Academic journal article The Journal of Business Forecasting

Demand Planning and Forecasting with POS Data: A Case Study

Academic journal article The Journal of Business Forecasting

Demand Planning and Forecasting with POS Data: A Case Study

Article excerpt

The retailer is the final frontier of supply chain planning. So, it is important for manufacturers to have a serious look at what is happening at the retail outlet level because that is where there is an interface with a real customer. If your customer does not get the products it wants, in the package it wants, and in the stores where they are needed, then your entire supply chain has failed. The most common shortcoming of retail execution is high levels of stock outs. Numerous studies by organizations like the Grocery Manufacturers Association and the National Association of Convenience Stores over the past 10 years have shown that the average manufacturer faces an out of stock rate of 6% to 8%, a number that can reach as high as 12% during price promotions.

The data that is needed for supply chain planning at the retail level is the actual sales to consumers at each retail outlet, which is called Point of Sale (POS) data. Twenty years ago, collecting all the POS data a manufacturer needed was extremely expensive, because it meant getting a massive amount of data from hundreds and thousands of independent businesses. Many of these businesses still had old-fashioned cash registers and did not store their POS data in a database. However, there have been three significant changes over the past 20 years that have dramatically reduced the cost of collecting this data:

1. There has been a massive consolidation of retail stores into large chains. This change has been led by Wal-Mart, which now accounts for nearly 10% of all U.S. domestic retail sales. AU the large chains consolidate the POS data from all their stores in a database at regional locations or at corporate headquarters. So it is now possible to get POS data for thousands of stores from one location. These large chains usually have an IT staff that can work with the manufacturers' IT staff to set up automated data exchange.

2. Data storage has become much less expensive and database engines have become powerful enough to move terabytes of data instead of megabytes of data.

3. Even most small retail outlets, like Mom and Pop Grocery Stores, now use scanners to record sales; with the result, POS data are stored in the database.

Retail chains do not account for 100% of retail sales yet because there are many independent retail outlets. But as the use of POS data by manufacturers grows, it has become profitable for independent IT businesses such as Nielsen and IRI to build data marts with the POS data gathered from independent retailers and then sell to manufacturers at a price.


The basic element of POS data is an individual sales transaction. It contains:

* Universal product code or UPC

* Price

* The number of units per transaction (over 97% of transactions that I have seen involve a single unit; sometimes there is a transaction of multiple units) POS data is converted into revenue by multiplying the quantity of SKUs by their price.

The accuracy of each transaction is generally very high because the UPC is usually scanned from product itself and the prices are rigorously maintained by the retailers because their profit depends on accurate prices.

The most basic supply chain decision that a retailer or manufacturer would make using POS data is the quantity to deliver to a store on a specific day. Therefore, we add the data across all transactions for a day or for each hour. When doing so, it is important to remember that retailers can charge different prices to different customers, so not all transactions will have the same price. I have found that the best daily price to use in forecasting models is the weighted average price of all the day's transactions. This can be computed simply by adding the revenue of all the transactions and dividing it by the sum of units sold.

There are still some channels of retail sales that do not have good POS data because some of the products they sell are not easily classified by a product number such as UPC. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.