Ricky & Stick - Model Something Real
Ricky & Stick main page
|
|
|
Models are great, but they're not the real thing. And they can be very deceptive when misused. It's all too easy to use a model that is grossly oversimplified; even worse is to use a model that takes no account of the real risk conditions that will be present (like the risk of getting stung by a flock of angry bees!). So we're constantly in danger of letting the most optimal scenario be the basis of our models, convincing ourselves that we're modeling the true context that the system will encounter.
This pitfall is sooo prevalent in software development; it's almost too easy to construct happy models that will give you happy results. But happiness isn't what you want, truthfulness is. If the model doesn't truly mimic the conditions the system will face in the field, then none of the simulations you run will tell you much about how the system will actually perform.
Bottom line: It's great if your testing plan includes using models and simulation. But don’t model what you hope to find; model what will really be out there. Two good questions for the contractor might be: "What does that model leave out? And what's the delta between the model and reality?"
For More Information
Send comments or questions to asp-requests@sei.cmu.edu
return to top
Ricky & Stick main page
Acquisition Support main page