How Does L.L. Bean Use Past Demand Data and a Specific Item Forecast to Decide How Many Units of That Item to Stock?

1. How does L.L. Bean use past demand data and a specific item forecast to decide how many units of that item to stock?

First L.L. Bean’s buyer for a demand center works with a team of four to five members, the inventory buyer and product people, to arrive at an item forecast by book. The item level forecast is then uploaded to excel and the preliminary forecast is checked for reality. This forecast is done on a per item per book basis.

The past demand and past forecast data is then checked for forecast errors. The forecast error is calculated by finding the ratio of actual demand to the forecasted demand (A/F) on an item level. The forecast errors are then compiled and a frequency distribution of these errors is developed. The frequency distribution developed by compiling past forecast errors is now treated as the probability distribution for future forecast errors. A percentage of past forecast errors for a “new item”, say 60%, falling between a range of 0.8 and 1.2 is then applied to the future forecasts. For current “new items” with a frozen forecast of 1000, the actual demand is assumed to lie between 800 and 1200 with a 60% probability.

Next, the item’s contribution margin if sold and liquidation cost if unsold are calculated. Here the contribution margin is the underage (U) and the liquidation cost is the overage (O). An optimal probability fraction is calculated by using (U/U+O). Lets say this fraction comes to 0.8, the 0.8 fractile value is then identified from the probability distribution identified from past forecast errors. Lets say the fractile value is 1.2 and the frozen forecast for that item is 1000, then the 0.8 fractile of demand distribution is considered as 1,200 and 1,200 items are ordered.

2. What item costs and revenues are relevant to the decision of how many units of that item to stock?

In order to determine the optimal probability for the forecast, Prob (Demand <= Quantity), the underage and overage need to be calculated. Subtracting the retailer cost of the item from the retail price of the item identifies underage. Subtracting the salvage cost (price the item can be sold at when demand is lesser than quantity) from the retailer cost of the item. Therefore the relevant costs and revenues that play a role in the forecast decision are: the item price L.L Bean charges the customer, the cost L.L Bean pays for the item and the price at which the item can be sold if the ordered quantity is greater than the market demand for the item.

3. What information should Scott Sklar have available to help him arrive at a demand forecast for a particular style of men’s shirt that is a new catalog item?

First Scott Sklar starts by identifying the demand for men’s shirts book by book, and then looks at the demand on a per-item basis. In order to forecast the demand for the particular style of men’s shirt, Scott has to identify how the new shirt will affect demand – whether the shirt’s presence in the catalog will generate incremental demand over the demand generated by other items or whether it will steal demand from the other items. Once this is identified, Scott has to rank the items in the book based on expected dollar sales from each item, adjusting the demand according to the effect of the new item. Once the ranking is in place, dollar values need to be assigned to the item and adjustments have to be made in accordance with the book’s overall forecast. The total shirt sales, keeping incremental or stolen demand has to undergo a reality check for which Scott will need to understand market trends, runways and other factors in the clothing industry. If there is a variance between