A quota is a target or limit for a group of respondents. A nationally representative study might require age, gender and region to match census proportions. A product test might need 150 current customers and 150 non-customers.
Quotas protect the validity of the data. They also generate the outcome respondents resent most: qualifying for a survey and being told it is full.
Kinds of quota
- Simple: each variable on its own, such as 50% women.
- Nested or interlocked: combinations, such as women aged 18 to 34 in the South.
- Minimums: at least a number, no cap.
- Maximums: a cap, after which the group is closed.
Nested quotas give tighter samples and many more small cells. Small cells fill unpredictably, and unpredictable cells produce over quotas.
What over quota means
A respondent is over quota when they qualify but their cell is already full. They were exactly who the study wanted, just late.
That is why over quota needs its own endlink. Suppliers use terminates to learn about targeting and over quotas to learn which groups to stop sending. Sending over quotas to the terminate link makes a good supplier look like it sends the wrong people. More on that in what each endlink means.
Where quotas are enforced
In the survey. Only the survey knows the respondent's answers, so it enforces demographic and behavioural cells and redirects to the over quota endlink when a cell is full.
At the router. The router knows the project's completes target and each supplier's allocation. On ResearchDart, once accepted completes reach the project target or a supplier's allocation cap, new entries go back to the supplier as over quota before they see a question, recorded with reason quota_full. Reversed completes stop counting, so quota reopens when you reject bad completes after cleaning.
Router quotas are coarse. They also save the most respondent time, because someone turned away at entry has not answered anything.
Where respondent time gets wasted
Several people enter for the last places in a cell. Some finish in time; the rest are turned away after ten minutes. Allowing a small overage on nearly full cells is kinder and usually cheaper than the goodwill it costs.
Suppliers keep sending closed groups. If a supplier cannot see which cells are closed, they keep inviting people who will be rejected. Share cell status and use the over quota endlink consistently so their systems learn.
Quotas stricter than the analysis needs. Interlocking every variable looks rigorous and multiplies cells. Weighting can often do part of the job.
Rare cells left to the end. The last few percent of a study can take as long as the first half. Start targeted recruitment for rare cells early.
Splitting quota across suppliers
Give each supplier an allocation cap that reflects what it can realistically deliver, and leave some headroom unallocated so faster suppliers can cover slower ones. Watch completes by allocation during field and move quota rather than letting one shortfall hold up the study.
For trackers, fix the supplier mix wave to wave. Changing it can move results more than the thing you are tracking did.