Statistics and Surveys
Statistical principles are very powerful in drawing logical conclusions from
data, and surveys of various groups of people are very useful in providing
that data. But good data collection and good logical inference are more
noted in the breach than the observance. A few basics to help assure good
Be sure your sample is representative. It needs to include the sort of people
at whom you are aiming a product or service. Just having a lot of people
isn't useful if most of them are only responding because they have nothing
better to do. You want the busy ones because they are better prospects. But
don't burn out your sample! Unfortunately, every entity now seems to be
surveying every customer immediately after some transaction (because they
CAN; technology makes it easy and cheap), but this annoys the very people
whose responses would be the most valuable).
Reward your sample. For the same cost you can get more actionable results
from a small sample of heavy hitters by giving them a relatively expensive
and/or clever gift than from a large sample of low/mediocre hitters who are
given a cheap or boring gift.
Elicit specific comments and recommendations. These "open-ended" questions
take a lot more time and skill to interpret, but they yield large nuggets of
value. A questionnaire with no such questions is almost certain to be a
useless one that is only being done because of a bureaucratic requirement.