There are things people will not type into a survey. A face-to-face interview reaches them, and the mechanism is not that the interviewer is friendlier — it is that a person can ask a second question.
The first answer is not the finding
Asked why they stopped using a service, most people give the socially acceptable reason: it was too expensive, they moved, they were busy. Asked "what made it feel expensive?" or "was there a particular occasion?", the answer changes — and the second or third answer is usually the one that would have changed the outcome.
A questionnaire cannot do this. It asked its question and it has to accept whatever came back.
Probing is a trained skill, not a longer script
Untrained interviewers do one of two things: accept the first answer, or lead the respondent to the answer they expected. Both produce data that looks fine.
Probes are written into the guide in advance, interviewers are briefed together on when to use them and when to stop, and a proportion of interviews are back-checked. The training is the cost of the method and it is what distinguishes an interview from a spoken questionnaire.
Showing something
Where a concept, a pack, a price list or an advertisement needs to be shown, it is presented in a controlled order so responses stay comparable across interviews. Rotation is planned rather than improvised, because an interviewer who shows the favourite first has already produced the result.
Reaching the people digital research misses
Online panels systematically under-represent older customers, lower-literacy respondents, and workers whose day does not include a screen. In Qatar that is a substantial part of several customer bases, and a digital-only programme quietly excludes them while reporting a national figure.
Face-to-face reaches them. It is the main reason the method survives in a market where everything else has moved online.
Language
Interviews run in Arabic, English, Hindi, Urdu, Malayalam, Tagalog and Nepali, and the interviewer is matched to the respondent rather than the respondent to a translation. A respondent working in their second language gives shorter answers and fewer of them, and the difference does not appear anywhere in the output — it looks like people simply had less to say. The full method is on the face-to-face interviews page.
What it costs
More per response than any other method, and it should be commissioned where that buys something. Sample sizes are smaller, so it does not replace a satisfaction survey. It answers why, where the survey has already established what.
Coding
Responses are coded to a frame built from the data rather than imposed on it, and the frame is supplied with the report. A coding frame written before the interviews were read will find what it was written to find.