A multibillion-dollar software manufacturer needed to drive down the total $191 million annual cost of supplying its workers with continuously updated workstations. Both the software manufacturer and its primary vendor/supplier were in the midst of an aggressive cost cutting initiative that would lead to a new compensation model. The initial, first year targeted savings goal was set at $12 million.
The classic Six Sigma DMAIC methodology (Define, Measure, Analyze, Improve, Control) is used to summarize this information technology success story.
The manufacturer and vendor disagreed over the causes for variation in workstation price and delivery times. A large number of vendor employees (80 heads) dedicated to the supply task and machine performance differences were only two of the key issues. Suspected causes (hypotheses) for the significant variations included:
- Hypothesis 1 (H1): Relaxed metrics – meaning there were no statistical methods used – led to casual story telling about the reasons for varying costs.
- Hypothesis 2 (H2): Standardized workstation systems were an unacceptable solution for knowledge workers with idiosyncratic production needs.
- Hypothesis 3 (H3): The current compensation scheme created perverse incentives that led to wasted resources.
- Hypothesis 4 (H4): Vendor costs, which were best in class, were already bare bones. There was little room for improvement due to incredible, prior breakthroughs.
A Master Black Belt ran a one-day Champion training session for managers on both sides of the negotiation table. A series of just-in-time, 1:1, half-day training sessions on data collection, designed experiments, and retrospective data mining were provided to technical support staff people.
Working under the direction of the hands-on Master Black Belt and a management charter for breakthrough improvement, these Green Belts developed a flowchart of the procurement process. (It is worth mentioning that no Japanese martial art metaphors were ever used during this process.) A relational database contained historical data as well as real time, daily transaction data.
Existing variable metric fields included but were not limited to the following: delivery times, cost per component, cost per workstation system, performance, reliability, and warranty costs. These statistical metrics were stratified into homogeneous fields and graphed using quality control charts. In addition, scatter diagrams with their regression analyses were completed to quantify correlations. For the first time, statistically rigorous analysis served as the measurement gold standard.
Using a Six Sigma data mining approach, the project team designed a screening, designed experiment data sampling plan to look at three variables at a time. For example, one two level, three factor design examined delivery times (fast and slow), OEM (Original Equipment Manufacturer number 1 and number 2), and purchasing volumes (low and high). The response measures included the all-important one, dollars.
Statistically or economically significant differences emerged across the board. Numerous patterns in one, two and three dimensions yielded new insights for breakthrough.
The significance of the results and magnitude of the data mining and analysis process cried out for an automated, real-time system that could be used by managers on a daily basis. A highly skilled, relational data base programmer received a three hour, 1:1 training session from the Master Black Belt.
The programmer linked together the relational databases, key metrics between the vendor and the software manufacturer, and Six Sigma analytic software applications. Using macros and Visual Basic programming, a one-click system was implemented that produced real time quality control charts, scatter diagrams, pareto charts and regression analyses. This programming task was completed in three hours time.
Proprietary restrictions prevents an open sharing of the exact annual cost savings achieved. However, it seemed apparent to casual observers that the initial goal of reducing costs by $12 million would easily be exceeded. Total time for this Master Black Belt consulting project was five days.