Case Studies

(or, you're from Missouri and I have to show you)

My clients are organizations undertaking difficult information technology projects that need extra help to succeed. The following case studies describe some of these organizations and how I was able to help them.

Specialty sales with rapid inventory

Unique Problem: The CFO of an industrial rock label called because their merchandising division's Apple-based accounting system was slow and made inventory-taking cumbersome. They posted transactions at day's end to minimize downtime and could perform physical inventory counts only quarterly. The 56K data line (this was before anybody had VPNs) between home office and warehouse became a communication bottleneck.

Implementation: I adapted an existing Linux accounting program (including general ledger, accounts payable and receivable, inventory, purchase order, and more) to the company's specific needs. We imported and converted data from their existing application so all past transactions were still available for review and payment.

Critical Result: As a character-based application, the new software was usable from dumb terminals, Windows PCs, and even the art director's MacIntosh. Both home office and warehouse used the software with equal speed because the load on the 56K connection was minimal. The time required for a physical inventory count was slashed from three days to one shift, enabling management to count monthly with very little fulfillment downtime.

Mall retail: legacy application, Internet-enabled platform

Unique Problem: A leading mall retailer wanted to preserve its existing point of sale (POS) system through a chain-wide replacement of in-store systems. The problem? The POS, in which they'd invested years of training and customization, ran on DOS, had little network functionality, and was seemingly nearing end of life. The POS vendor's new product line for Windows would need to be customized all over again at high cost, and retraining would cost again.

Furthermore, the retailer saw opportunities to integrate web client technology as well as credit authorization over IP, both of which would be difficult or impossible on DOS.

Implementation: Working with the POS vendor, I piloted a simplified port of the DOS software to Red Hat Linux. In-house demonstrations showed the concept was viable, so we continued development of all features and functionality that were specific to the new operating system and environment.

Critical Result: The retailer saved at least $200,000 simply by virtue of not requiring Windows licensing on the new systems, preserved their investment in the POS software, was able to migrate all pre-existing data, required only trivial modifications to their back-office systems, and eliminated retraining costs. Additionally, the stability and network-friendliness of the Linux platform reduced help desk support by at least 500 person-hours and completely eliminated all spyware and virus issues. I also provided training and support that empowered the POS vendor to perform their own future maintenance on the new Linux system.

Medical imaging patient records preserved!

Unique Problem: A medical imaging practice's one-of-a-kind health record management system was tied to a DOS platform. They were seeing apparently random corruption and data loss in the files that recorded the radiologist's detailed comments on each mammogram. As a result, patients and primary care physicians were receiving incorrect data, or none at all, on their followup letters and an accurate history of a patient's followups over the years was unreliable. Worse yet, the system's original developer was unavailable.

Implementation: After analyzing the disappearing data and seeking patterns, I determined that the most likely cause was overflow of a hard limit in the system's replaceable database driver (RDD) for free-text memo fields. I rolled in a more sophisticated RDD, with a much higher record limit, into the system's source code; then I migrated all existing data from the old format into the new format.

Critical Result: This solution immediately eliminated the data corruption problem. It also made the software significantly faster and radically reduced the time taken by the nightly index maintenance routine. My client's records have been consistently correct from that point forward.

Supply Chain EDI: flexible and reliable

Unique Problem: The logistics division of a major overnight carrier runs EDI and similar processes, for legacy and compatibility reasons, on no fewer than three versions of Unix and three other operating systems. On two continents. Among corporate departments and outside trading partners. All night. All of the files need to get to their destinations in good time, reliably, with followup in case of errors. About half of the processes required additional data massaging at the destination.

Implementation: I was given a straightforward Perl application called Distribution File System (DFS). I consulted extensively with EDI analysts while gradually rewriting DFS into a more robust data management platform. We developed ways to rename files or clean up their contents, convert to and from EBCDIC, and automatically signal the help desk when anything went wrong. The improved DFS restarted itself as needed, supported user-supplied Perl code in the system configuration file, and actually contained its own user documentation.

Critical Result: Analyst requests for new functionality were turned around in minutes instead of days. The help desk reduced its need to page system admins, and gradually gathered enough experience to solve almost all problems on its own. DFS was affectionately renamed DOOFUS, for Distributed Object-Oriented File Update System.

The Epilogue: Before moving along to the next project, I transferred knowledge and inline documentation to salaried admins who took over further maintenance of DOOFUS. They're still using it today!