Skip navigation.

Managing with Store Level Data


Henry M. Steele
July 31, 2010

During the summer of 1974 the first IBM supermarket scanning system (3660) in the United States was installed at the Pathmark (Supermarkets General Corp.) store in South Plainfield, N.J. I managed the on-site IBM installation team, working with a similar SGC group. In addition to many preinstallation functions (e.g., front end modifications, check stand design, training of store personnel, measurement of various factors to determine costs and benefits, etc.), we worked with the IBM development lab to validate function in the 3660 hardware and software. Our application development work, both pre and post installation, focused on front end measurements to facilitate labor scheduling and the use of item movement data (e.g., store inventory replenishment, shelf allocation, shrink identification, direct store delivery control). We eagerly anticipated use of item movement data for merchandising analysis, but did not quantify any work in this area.

The first use of point-of-sale item movement data for merchandising analysis of which I am aware was done by Ralphs in Los Angeles beginning in 1975. Like a number of other retailers, Ralphs did an early installation of scanning in four stores to test this new technology in their environment. They used item movement data from these stores as a sample representing their approximately 100 stores. One concept tested was item price sensitivity for private label orange juice versus the national brand competition. They found they could raise the private label price, i.e. narrow the price gap between private label and national brand, with a significant increase in profitability. Pat Collins, president of Ralphs, presented these results at one or more trade association meetings. Fortune magazine ran an article on this use of scanning data, and a scathing letter from a consumer advocate to the editors appeared in the next issue, blasting Ralphs for ripping off the consumer. Retailers became reluctant to discuss their use of scanning data.

Within IBM and to many of our customers, I became an advocate for developing applications based on point-of-sale data. To IBM I gave three justifications: (l) point-of-sale systems were required, (2) a vast amount of storage was required for the store data base, and (3) many processing cycles (MIPS) were required. To our customers I stressed the large potential benefits, and that merchandising analysis benefits could be derived chain wide based on a sample, long before chain wide scanning was installed.

During 1977 and 1978 I functioned under a program called ATT for Application Transfer Team. I worked, for periods from a few days to a few weeks, with many of our customers designing systems. My favorite system I called MARS, for Merchandising Analysis Reporting System. Because of the cost differential between disk and tape storage, we designed those systems using tape storage. My second favorite system I called CPS for Cashier Performance System. This provided various measurements of cashier activity for recognition and rewarding productivity, but the primary benefit was a recognition of likely sweethearting (cashier underringing and nonringing items for friends and relatives) and a major reduction of that cost. CPS required little storage and was simple to design. With AG of Colorado I designed one such system that we made an IUP (Installed User Program) that anyone could buy the code for $3,000. That low price, even at IBM prices, indicates how simple the application code was.

By the end of the 1970’s, disk storage had declined in cost and improved in performance and capacity so much that it was the preferred storage medium, even for the voluminous files of store level item movement data. I obtained funding approval within IBM to develop with an IBM customer a disk-based merchandising analysis reporting system. My objective was to document such a system, so that others could more easily develop and install a MARS type system. I found a willing partner in Lucky Stores of Dublin, California. We agreed that IBM could publicise and document the system, but would release no Lucky Stores financial data. For six months, during the latter part of 1980 and early 1981, a talented industry specialist from the IBM Oakland office and I worked with four Lucky Stores people to design and code a system named LIMES (Lucky Item Movement Evaluation System). The system became operational in 1981.

From 1984 to my retirement in mid 1991, I served as Industry Consultant for my industry. For those of you who knew him, my predecessor was Ed Igler. It was the best job in the IBM Company. My task was to serve as the IBM interface to the trade associations and to stay abreast of what was occurring in the industry to serve as a consultant to IBM management and to customer management. It gave me the opportunity to track the activities of many of the industry leaders. Because of a trust built up over the years, many of them shared with me what they were accomplishing, knowing I would honor their request not to reveal their results. As this was an area that I had been excited about for almost 20 years, it was gratifying to learn of outstanding work in mechandising analysis and the maintenance of store data bases.