When an interview is stopped or times out, any cells that were pended for the interview are rolled back (decremented by 1) so that it is as if the interview had never taken place. This ensures that unfinished interviews do not contribute towards the targets when the interviewing program is checking quotas for new interviews.

Data for timed out interviews is held for a certain amount of time in the value cache, and can be used for restarting an interview at or very near the point at which it timed out. If the interview is restarted by clicking Next, the interviewing program reloads the interview from the value cache and repends any quotas that were previously pended, even if this means exceeding the quota targets for those cells.

If the interview is restarted by the interviewer or respondent clicking Previous, the interview is restarted as described above and then the previous page of the interview is displayed. If the answer to the quota-controlled question is changed, the quota tests are carried out using the new answers. In this way, an interview that may have resulted in an over-quota cell count may actually result in a completely different quota cell being pended by the time the interview ends. For more information about snapbacks and quota control, see Quota control and snapbacks.

Interviews that are stopped by clicking the Stop button or by a statement in the script can only be restarted if the project uses sample management, since this allows the interviewing program to locate the interview data using the ID of the respondent’s sample record.

When interviews are restarted via sample management, the interviewing program silently replays the stopped portion of the interview using the data in the case database. If it reaches an instruction to pend quota cells, the interviewing program passes the instruction to the quota object model and the cells are pended as if this were a new interview. What happens next depends on the amount of time that has elapsed between the original interview and the restart, how busy the project is, and the data type of the quota-controlled question.

If the quota is based on a single-choice response list and the target for the respondent’s answer has still not been met, the quota cell is repended and the replay continues. If the target has been met, the interview is terminated and any data associated with subsequent questions is discarded.

If the quota is based on a multiple-choice response list and none of the targets for the respondent’s answer have been met, the quota cells are repended and the replay continues using the original data. If some cells are still open and others are now closed, the cells for the answers with open quota cells are repended and the replay continues using the original data. If all targets have been met for all the respondent’s answers, the interview is terminated and data for any subsequent questions is discarded.

For example, suppose the BrandsBought question is quota controlled. The respondent chooses brands A, B, and C, answers additional questions about those brands, and then clicks Stop. When the respondent restarts the interview, the interviewing program reads the answer A, B, C for BrandsBought but finds that the target for brand C has been met, so it pends just the cells for brands A and B. The interviewing program comes to the questions based on BrandsBought and finds that it still has valid data for the questions, so the replay continues. The data for brand C remains in the case data even though it is no longer valid in terms of the current state of the interview.