Quick answer: There is no fixed limit, but every quota layer multiplies the number of cells you must fill. As a rule of thumb, keep hard quotas to two or three dimensions that you will report on (for example age, gender and region), treat everything else as a soft target or weight it afterwards, and check the smallest cell before you commission fieldwork. If a cell is likely to hold only a handful of completes, merge it or relax it.
Quotas do not add up, they multiply. Five age bands, two genders, four regions and three ethnic groups sounds like four simple requirements. In practice it is one hundred and twenty separate cells, and every cell needs its own completes.
Real populations are also uneven. Some cells, such as younger respondents in smaller states or a specific ethnic group in one region, are small in the population and even smaller in an online panel. The more layers you stack, the more your study depends on the scarcest cell, not the average one.
| Quota layers | Example | Cells | Completes per cell (n = 400, even split) |
|---|---|---|---|
| 1 | Age (5 bands) | 5 | 80 |
| 2 | Age x gender | 10 | 40 |
| 3 | Age x gender x region (4) | 40 | 10 |
| 4 | Age x gender x region x ethnicity (3) | 120 | About 3 |
The last row is why projects stall. Even with a generous sample, three completes per cell leaves no room for the cells that fill slowly, and fieldwork drags while the last few cells wait for respondents.
A hard quota is a target you will not close fieldwork without. A soft quota is a guide you monitor but allow to drift, usually correcting it later with weighting. The skill is deciding which is which.
Independent quotas are set one dimension at a time: so many per age band, so many per gender, so many per region. Each is filled on its own, which is easy to achieve but means combinations can drift, for example most of your younger respondents ending up in one gender.
Interlocking quotas set a target for every combination, such as each age band within each gender. They give you tighter control, but they create many more cells. A sensible compromise is to interlock only the two dimensions that matter jointly (often age and gender) and keep the others independent.
Do this on paper before anything is scripted. A short feasibility pass saves weeks of chasing the last few completes.
One caution on arithmetic: multiplying incidence rates assumes the filters are unrelated. In practice behaviours are often correlated, for example people who buy a product are frequently the household decision makers, so naive multiplication can be pessimistic. Treat it as a first estimate and confirm it against real availability. Our guide to checking a niche audience is feasible in Malaysia goes through this in more depth.
Order matters. Start with the changes that cost you the least in reporting accuracy.
A worked example. A brand wants four hundred Malaysian online shoppers with quotas on age, gender, region, ethnicity and income, plus a purchase filter. That is far too many hard layers. A better design keeps age by gender interlocked, keeps region and the purchase filter hard, and treats ethnicity and income as soft targets monitored during fieldwork and weighted if needed. The sample is still credible, and the fieldwork is far more likely to finish on time.
If you are deciding how many completes the whole study needs, our survey sample size guide covers the rules of thumb, and how long market research takes shows where fieldwork fits in the overall timeline.
We build Vase.ai, a research platform with a verified panel of SEA consumers, AI-validated responses and quota controls that let you set interlocking and grouped targets without heavy scripting.
The best starting point is the decision behind the study. Share the business question and who you need to hear from, and we will recommend which quotas to hard-set, which to leave soft, and how to keep the fieldwork on track.
Is it better to have fewer quotas or a bigger sample?
Usually fewer, better chosen quotas. A bigger sample helps, but it does not fix a design with more cells than the audience can support. Choose quotas tied to decisions first, then size the sample to the smallest cell you need to read.
Can I fix an imbalanced sample with weighting instead of quotas?
Often yes for modest imbalances on demographics. Weighting works best when every group is reasonably represented. If a group has only a handful of respondents, weighting makes those few answers count for a lot, so a hard quota is safer.
Should I quota on income in Malaysia?
Only if income drives the decision you are researching, and only if you can define the bands clearly. Income is a sensitive question, response quality varies, and it narrows the audience quickly, so it is often better as a soft target or a reporting cut.