Quota setup considerations

While quotas are very powerful and reasonably easy to use, there are certain aspects of them that must be understood by all parties involved in order to assure that the results reflect what is sought with regards to the sample. Any issues or complications should be discussed early in the process to ensure smooth fielding, as misunderstandings can often be costly both in terms of time and money. It is generally always a good idea to go over the quotas prior to launching a survey, and there are a few specific topics which should be addressed:

selectby — If a respondent can qualify for more than one quota bucket, is it well understood how the survey will prioritize quota assignment? For instance, prioritizing options via selectby: weight can become quite nuanced and complex; or, making sure clients understand the difference between counts and random.

Minimum quotas — Survey field managers cannot directly control who takes a survey and what answers they give. In this sense, a "minimum quota" is not technically possible. In some cases, a minimum requirement can be achieved by creating a quota with caps on every option except the one with the minimum. As the other buckets fill up, eventually only respondents who meet the criteria for the minimum will be allowed to complete the survey. However, this is really something that needs to be tackled with slow sampling.

Maxgroups greater than one (i.e., a "multi-punch" quota) — This is easily doable, but can be a dangerous thing to do. First, different respondents will be able to qualify for a different number of groups. Thus, the count will not neatly add up to match the survey's total N. Second, a respondent will only need to qualify for one group to enter the survey, even if some of the groups they qualify for are full. These factors can make managing quotas which allow more than one group tricky.

Late quotas — Quotas can be placed anywhere in the programming, but questionnaire designs usually dictate that they should occur early in the survey (within the first ~10 questions). Late terminations are inconsiderate with regards to the respondents' time and compensation, and thus troublesome for the sample providers.

Nested quotas — Nested quotas, or lack thereof, can often times lead to excessive fielding times or skewed results. Too many nested quotas, or too many layers, will make the quotas difficult to fill.  A lack of any may lead to too much asymmetry in the final data (i.e., one concept ends up being assigned to primarily one gender). In many cases, selectby: balance can greatly simplify the programming requirements.