REPEATABILITY INDEX TO ENHANCE SEASONAL PRODUCT FORECASTING

• US 20100138273A1
• Filed: 12/01/2008
• Published: 06/03/2010
• Est. Priority Date: 12/01/2008
• Status: Abandoned Application

• Current Assignee: Teradata Corporation
• Original Assignee: Teradata Corporation
• Inventors: Bateni, Arash, Kim, Edward, Chan, David
• Application Number:
  US12/325,414
• Publication Number:
  US 20100138273A1
• Time in Patent Office: Days
• US Class Current: 705/10
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities

Abstract

A repeatability score is described for determining the quality and reliability of product sales data for generating seasonal demand forecasts. The repeatability scores are calculated from seasonal sales data stored in a data warehouse. Products are sorted based on their reliability scores such that those products that are highly seasonal and have a reliable year-to-year demand pattern are used to form initial or unique demand models. Products that are determined to be less reliable based on their repeatability score are added to the unique demand models through an iterative matching process or left out of the unique demand models.

Claims

1. A machine implemented method comprising:

determining a plurality of repeatability scores based on sales data, each of the repeatability scores associated with one of a plurality of products;

selecting at least one of the products based at least in part on the repeatability score associated with the selected product; and

generating a model of future demand for the selected at least one product.

2. The method of claim 1, wherein the repeatability scores comprise a plurality of quality metrics.

3. The method of claim 2, wherein determining each of the quality metrics comprises:

calculating a seasonal demand for each of a plurality of periods based on the sales data;

calculating a standard deviation for the seasonal demand based on the seasonal demand for each of the plurality of periods;

determining an average residual based on the seasonal demand for each of the plurality of products and the sales data; and

dividing the average residual by the standard deviation of the seasonal demand.

4. The method of claim 3, wherein determining each of the quality metrics further comprises:

obtaining the sales data from a database, the sales data comprising a plurality weekly sales figures over at least two years;

determining an overlap percentage based on the presence of the weekly sales figures for each of a plurality of weeks in at least two of the at least two years; and

comparing the overlap percentage to a preselected overlap threshold.

5. The method of claim 4, wherein determining each of the quality metrics further comprises setting a default condition for the associated product when the overlap percentage is less than the preselected overlap threshold, the default condition comprising one of:

assigning the associated product to a master model;

assigning the associated product to a unique model;

or assigning the associated product to be used in a clustering process.

6. The method of claim 1, wherein selecting at least one of the products comprises:

comparing the plurality of repeatability scores to a preselected value;

sorting the products into two or more categories based on whether the repeatability score associated with each of the products is greater or less than the preselected value; and

selecting at least one of the products from a first group corresponding to one of the two or more categories.

7. The method of claim 1, wherein selecting at least one of the products comprises:

comparing the plurality of repeatability scores to a first preselected value;

comparing the plurality of repeatability scores to a second preselected value;

sorting the associated products into three categories based on the comparisons to the first and second preselected values; and

selecting at least one of the products from a first group corresponding to one of the three categories.

8. The method of claim 7, further comprising:

selecting an additional product from a second group corresponding to one of the three categories;

matching the additional product from the second group with one of the at least one selected products from the first group; and

generating a second model of future demand for a cluster, the cluster comprising the matched product from the first group and the additional product from the second group.

9. The method of claim 7, further comprising:

selecting a plurality of additional products from a second group corresponding to one of the three categories;

matching the plurality of additional products from the second group with one of the at least one selected products from the first group; and

generating a second model of future demand for a cluster, the cluster comprising the matched product from the first group and the plurality of additional products from the second group.

10. The method of claim 7, further comprising generating a master model, the master model comprising a second model of future demand for a plurality of additional products form the third group corresponding to one of the three categories, wherein the third group corresponds to a subset of the plurality of products that have non-seasonal demand patterns based on the repeatability scores associated with the subset of the plurality of products.

11. The method of claim 7, wherein the first preselected parameter is between approximately 0.5 and 1.0.

12. The method of claim 7, wherein the second preselected parameter is between approximately 0.7 and 1.5.

13. A machine implemented method for generating a quality metric comprising:

calculating a seasonal demand for a product using stored demand data;

calculating a residual for the product based on the seasonal demand and the stored demand data; and

generating a quality metric by comparing the residual to a variation of the seasonal demand.

14. The machine implemented method of claim 13, wherein calculating the seasonal demand for the product comprises determining a plurality of average weekly sales volumes based on the stored demand data.

15. The machine implemented method of claim 14, wherein calculating the residual comprises comparing the plurality of average weekly sales volumes to a plurality of corresponding weekly sales values.

16. The machine implemented method of claim 13, wherein generating a quality metric comprises dividing the residual by a standard deviation of the seasonal demand.

17. The machine implemented method of claim 13, further comprising:

determining whether the quality metric is within a preselected range corresponding to a repeatable product; and

generating a demand forecast for the product when it is determined that the quality metric is within the preselected range.

18. A system comprising:

a database comprising a plurality of entries;

each of the entries corresponding to a product-location and comprising sales data;

a repeatability score module configured to access the database and determine a repeatability score for each of the plurality of entries; and

a demand model generator configured to access the database and receive the repeatability score for each of the plurality of entries, the demand model generator further configured to generate seasonal demand forecasts for a subset of the plurality of entries using the corresponding sales data, the demand model generator further configured to select the subset based at least in part on the repeatability score for each of the plurality of entries.

19. The system of claim 18, wherein the repeatability score comprises a quality metric.

20. The system of claim 18, wherein the demand model generator is configured to select the subset by comparing the repeatability score for each of the plurality of entries to a preselected parameter.