STOCHASTIC METHODS AND SYSTEMS FOR DETERMINING DISTRIBUTION CENTER AND WAREHOUSE DEMAND FORECASTS FOR SLOW MOVING PRODUCTS

Abstract

A method and system for determining distribution center or warehouse product order quantities of a slow selling product. The method includes the step of determining for each one of a plurality of stores supplied by the distribution center, a store sales forecast for the slow selling product. The method converts the store sales forecast to a stochastic forecast when the average rate of sale of the product is less than a minimum average rate of sale threshold value. Store order forecasts are thereafter determined by subtracting a store inventory value from the stochastic forecast when average rate of sale is less than the average rate of sale threshold value, and subtracting the store inventory value from the sales forecast when the average rate of sale is not less than said average rate of sale threshold value. The individual store order forecasts are accumulated to generate a distribution center demand forecast; which is compared with current and projected inventory levels for the product at the distribution center to determine distribution center order quantities necessary for maintaining a product inventory level sufficient to meet the distribution center demand forecast for the product.

Description

FIELD OF THE INVENTION

The present invention relates to methods and systems for forecasting product demand for distribution center or warehouse operations; and in particular to an improved method and system for determining distribution center or warehouse order forecasts from store forecasts of slow selling products.

BACKGROUND OF THE INVENTION

Today's competitive business environment demands that retailers be more efficient in managing their inventory levels to reduce costs and yet fulfill demand. To accomplish this, many retailers are developing strong partnerships with their vendors/suppliers to set and deliver common goals. One of the key business objectives both the retailer and vendor are striving to meet is customer satisfaction by having the right merchandise in the right locations at the right time. To that effect it is important that vendor production and deliveries become more efficient. The inability of retailers and suppliers to synchronize the effective distribution of goods through the distribution facilities to the stores has been a major impediment to both maximizing productivity throughout the demand chain and effectively responding to the needs of the consumer.

Teradata Corporation has developed a suite of analytical applications for the retail business, referred to as Teradata Demand Chain Management (DCM), which provides retailers with the tools they need for product demand forecasting, planning and replenishment. Teradata Demand Chain Management assists retailers in accurately forecasting product sales at the store/SKU (Stock Keeping Unit) level to ensure high customer service levels are met, and inventory stock at the store level is optimized and automatically replenished. The individual store product forecasts can thereafter be accumulated and used to determine the appropriate amounts of products to order from a product warehouse or distribution center to meet customer demand. The warehouse must in turn order appropriate amounts from suppliers and vendors based on its demand forecast.

Some currently used methods for forecasting product sales and determining suggested store order quantities (SOQs) suffer when dealing with slow moving products and may produce problematic results when used to determine warehouse or distribution center orders for low inventory, very slow selling products. Problems may include periodic spikes in order forecasts, a drop in the size of an order from week to week, and a large discrepancy between forecasted and actual orders. Described below is an improved methodology for forecasting product sales and determining suggested store order quantities and warehouse demand forecasts for slow selling products.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 provides an illustration of a product supply/demand chain from a supplier and manufacturer to a retail store and customer.

FIG. 2 is process flow diagram illustrating a synchronized DC/warehouse forecasting and replenishment process.

FIG. 3 is a high level block diagram illustration of a process for determining DC/warehouse demand from an accumulation of store suggested order quantity (SOQ) data.

FIG. 4 is a high level block diagram illustration of a process for determining DC/warehouse demand from a roll-up of store long range order forecasts.

FIG. 5A illustrates the total demand forecast and accumulated suggested order quantity forecast for a very low selling product sold at a number of stores over a sixty-five week period.

FIG. 5B illustrates the effective total inventory of the product of FIG. 5A over the same sixty-five week period.

FIG. 6 provides a simple flow diagram of a process for determining product demand forecasts and suggested order quantities for slow selling products in accordance with the resent invention.

FIG. 7 illustrates an accumulated suggested order quantity forecast for a very low selling product sold at a number of stores over a forty-four week period following implementation of the process illustrated in FIG. 6.

FIG. 8 illustrates the effective total inventory of the product of FIG. 7 following implementation of the process illustrated in FIG. 6.

DETAILED DESCRIPTION OF THE INVENTION

In the following description, reference is made to the accompanying drawings that form a part hereof, and in which is shown by way of illustration specific embodiments in which the invention may be practiced. These embodiments are described in sufficient detail to enable one of ordinary skill in the art to practice the invention, and it is to be understood that other embodiments may be utilized and that structural, logical, optical, and electrical changes may be made without departing from the scope of the present invention. The following description is, therefore, not to be taken in a limited sense, and the scope of the present invention is defined by the appended claims.

FIG. 1 provides an illustration of a retail demand/supply chain from a customer 101 to a retail store 103, retail distribution center/warehouse 105, manufacturer distribution center/warehouse 107, manufacturer 109 and supplier 111. Arrows 115 are used to illustrate communication between the demand/supply chain entities. The Teradata Demand Chain Management system 151 includes product demand forecasting, planning and replenishment applications executed on server 153 determines store order quantities 155 and distribution center forecasts 157, and provides for the synchronization of the warehouse/distribution center replenishment system with the replenishment ordering system from their stores.

A synchronized DC/warehouse forecasting and replenishment process is illustrated in the process flow diagram of FIG. 2. Beginning at step 205, each retail store 201 supplied by warehouse 203 creates a store forecast and order forecast utilizing a methodology such as the methods illustrated in FIG. 3 or 4. In step 207, the individual store order forecasts are accumulated to the DC/warehouse level. This rolled-up order forecast is provided to the DC/warehouse 203 for use as the DC/warehouse demand forecast, as shown in step 211.

In step 213, DC/warehouse level policies may be established for RT (Review Time from last time the replenishment system was run), LT (Lead Time from the order being cut to the delivery of product), PSD (Planned Sales Days, the amount of time the Effective Inventory should service the forecast demand), Replenishment Strategy, and Service Level. In step 215, forecast error is calculated comparing actual store suggested order quantities (SOQs) to DC/warehouse forecast orders. Finally, in step 217, weekly forecasts are broken down to determine daily forecasts, calculate safety stock and SOQs. Safety Stock is the statistical risk stock needed to meet a certain service level for a given order quantity. The safety stock is a function of lead times, planned sales days, service level and forecast error.

There are several methods that can be utilized to produce DC/warehouse demand forecasts. Two methods for generating DC/warehouse demand forecasts, illustrated in FIGS. 3 and 4, are described below. FIG. 3 illustrates a process where DC/warehouse demand forecasts are determined from roll up of Suggested Order Quantities (SOQs). Suggested Order Quantity information from numerous store locations 301-304 is aggregated 305 and used to generate DC/warehouse profile and weekly, monthly or quarterly forecasts 307. This method takes into account lead times, seasonality and recent trends in both store and DC/warehouse requirements. The SOQ represents true DC/warehouse demand from stores as it calculates demand for the stocking period (planned sales days), considers lost sales where they exist and subtracts the effective inventory (on hand and on order) in building the correct store orders.

FIG. 4 is a high level illustration of a process wherein store order forecasts determined for numerous retail stores 401-404 are accumulated 405 to create the DC/warehouse Synchronized Demand 407. Store order forecasts are determined through the process described in application Ser. No. 10/737,056, referred to above and incorporated by reference herein. The DC/warehouse replenishment orders will be executed considering all stores' time-phased needs net of effective inventory and applying the DC/warehouse's lead time, planned sales days, forecast error and service levels.

In the processes shown in FIGS. 3 and 4 discussed above, the Suggested Order Quantity (SOQ) or store order forecast for a product is determined by subtracting the effective inventory of the product from the DCM demand forecast for the product. The effective inventory of the product includes the current or beginning inventory of the product, also referred to a beginning on-hand (BOH) stock, plus additional inventory expected to be received by the store prior to the demand forecast period, less expected sales of the product prior to the demand forecast period.

As stated above, some currently used methods for forecasting product sales and determining suggested store order quantities (SOQs) may produce problematic results when used to determine warehouse or distribution center orders for low inventory, very slow selling products. FIGS. 5A and 5B are provided to illustrate this problem. The graphs of FIG. 5A illustrate the total demand forecast and accumulated suggested order quantity forecast for a very low selling product sold at 1100 stores over a sixty-five week period. The graphs of FIG. 5B show the effective total inventory level of the product over that same sixty-five week period. In this example, the most stores have a beginning on-hand inventory of 1 unit, the same weekly average rate of sales (ARS), and decrement on-hand inventory by the same amount every week. Product forecast unit sales and inventory levels are measured against the vertical axis in FIGS. 5A and 5B, respectively. Sales weeks are measured along the horizontal axis in both figures.

Graph 501 of FIG. 5A illustrates the DCM system generated sales forecast for a representative product with a low average rate of sales of 0.024 units/week, i.e., approximately 1 sale every 42 weeks. With a requirement that a minimum stock of 1 unit be maintained at each store, the warehouse or distribution center (DC) suggested order quantities and total store effective inventory levels generated by the DCM system are illustrated by graph 503 of FIG. 5A and graph 513 of FIG. 5B, respectively. Without the requirement that a minimum stock of 1 unit be maintained at each store, the DC suggested order quantities and total store effective inventory levels generated by the DCM system are illustrated by graph 505 of FIG. 5A and graph 515 of FIG. 5B, respectively.

As can be seen in graphs 501, 503, and 513, for the product having an ARS of 0.24, a beginning inventory of 1 at most stores, and a requirement that a minimum stock of 1 unit be maintained at each store, the DCM system will forecast a significant number of product sales near week 42 of the forecast period, followed by a drop in the effective inventory of the product, and a very large DC SOQ at week 46. In this scenario, most of the 1100 stores will order replenishment stock during the same week, week 46, a potentially problematic situation for the warehouse, distribution center, or product manufacturer. A higher or lower ARS for the product will vary the week in which the week in which the spike in SOQ occurs.

Without the requirement that a minimum stock of 1 unit be maintained at each store, graphs 501, 505, and 515, show that the DCM system will forecast a significant number of product sales near week 42 of the forecast period, followed by a drop in the effective inventory of the product, but a replenishment SOQ will not be generated until after the 65 week forecast period. The effective inventory levels are significantly lower without the requirement that a minimum stock of 1 unit be maintained at each store. Following week 46, the effective inventory for the product drops to below 600 units, well below the inventory level needed to meet the potential demand at all locations. This may cause insufficient orders and frequent stock-outs, resulting in lost product sales.

Some of the problems with the currently used methods for determining store and distribution center orders are rooted in the way the way product demand forecasts are used in the order calculations. Currently, a weekly product demand forecast, or Average Rate of Sales (ARS), is a real number, which for a slow selling product is less than one and close to zero: 0≦ARS<1. However, the actual weekly demand in reality is a nonnegative integer, which for a slow selling product is either zero or one: demand={0,1}.

The difference between the nature of actual demands and the way forecasts are defined and used creates a discrepancy between reality and the replenishment model calculations. This discrepancy is particularly substantial when dealing with slow selling products:

• • Orders need to rounded up or down to be whole numbers;
  • The rounding error is significant when dealing with small values; and
  • The errors are accumulated and magnified when orders are rolled up to a distribution center or warehouse level.

A close inspection of demand and forecast values indicates that demand values are probabilistic, or stochastic, by nature, and the outcome of each week demand is either one or zero with probabilities that can be estimated in advance. The forecast values are in fact the estimators of expected or average weekly demand and are not the estimators of each individual outcome.

It is therefore proposed that within the distribution center order forecasting process, the store demand forecasts for slow selling products be converted into stochastic values which are compatible with actual demands. A stochastic process is a probabilistic method for determining the value of a random variable over time.

FIG. 6 provides a simple flow diagram of a process for determining product demand forecasts and suggested order quantities, which utilizes a stochastic process for determining product demand forecasts of slow selling products. Referring to FIG. 6, the DCM forecasting system provides a weekly store demand forecast, a beginning on-hand inventory level, an on-order inventory value, and an average rate of sale value for a product in step 601. In step 603, the average rate of sale value is compared to an average rate of sale (ARS) limit values to determine if the product is to be treated as a very slow selling product. In the example discussed herein, the ARS limit is 0.1 units per week.

If the average rate of sale value exceeds the ARS limit value, the product will not be considered a very low selling product, and in accordance with step 605 the suggested order quantity for the product is determined by subtracting the effective inventory value, i.e., the on-hand and on-order inventory values, of the product from the DCM demand forecast for the product. The DCM forecasting process continues in step 611 with the SOQ determined in step 605 for these products.

When the average rate of sale value for a product falls below the ARS limit value, the product will be considered a very low selling product, and a stochastic process is employed in step 607 to convert the weekly demand forecast into a stochastic forecast. Using a Bernoulli distribution, the stochastic demand forecast is determined as described below:

$f\left(k;p\right)=\{\begin{array}{cc}p& \mathrm{if}k=1,\\ 1-p& \mathrm{if}k=0,\\ 0& \mathrm{otherwise}.\end{array}$

where:

p is the expected value of the distribution, i.e., the average weekly demand;

k is the outcome of the distribution, i.e., the demand of a given week;

0≦p≦1; and

k={0,1}.

In step 609, the suggested order quantity for the product is determined by subtracting the beginning on-hand inventory value and the on-order inventory value from the stochastic demand forecast for the product. The DCM forecasting process continues in step 611 with the SOQ determined in step 609 for the very low selling products. Store SOQs are accumulated to determine the warehouse or distribution center SOQs.

The use of stochastic forecasts within the process of FIG. 6 significantly improves the stability and consistency of order forecasts for slow selling products, and more stable inventory levels at the distribution center level. The use of stochastic forecasts within the process of FIG. 6 also improves the accuracy of order forecasts compared to actual orders, reduces the drop between the first and the second week SOQs, and generates more effective order triggers and rounding. FIGS. 7 and 8 illustrate some of these improvements in order forecasting for slow selling products.

FIG. 7 provides a comparison between order forecasts for a very low selling product determined through prior DCM forecasting methods, graph 701, and through the stochastic process described above, graph 703. The graphs of FIG. 7 show weekly order forecasts calculated at week 32 and rolled-up to the distribution center level, for 1970 slow selling products. These products comprise three products at 703 locations with an ARS between 0 and 0.33 units per week (0≦ARS<0.33). As can be seen in graph 701, the prior DCM forecasting method produces large variations in order quantities, particularly a large order spike at week 42. In contrast, the order forecast provided by the stochastic method is far more stable.

FIG. 8 provides a comparison between on-hand inventory levels for the same product locations shown in FIG. 7. Graph 801 shows on-hand inventory levels resulting from the use of the prior DCM forecasting method, while graph 803 shows inventory levels resulting from the improved forecasting methodology using stochastic demand forecasts for the slow selling products. Again, the inventory levels associated with the stochastic method are far more stable those associated with the prior forecasting method.

CONCLUSION

The improved methodology for forecasting product sales and determining suggested store order quantities and warehouse demand forecasts using stochastic demand forecasts for slow selling products better represents the supply chain reality. Converting forecast values into stochastic forecast values is simple, scalable, easily implemented within the DCM forecasting system, and performed with little computational effort. Using stochastic forecasts can eliminate the need for rounding of order quantities and therefore reduces rounding error in the calculations. Use of stochastic demand forecasts for slow selling products improves the accuracy of order forecasts, reduces the drop between the first and the second week SOQs, and generates more effective order triggers and rounding.

Instructions of the various software routines discussed herein, such as the methods illustrated in FIG. 6, are stored on one or more storage modules in the system shown in FIG. 1 and loaded for execution on corresponding control units or processors. The control units or processors include microprocessors, microcontrollers, processor modules or subsystems, or other control or computing devices. As used here, a “controller” refers to hardware, software, or a combination thereof. A “controller” can refer to a single component or to plural components, whether software or hardware.

Data and instructions of the various software routines are stored in respective storage modules, which are implemented as one or more machine-readable storage media. The storage media include different forms of memory including semiconductor memory devices such as dynamic or static random access memories (DRAMs or SRAMs), erasable and programmable read-only memories (EPROMs), electrically erasable and programmable read-only memories (EEPROMs) and flash memories; magnetic disks such as fixed, floppy and removable disks; other magnetic media including tape; and optical media such as compact disks (CDs) or digital video disks (DVDs).

The instructions of the software routines are loaded or transported to each device or system in one of many different ways. For example, code segments including instructions stored on floppy disks, CD or DVD media, a hard disk, or transported through a network interface card, modem, or other interface device are loaded into the device or system and executed as corresponding software modules or layers.

The foregoing description of various embodiments of the invention has been presented for purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise form disclosed. Many alternatives, modifications, and variations will be apparent to those skilled in the art in light of the above teaching. Accordingly, this invention is intended to embrace all alternatives, modifications, equivalents, and variations that fall within the spirit and broad scope of the attached claims.

Claims

1. A computer-implemented method for determining product order quantities required to meet future product demands for a distribution center, the method comprising the steps of: for each one of a plurality of stores: determining, by said computer, an average rate of sale of said product;comparing, by said computer, said average rate of sale to an average rate of sale threshold value;determining, by said computer, a sales forecast for said product;converting, by said computer, said sales forecast into a stochastic forecast when said average rate of sale is less than said average rate of sale threshold value; anddetermining, by said computer, a store order forecast by subtracting a store inventory value from said stochastic forecast when said average rate of sale is less than said average rate of sale threshold value, and subtracting said store inventory value from said sales forecast when said average rate of sale is not less than said average rate of sale threshold value;accumulating, by said computer, said store order forecasts for said plurality of retail stores to generate a distribution center demand forecast for said distribution center;comparing, by said computer, said distribution center demand forecast with current and projected future inventory levels at said distribution center of said product; anddetermining, by said computer, from distribution center demand forecast and said current and projected future inventory levels distribution center suggested order quantities necessary for maintaining a minimum inventory level sufficient to meet said distribution center demand forecast for said product.

2. The computer-implemented method for determining product order quantities in accordance with claim 1, wherein said stochastic forecast is determined through use of a Bernoulli distribution:

3. A computer-implemented method for determining product order quantities for a store, the method comprising the steps of: determining, by a computer, an average rate of sale of a product;comparing, by said computer, said average rate of sale to an average rate of sale threshold value;converting, by said computer, said sales forecast into a stochastic forecast when said average rate of sale is less than said average rate of sale threshold value; anddetermining, by said computer, a store order forecast by subtracting a store inventory value from said stochastic forecast when said average rate of sale is less than said average rate of sale threshold value, and subtracting said store inventory value from said sales forecast when said average rate of sale is not less than said average rate of sale threshold value.

4. The computer-implemented method for determining product order quantities in accordance with claim 2, wherein said stochastic forecast is determined through use of a Bernoulli distribution:

5. A system for determining product order quantities required to meet future product demands for a distribution center, the system comprising: a computer for:determining, for each one of a plurality of stores, an average rate of sale of a product;comparing, for each one of a plurality of stores, said average rate of sale to an average rate of sale threshold value;determining, for each one of a plurality of stores, a sales forecast for said product;converting, for each one of a plurality of stores, said sales forecast into a stochastic forecast when said average rate of sale is less than said average rate of sale threshold value;determining, for each one of a plurality of stores, a store order forecast by subtracting a store inventory value from said stochastic forecast when said average rate of sale is less than said average rate of sale threshold value, and subtracting said store inventory value from said sales forecast when said average rate of sale is not less than said average rate of sale threshold value;accumulating, said store order forecasts for said plurality of stores to generate a distribution center demand forecast for said distribution center;comparing said distribution center demand forecast with current and projected future inventory levels at said distribution center of said product; anddetermining from distribution center demand forecast and said current and projected future inventory levels distribution center suggested order quantities necessary for maintaining a minimum inventory level sufficient to meet said distribution center demand forecast for said product.

6. The system according to claim 5, wherein: said stochastic forecast is determined through use of a Bernoulli distribution:

7. A system for determining product order quantities for a store, the system comprising: a computer for:determining, for each one of a plurality of stores, an average rate of sale of a product;comparing, for each one of a plurality of stores, said average rate of sale to an average rate of sale threshold value;determining, for each one of a plurality of stores, a sales forecast for said product;converting, for each one of a plurality of stores, said sales forecast into a stochastic forecast when said average rate of sale is less than said average rate of sale threshold value; anddetermining, for each one of a plurality of stores, a store order forecast by subtracting a store inventory value from said stochastic forecast when said average rate of sale is less than said average rate of sale threshold value, and subtracting said store inventory value from said sales forecast when said average rate of sale is not less than said average rate of sale threshold value.

8. The system according to claim 5, wherein: said stochastic forecast is determined through use of a Bernoulli distribution:

9. A computer program, stored on a tangible storage medium, for determining product order quantities required to meet future product demands for a distribution center, the program including executable instructions that cause a computer to: for each one of a plurality of stores: determine an average rate of sale of said product;compare said average rate of sale to an average rate of sale threshold value;determine a sales forecast for said product;convert said sales forecast into a stochastic forecast when said average rate of sale is less than said average rate of sale threshold value; anddetermine a store order forecast by subtracting a store inventory value from said stochastic forecast when said average rate of sale is less than said average rate of sale threshold value, and subtracting said store inventory value from said sales forecast when said average rate of sale is not less than said average rate of sale threshold value;accumulate said store order forecasts for said plurality of retail stores to generate a distribution center demand forecast for said distribution center;compare said distribution center demand forecast with current and projected future inventory levels at said distribution center of said product; anddetermine from distribution center demand forecast and said current and projected future inventory levels distribution center suggested order quantities necessary for maintaining a minimum inventory level sufficient to meet said distribution center demand forecast for said product.

10. The computer program, stored on a tangible storage medium, for determining product order quantities according to claim 9, wherein said stochastic forecast is determined through use of a Bernoulli distribution:

11. A computer program, stored on a tangible storage medium, for determining product order quantities for a store, the program including executable instructions that cause a computer to: determine an average rate of sale of a product;compare said average rate of sale to an average rate of sale threshold value;convert said sales forecast into a stochastic forecast when said average rate of sale is less than said average rate of sale threshold value; anddetermine a store order forecast by subtracting a store inventory value from said stochastic forecast when said average rate of sale is less than said average rate of sale threshold value, and subtracting said store inventory value from said sales forecast when said average rate of sale is not less than said average rate of sale threshold value.

12. The computer program, stored on a tangible storage medium, for determining product order quantities according to claim 11, wherein said stochastic forecast is determined through use of a Bernoulli distribution: