How Many People Must You Survey

If you think you've got a good idea, run it by half a dozen colleagues and you will probably get a quick idea of whether it's likely to be a clear winner or a flop. But very often the responses from a close circle of friends can polarize both ways. Statisticians think that if you put the question to two dozen of the right people you will get a reasonably indicative, initial reaction, which is fine if you are testing a drug as one body reacts much like another. However, if you are dealing with minds, reactions are highly individual. So the only way to be sure you are asking the right people is to ask a larger sample. What this larger sample gives you is never the right answer but an answer that is less prone to error.

The chart depicted in Figure 2-1 shows how many people you'd have to ask to get better than 90 percent reliable answers for various sizes of customer universe. In each case the sample allows for 5 percent of respondents giving wrong answers.

Market Size

5,000

50,000

500,000

5,000,000

Reliability

Number of Respondents Required

90%

258

271

272

272

95%

357

381

384

384

99%

586

655

663

663

Figure 2-1: Predictor of required respondents.

Figure 2-1: Predictor of required respondents.

Note how quickly the number of people required for a quorum levels as market numbers grow. When highly specialized software is being written for a limited number of potential users, you should aim to consult them individually. Because their appreciation of what you are doing is likely to be high, you can usually ask quite complex questions. They may also volunteer a lot of further useful information. But be very circumspect: while most are undoubtedly potential clients, a few may turn out to be competitors.

Was this article helpful?

0 0

Post a comment