An online retailer will start to sell a new product to its customers next month. The price of the new product has been set at $100/unit. The buyer of the retailer has negotiated with its supplier the following ordering costs:
- If the order is less than 120 units: $80/unit
- If the order is between 120 and 280 units: $75/unit
- If the order is more than 280 units: $70/unit
All unsold items can be sold at a different channel at a much lower price of $25/unit. The buyer has to decide the order quantity now. This is a new product and no historical sales data is available. However, the Analytics team at this retailer has been using machine learning models to predict the demand of new products. Their approach suggests that the demand for this new product is going to be 200 units. The buyer analyzed the historical sales data of 500 items that share similar features of the new product. For these 500 items, the Analytics team had applied the same machine learning model to predict their demand. The predicted demand and the actual demand of these 500 items can be found in OrderingNewProduct_Data.xls. The table below shows the data for 10 items. For example, for item 1, the demand predicted by the Analytics team is 199 units and the actual demand is 179 units.
Question: How many units should be ordered for the new product?
Item Forecast Actual Demand
1 199 179
2 159 167
3 186 263
4 164 101
5 161 196
496 149 144
497 187 271
498 66 44
499 172 217
500 228 183