Documentation

Combining multiple quota groups

Most real surveys enforce more than one set of quotas at once: age, gender, ethnicity, income, and region, each balanced to its own target. This page explains how MX8 Labs combines multiple quota groups when it decides whether to accept a respondent, and how that interacts with the fielding strategy. It is the detailed companion to Quota management methodology; read that page first for the single-group concepts (strategies, interlocked vs. marginal, click balancing).

One interlocked group vs. several marginal groups

There are two ways to enforce targets on several dimensions, and they are structurally different.

One interlocked group. A single set_quota() whose lines are the full crossing of the dimensions (age × gender × region), so every respondent lands in exactly one cell. This controls the joint distribution but multiplies cells and makes rare combinations hard to fill.

Several marginal groups. One set_quota() per dimension (a group for age, a group for gender, a group for region), each summing to 100% on its own. A respondent counts toward one line in each group simultaneously. This controls the margins independently, is far easier to fill, and is the more common setup for large studies. The live quota report below shows exactly this pattern.

You can also mix the two: interlock the two or three dimensions whose joint distribution genuinely matters, and leave the rest as separate marginal groups.

How the fielding strategy combines groups

When a respondent reaches the quota checks, MX8 Labs assigns them, within each group, to the least-filled line they qualify for (see click balancing). A group is considered open for that respondent if it still has room in a line they match, and full for them if every line they match in that group has already hit its target. What happens next depends on the strategy:

  • Strict. The respondent must be open in every group. If even one group is full for them, they are terminated, even though other groups still need them. This is why strict fielding with several groups stalls hardest: the more groups you enforce, the more ways a respondent can be blocked.
  • Quota driven and respondent driven. The respondent is accepted if they are open in any group. They still fill the least-filled qualifying line in each open group, but a single full group does not terminate them.

This asymmetry is the crux of multi-group design. Under strict, adding a group can only make a respondent harder to accept; under the driven strategies, groups are enforced more leniently and residual imbalance is corrected by weighting.

A respondent who is terminated at the quota checks is recorded with a quota-screenout status, which is what maps to your sample provider's over-quota exit link. See Setting up any sample provider.

Reading the live quota report

The quota report shows each group as its own table and updates in real time as respondents complete. Counts increment only on completion, so the figures always reflect finished interviews, not respondents still in progress.

The live quota report, showing several marginal quota groups each balanced to its own target

Each group is reported independently, with one row per line and these columns:

  • Percent: the target proportion for that line.
  • Target: the number of completes that proportion implies at the current sample size.
  • Progress: a bar showing how close the line is to its target. Filled lines and still-filling lines are shown in different colors.
  • Current: completes achieved for that line so far.
  • Remaining: completes still needed. A line that has reached or passed its target shows 0; a line still open shows a positive number.

Because the groups are independent, they fill at different rates. In the example above, every group still has one line filling (the 65+ age band, the "Other" ethnicity, Hispanic "Yes", male respondents, the top income band, and the West region), while the easier lines in each group have already reached their targets. It is the hardest cell in each dimension that lags. Under strict fielding, those open lines would keep the survey running (and reject anyone who only fits already-full lines); under a driven strategy, the study would keep accepting anyone who helps an open line and close on its overall rule.

The Dataset toggle switches the report between live and other datasets, and Edit Quotas lets you change targets while the study is in field. See Adjusting quotas during fielding. Note that the report only refreshes when a respondent completes, so a change you make will not appear reflected until the next completion.

Practical guidance

  • Prefer marginal groups unless the joint distribution matters. Several marginal groups fill far more easily than one interlocked grid of the same dimensions, and weighting recovers most of what you lose. Reserve interlocking for the dimensions where the combination genuinely matters.
  • Be cautious with strict + many groups. Every group you add under strict is another way for a respondent to be blocked. If you need many dimensions enforced, a driven strategy plus weighting is usually more realistic.
  • Watch the slowest line, not the average. A group is only as done as its least-filled line. The report's Remaining column tells you which specific lines are holding the study open.

Related