Technology Intervention Case Studies

In the early 1980's Kmart decided to implement a Point of Sale (POS) store system for scanning UPC symbols. The justification for a retailer like Kmart to know what items were selling in which stores by week and the actual prices charged was compelling. In fact the financial results in the fiscal year ended January 31, 1984 were a huge miss of plan. The cause was that the end of year physical inventory revealed that the accounting for markdowns was so flawed that year end results were severely below the company's financial forecasts. As a result the then CEO, Ben Fauber, launched a series of meetings to determine how far the Kmart scanning system was from full availability.

Unfortunately, at that time there were two Kmart technology departments working on two completely different scanning systems; the Store Systems group was working with the NCR Company and the corporate communications and networking organization (CC&N) organization was developing a new proprietary system which was being built on IBM Series 1 technology.

This latter effort had grown out of a very successful Series 1 system built to manage information flow into and out of the stores (called “KIN” for Kmart Information Network.) The new Series 1 POS project required a new proprietary operating system for the Series 1 (which was called “son of KIN.”) Although the KIN systems only needed to handle a single application at a time, the scanning version required the new Series 1 system to manage the activities of up to 25 IBM electronic cash registers (ECR.) In fact, the substantial intelligence of the 3683 was disabled and the entire burden for processing transactions fell to the operating system in the Series 1; this constituted a huge technical burden on the hardware, the new operating system and the new POS application. It was also being built on a single Series 1 controller which meant that if the Series 1 managing the scanning system failed it would shut down POS scanning in the entire store. In early 1984 the CC&N group predicted that they would have a fully functioning system within 90 days but within 60 days thereafter the CC&N announced that the project was going to be delayed at least 6 more months.

Soon thereafter, Joe Thomas, a Vice President, and Kmart's premier store operations executive, received a double promotion to Executive Vice President and Chief Administrative Officer. Among other responsibilities Joe was given responsibility for all of technology. He and Ben decided to allow both projects to go forward but Joe began to favor the NCR option. Joe and Ben also decided to recruit and hire a Vice President of Merchandizing Systems. This recruiting started in October of 1984 and ended on July 15, 1985 when I was hired as the first officer brought in from outside the company in its 86 year history. I became a candidate because I had had the great fortune of moving into a house next to Ben in the late 1970's. In fact between that time and when I finally joined Kmart, Ben and I had had several conversations during which I reviewed the justifications for scanning for general merchandise retailers. 

By the date that I started Kmart had installed 23 NCR scanning systems and one Series 1 scanning system. None were operating reliably and the functionality did not meet Kmart requirements.

Ironically, the large painting over my desk was of a Boeing 707 wing that showed two large engines; a picture to which I referred many times in the context of our stores's needing two redundant POS controllers; and my insistence that the long term strategy for scanning in the stores demanded that approach.

The major aspects of the Kmart technology intervention were:

• An assessment of each of the scanning projects
  • The statement of requirements and whether or not other parts of the business agreed with them
  • The technical architecture
    • Hardware
    • Operating System
    • Application development language
  • The project management
• An assessment of technology leadership

One of the most telling conversations that I had in the first few weeks I was at Kmart was with the senior manager of store systems development, Ron Primrose. My question to him was, “How many more months will it take to complete the initial release of the full scanning system on the Series 1?“ He told me that he would have to get back to me with a good estimate. I knew that at that time he had 42 full time people programming the Series 1 POS scanning application. I immediately concluded that there was at least another 42 developer years that needed to be invested. Given that conclusion and the fact that when the system was fully developed it would still provide only a single controller for POS scanning in the stores; I decided that the Series 1 implementation needed to be totally abandoned. My recommendation to terminate the Series 1 project was made to Ben, Joe and the entire Executive team on the 75thday I was there. The termination of the project meant that Kmart acknowledged that the more than 100 years of development needed to be written off.

This left the NCR option as the only alternative being worked on. At that time Kmart had tens of thousands of NCR mechanical cash registers (the NCR Class 5) which were reaching their end of life. The 1985 situation was further complicated by the fact that the NCR scanning systems were not performing well and NCR had failed to meet any of their commitments for enhanced and reliable functionality.

About this time IBM announced the next generation of PC: the IBM PC AT. It was my opinion that this was a tipping point in the price performance curve and decided to pursue the use of AT's as store scanning controllers. After some investigation I came across a small company in Raleigh, North Carolina that had developed scanning software that ran on redundant AT's. The company had been created by former employees of the IBM retail organization in Raleigh and was run by a young executive, Horace Clemmons. I called Horace, introduced myself and explained my interest. Horace's response was astounding. He said that he was not interested because “Kmart would become too large a customer and my company would be too dependent on you.“ After additional study of alternatives, I determined that Horace's company's software was the only viable option. After substantial negotiations an arrangement was finalized under which EDS would act as systems integrator for the scanning controllers using Horace's company's software.

Two huge benefits accrued from this decision: 1) the total costs of scanning controllers for Kmart stores was reduced by $100 million and, 2) the flexibility of the approach allowed five different registers to be managed by common scanning controllers. The new system was called PRISM (for Point of sale Retail Information Systems Management.) As various registers were added to the system the first digit of the register system was added as a suffix. For example when IBM 3683 registers were added, the designation became PRISM-3.

A huge percentage of the code of each PRISM-x system was shared by all PRISM systems (for example the recording of the actual price charged for the item scanned for later uploading to Kmart's central merchandising systems.) It is estimated that being able to leverage the common PRISM platforms for all common functionality saved Kmart an additional $50 million.

Another interesting IBM announcement occurred in this timeframe. IBM announced a major new POS register: the IBM 4683 which was the replacement for the 3683. We were of course interested in this new register and by this time were confident that we could incorporate it into the PRISM architecture because we had very successfully installed a few PRISM-3 stores. Unfortunately, the IBM folks that managed the introduction of the 4683 had determined they would only allow the 4683 to be installed and piloted if it were connected to the proprietary IBM controllers. This constraint was not acceptable to us because under no conditions were we interested in another controller option. At that time, Kmart had an executive sponsor by the name of Marvin Mann. He and I met a few times and were at an impasse. Then, as he was leaving Kmart headquarters on one of his trips, he and I sat in the lobby and had the conversation that determined nearly $90 million of Kmart's register purchasing for the next few years. As he was leaving I reminded him of our commitment to the PRISM architecture. I also told him that we could proceed the “easy way or the hard way.” I said this was the easy way: if he could quickly secure enough 4683's to test I was prepared to commit to an immediate order for 10,000 4683's (roughly a $30 million commitment) contingent only on their successful integration into PRISM. Marvin went back to IBM and, armed with my commitment, convinced his peers to allow Kmart to get about 25 4683's within a few weeks. Within a few months we had PRISM-4 installed in a store and executed an order for the first 10,000 4683's. For the next few years Kmart purchased about 25% of IBM's worldwide 4683 production.

Decisions made in the first 100 days

It is interesting to note that the fundamental decisions which lead to these extraordinary savings were made within the first 100 days after the technology intervention at Kmart. These decisions were:

• The abandonment of the severely flawed Series 1 POS scanning system development initiative
• The decision to explore the possibility of using twin IBM AT's as store scanning controllers
• The abandonment of the NCR scanning systems.
• The decision to use scanning software on the twin AT controllers that had been developed by Horace's company
• The decision to use EDS as the systems integrator.
• The decision to reach into the store systems organization and assign an extremely talented systems manager, Diane Hanes, to lead the project

The corporate environment in existence that led to the need for an intervention included the following:

• Corporate executive leadership had lost touch with one of the most important projects for the company
• Technology leadership had been allowed to perpetuate and even extend technology that had been obsoleted from both a price-performance and ease of development standpoint.
• The culture of much of the technology organization did not endorse or even tolerate the notion that there may have been solutions that were not developed internally.

As the scanning systems began rolling out, the connectivity between them and other store systems (e.g., layaway, pharmacy, time clocks, KIN, automotive, printers, etc.) became a real concern. In fact, it became clear that every store system needed to exchange information from at least one other. This observation led to the conclusion that we needed a local area network technology so that we did not have to write point to point interfaces. The architecture which included all of the various store systems technologies was diagrammed and circulated as the Strategic Store Environment. A subsequent issue of a retail magazine carried a cover story with the following quote from me: “I do not buy products, I invest in architecture.“