OUR PRINCIPAL METHOD
for finding errors in implementing the specification is to take
on the mentality of the most devious and obtuse user we have ever
encountered and intentionally try to break the product by doing
incorrectly everything the instruction manual says to do. Essentially
we are exploring the boundary cases of the specification by using
We also test more formally for:
- bugs found in a previous release
- errors (and/or usability issues) found in an equivalent product
from another vendor
- errors which typically arise from poor programming practices
we've seen at other companies while consulting for them
- errors which might come from potential limitations of the
underlying algorithms used in the product
A list of such previous errors we have collected for desktop
speech products can be found here.
In addition to experienced testers, when appropriate, we also
employ naive testers. Their inexpert understanding of both the
technology and computers in general cause them to generate sequences
of operations which are so outside of the norm as to set earthquakes
into even the most robust code.
If you have speech applications under development and wish to
have an analysis performed along these lines please contact
More on usability...
More on expandability...