Academic journal article The Journal of Business Forecasting

Forecasting Demand with Point of Sales Data-A Case Study of Fashion Products

Academic journal article The Journal of Business Forecasting

Forecasting Demand with Point of Sales Data-A Case Study of Fashion Products

Article excerpt

The retail industry is faced with increasingly shorter lead times due to changing customer-supplier relationships and overall competitive and profitability pressures. Many retailers utilize weekly POS data to improve forecasting accuracy of their products by store location. In this article we describe methods to improve weekly demand forecasts by using the Point of Sales (POS) data. This data, which represents retail store sales to their final consumers, are captured electronically from retail accounts. In forecasting consumer demand trends, POS data represents the most current indicator of actual consumer demand; in fact, it is the first indicator of changes in consumer demand patterns. In consideration of lead times and the potential short duration of trends, the fashion industry requires a weekly forecasting technique, which detects early changes in consumer demand so that it can quickly respond by revising forecasts, as well as production plans.

The Monet Group, acquired by Liz Claiborne, is the world leader in the design, production, and distribution of costume jewelry. End customers include most large retailers such as Macy's (USA), Breuninger (Germany), Harrods (UK), Galeries Lafayette (France), and De Bijenkorf (Holland), as well as many other smaller retail outlets.

METHODOLOGY

POS data is transferred via EDI (Electronic Data Interchange), which is to say, it is transferred by way of computer-to-computer data transfers. Retail POS data is transmitted from retail stores to our computer facility once each week. The POS data is modeled for seasonality patterns and then an annual (single number) forecast of expected sales to POS accounts is produced which is then broken down into 52 weekly periods. The derived annual forecast of retail POS sales is inflated for non-POS customers, suchas international and military customers. Due to the difference in seasonality between retail POS sales and shipments from the distribution facility, the inflated annual estimate of POS sales is re-seasonalized into monthly buckets based on seasonality patterns derived from historical shipping patterns.

EXPECTATIONS FROM IMPROVED FORECASTS

The expectations from improved forecasts both by vendors and customers are:

* Lower Inventory Cash Flow, which will allow for increased expenditures in areas such as advertising and in-store displays to further promote sales.

* Minimum store-level stock outs (lost sales) with a rapid vendor replenishment system.

* Maximum resource utility under constrained production environment by producing the right product at the right time.

HANDLING THE SEASONAL COMPONENT OF A FORECAST

There are five components of a forecast: baseline, seasonality, promotions, events, and outliers. This article, however, discusses only the seasonality component of the entire forecast process. The procedure discussed here can be easily programmed into an automated system - particularly if it is written in a user-friendly programming language.

The costume jewelry business is a highly seasonal business. Therefore, the first step here should be to aggregate POS data to a highest product level by adding up all product items with similar characteristics. In the costume jewelry industry, products are categorized by pierced earrings, clip earrings, necklaces, bracelets, color pierced earrings, etc. This is called category level categorization (in comparison to the brand level, which would be a higher level, or the subcategory level, which would be lower). After aggregation along the product line, the weekly POS data needs to be aggregated into monthly buckets. Figure 1 gives the monthly seasonal indexes of POS data of two categories - Metal Pierced Earrings and Color Pierced Earrings. …

Author Advanced search

Oops!

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.