Know the method's capabilities and limitations. Experimentation is especially
well suited to some types of information needs but ill-suited or incapable
of satisfying others.
Evaluate a laboratory experiment's potential. Lab experiments provide
much greater control over the threats to internal validity.
Examine a field experiments promise. Experiments in the field provide
a better parallel with the real word conditions of the population.
Avoid seriously flawed designs. Some designs are so faulty they're likely
to yield information that is worse than none at all.
Consider genuine experiments first. Genuine experimental designs provide
the greatest control over all the threats to internal validity.
Review the need for premeasures. If assignment to treatments is random
and sponsors are comfortable with them, postmeasure-only designs are preferable.
Compromise when necessary. If genuine experimental designs aren't feasible,
search for a viable design even if it's less than perfect.
Examine external validity. Lab experiments and field experiments that
don't parallel the real world may suffer from artificiality and lack generalizability.
Apply appropriate survey research principals. Many of the same principals
and techniques used in survey research apply equally well to experiments.
