# Forcing Quota Groups

Any survey population can be broken down into segments, from a simple male/female segmentation to a more complicated very profitable/lesser profitable/unprofitable customer segmentation. For larger surveys with broad segments such as male/female, representation is rarely an issue and therefore you rarely need to force quota groups.�

For example, if you conduct a 1,000 respondent survey and the male to female ratio is 60/40, then you would expect to have approximately 600 males and 400 females completing the survey, which is plenty for additional analysis of those two segments. This tends to happen more often with general consumer surveys then with business-to-business or customer surveys, where total populations and sample sizes are sometimes smaller.

Taking a real world scenario, if a company has 1,000 customers, they might wish to complete a 300 respondent survey, resulting in a total margin of error of +/- 4.7 percent with a 95 percent confidence level. More often then not, the company will have customers broken out in some way, such as 20 percent that are very profitable, 50 percent that are lesser profitable and 30 percent that are unprofitable.�

If the company did a random survey of all customers, it would be reasonable to expect to complete approximately 60 surveys with very profitable customers, 150 surveys with the lesser profitable customers�and 90 with unprofitable customers. If this company is like most, however, they are much more interested in their very profitable customers then their unprofitable customers. Yet by doing random surveys, they will likely end up with only 60 completes from this key group to analyze.�

The answer for this particular company is to force quota groups of perhaps 100 completes per customer segment. This would give the company a much better +/- 6.9 percent margin of error for the profitable group with a 95 percent confidence level, as opposed to +/- 10.6 percent for 60 completed surveys.�

Be aware though, that there are additional costs associated with forcing quota groups. To complete 300 random surveys with a total sample population of 1,000, you would have to achieve a 30 percent response rate (300/1,000), which should be easy enough. If you wanted to establish quota groups as suggested, since there are only 200 (20% of 1,000) very profitable customers to potentially survey and you would be looking to complete 100 surveys, you would have to achieve a 50 percent response rate with that key group. That can be done with additional creativity and work during the data collection process, but doing so will likely translate into additional time and cost.