Levinson Productivity Systems, P.C. is a Certified Stochos Partner and, as such, is authorized to promote and to present Licensed Programs in the State of Pennsylvania.

1. Levinson Productivity Systems is professionally obligated to disclose to its clients and prospective clients that it receives a commission on sales leads for Stochos products. Clients are encouraged to select the software package that best meets their needs, and links to other vendors are provided in the Resources directory.
2. The representations on these pages are accurate to the best of this company's knowledge. It is this company's opinion that, in general, Stochos' products are powerful and effective instruments for implementing and managing a first-rate quality management system ("ISO 9000 in a Box"), improving quality and productivity, and managing production. Only Stochos, however, can provide specific advice regarding the suitability of its products for your specific application.

PPI differs from Design of Experiments (DOE) in that DOE is an offline technique. In other words, you can rarely perform DOE on product that you intend to ship to your customers because the experimental conditions might be outside the normal specifications. Differences in factor levels (e.g. in a factorial or fractional factorial design) must often be substantial to identify key design variables, and such differences are often unacceptable in a running process.

There is an important tie-in here to Goldratt's Theory of Constraints (TOC). PPI is an online technique that can be used while the process is running, and without interfering with its normal operation. It makes gradual and incremental changes in the process that, while remaining within specification, drive the process toward an optimum.

Illustration of Simplex EVOP with PPI (212 Kb .GIF animation).

• There are two adjustable process variables (x and y axis)
• The idea is to maximize the response variable, whose values appear on the contour plot. This could, for example, be a process yield.
• Simplex EVOP begins with a triangular experimental design. Unlike the situation in a factorial experiment, the increments between factor levels can be very small. This allows use of this method in an operating process because the settings won't be out of specification.
• A 3-factor system would use a tetrahedral experimental design (four values).
• The next experiment is always a rotation away from the worst-performing point. It really involves running one (not three) additional experiment because we already have results for two of the corner points.
• Rotation continues until an optimum is achieved.