to your advertising strategy i.e. significant increases or decreases in ad spend, running promotional offers or lightning deals, and changes in competition.
Footnote: The scientific paper which details how Flywheel’s Demand Forecasting works can be viewed here.