A Case for Data Scrubbing

Often maintenance systems do not reap the benefits that they promise through no fault of their own. How do you expect a method to enhance underlying data? The solution is that you simply can't. Everything you need is to get good data in the machine so that it may be obtained, processed and utilized to provide useful information for your company.

Allow me to illustrate the price of not having good information with an example. A multi-site maker has four locations, three of which are in fairly close proximity to one another. Each website has its own autonomous storeroom with stock components. At each site, there's a part-time catalogue manager responsible for all database activity. Since the plant is unionized and positions frequently change, the catalog manager may be replaced every month or two.

The resulting stock standards represent this: inconsistent maker naming; lost manufacturer part numbers; inconsistent use of symbols/abbreviations; punctuation mistakes; incomplete descriptions and; duplicate items. System word searches are next to impossible and finding a part is a frustrating, challenging, usually unsuccessful encounter.

Care workers at all places had lost faith in shops; each kept a stash of components concealed somewhere for his own use. To plan for a repair job, they would try to locate parts throughout the system, but if not able to find what they desired, they would abandon the hunt and only order the part right; in the event of a crisis, they may call another location to ask the loan of a part. Inventory value throughout the company topped $80 million.

Recognizing that something needed to be achieved, the organization tried to tackle the data cleaning themselves. They established a group of nineteen internal folks comprised of care employees (Electrical, Mechanical, Instrumentation & Pipe Fitters) from all four sites as well as two support individuals and one Inventory Specialist.

After over a year of work, and with only half of the database cleaned, they chose to participate external data cleaning specialists to revitalize the effort. Systematically, the information from each site was cleaned. In combination with maintenance employees from all sites, a typical design for product descriptions with acceptable noun/modifier pairs has been designed; the order of attributes was negotiated to meet all locations; language, symbols, abbreviations and business nomenclature were agreed upon. It took six months to rework the whole database.

Having good data brings with it quantifiable rewards. Duplicates in sites were shown to be in the 10 percent range. Frequent items across sites were identified in the 25% range. Merging the three regional stores to some central warehouse diminished total stocking levels and allowed websites to share common critical spares. It also freed up tens of thousands in money savings.

Item searches successfully demonstrated part information that maintenance employees could count on. As confidence in the fundamental stores grew, additional inventory from private caches was repatriated, further adding to the savings realized. Overall, across the business, stock was reduced by greater than 20%.

The data cleanup effort clearly paid for itself several times over. It also became the impetus for other corporate initiatives. The business went on to boost its item-equipment connections to further improve the maintenance system. Additionally, it merged items along product lines and reduced its supplier base for volume discounts.

Clearly great data yields great results.