FF&P Store Report for 2025-02
In “Forecasting for Fun and Profit” I described how I would simulate the process of running the inventory replenishment decisions of a store for the remainder of a year. Every month I would publish a report on the preceding month’s results.
I just published the report summarizing the month of February.
A few quick thoughts:
- To properly reflect fixed order costs in cost-of-goods-sold, I updated the
shopr
package I use for the simulation. The derivation of cost-of-goods-sold now attributes part of an order’s fixed cost to each unit of product sold based on how many units were ordered. - Simulating realistic data is hard. While the real world has a way of keeping only those combinations of purchase prices, delivery times, order and holding costs that fit to the demand for a product, picking these parameters at random for each product while keeping the overall business reasonably profitable is hard. To avoid a simulation that implies an unprofitable business model, I adjusted the upfront simulation of the products’ attributes.
- The previous two bullets imply that the simulation for February would be nothing like the one I had previously reported for January. In fact, I needed to re-simulate January as basis for February. So I also published a new version of the January report.
- The reports now feature additional product-level information (private-brand products, new products, discontinued products) and a revised out-of-stock impact.
- While the shop is not yet unprofitable, February shows an increase in holding costs and out-of-stock periods, and thus first signs that the (still) default forecast and replenishment optimization are not good enough and will lead the shop into an untenable downward spiral.