DOES Test Designs

DOES has developed innovative new test designs which enable experimenters to collect more accurate data with less testing. These designs have been proven using our vast experience in designing tests, having  designed over 400 major tests.

The following table shows the number of test parts  needed in a DOES test design in order to measure variable effects:


For Variables at Three Levels
Number of Variables Number of Test Parts
2-3 6
4-7 10
8-15 19
16-31 36
Using DOES test designs is a big savings benefit in reducing test parts compared to typical three level orthogonal arrays recommended by Taguchi and others. For example, with four variables at three levels one could expect to run 27 test parts using a Box-Behnken test design. Running 10 test parts using a DOES test design instead of 27 test parts saves 17 test parts, a savings of 63%.

Using three level variables is vastly superior to using two level variables. The use of three levels allows us to measure curvilinear (quadratic) effects. The use of only two levels of each variable requires us to assume linear effects, a dangerous assumption. Many variable effects are not linear.

With DOES test designs we can incorporate selected interactions into the test. This is superior to most of the Taguchi test designs that ignore interactions.

A big advantage of using DOES Test designs is that these allow sequential testing. Using these test designs enables you to measure 1st order (linear) and 2nd order (curvilinear) effects. Then add 3rd order and 4th order effects if needed to make accurate predictions.

DOES test designs have built-in measures of repeatability which improve confidence in the test results. DOES test designs work equally well for mixture (blend) formulation tests. This is a big advantage because commercially available test designs do not handle mixtures.


DOES Inc. Clients  Experience  Staff  References  DOE Knowledge Services Contact

DOES, Inc. 2531 Woodbine Road. Winston-Salem, NC 27104 Phone:(336) 624-2294
(336)-721-2441 email: