- CAPI means a face-to-face interview with a questionnaire programmed on a tablet: filters, validations, GPS and audio recording are automatic.
- It is the compulsory method for nationally representative samples, for rural and older populations, for questionnaires longer than 30 minutes and for sensitive topics.
- The sample is built in stages: localities, then households, then the person within the household, with strict replacement rules.
- Quality control verifies a defined share of interviews through GPS, duration, audio and telephone re-contact; without it, a face-to-face interview is no better than an online survey.
- A national CAPI study usually takes 4-8 weeks and costs several times more per interview than CATI or CAWI; the cost is justified by coverage and depth, not by tradition.
1. What CAPI is and how an interview unfolds
CAPI (Computer-Assisted Personal Interviewing) is the face-to-face interview in which the interviewer administers the questionnaire from a tablet rather than on paper. The difference sounds technical, but it changes data quality in four ways. Filters and skip logic are programmed, so the interviewer cannot take a wrong route through the questionnaire. Validations are automatic: an age of 150 or an income left blank is flagged on the spot. The duration of every question is recorded, which makes it possible to detect interviews completed suspiciously fast. And the tablet records GPS coordinates and, with the respondent's consent, audio fragments, which make later verification possible.
A typical CAPI interview for a national survey lasts 30-45 minutes and takes place at the respondent's home. The interviewer introduces themselves, explains the purpose of the study and the confidentiality rules, selects the person in the household according to the set rule, then reads the questions exactly as worded and records the answers. For questions with many options, the respondent is shown cards or the tablet screen. At the end, the data are synchronised with the server, and the field coordinator sees the same day how many interviews were completed, where and how long they took.
2. When the face-to-face interview is the only valid method
The face-to-face interview is not "better" in general. It is better for certain populations and certain questions, and outside them it is merely more expensive. The situations in which CAPI is compulsory:
- Nationally representative samples of the general population. Telephone and internet do not cover every group evenly; the only sampling frame that covers everyone is the frame of localities and households, and it can only be used at the doorstep.
- Rural, older or low-income populations, where internet access is limited and telephone numbers do not exist in a database from which they could be selected at random.
- Long questionnaires, over 30 minutes. On the telephone, attention fades after 15-20 minutes, and online after 10-12. Face to face, a 45-minute questionnaire is completed normally.
- Visual materials: packaging, advertisements, long brand lists, complex scales. They can be shown only in the interviewer's presence or online, and online excludes part of the population.
- Sensitive topics that require trust to be built: income, health, voting intention, experiences of discrimination, violence. A trained interviewer obtains answers that a telephone call does not.
- KAP studies and programme evaluations in communities, where beneficiaries must be identified in the field and where comparability with the baseline requires the same method.
If your population is a company's existing customers, with telephone numbers in a database, CAPI is almost always a waste; a telephone survey does the same job in a third of the time. If the population is urban internet users under 45, an online survey is appropriate. The criterion is the population, not the method.
3. Sampling for CAPI: three stages and strict rules
A national CAPI study uses multistage probability sampling. The population is stratified by region, urban or rural residence and locality size, so that each stratum is represented proportionally. Selection then proceeds in three stages.
Stage one: localities. Within each stratum, localities are selected at random with probability proportional to population, so a large town has a higher chance of entering the sample than a small village, but every resident ends up with the same probability of selection. For a sample of 1,100 people, the number of sampling points is usually between 100 and 150, with 8-12 interviews at each point. More points with fewer interviews each means a smaller design effect and better precision, but higher travel costs.
Stage two: households. At each sampling point, the interviewer starts from a randomly chosen address and follows a fixed rule for walking the streets, known as a random route, visiting for example every third dwelling. The rule removes the interviewer's natural tendency to choose the more accessible houses or the friendlier people.
Stage three: the person. In the selected household, the respondent is not whoever opens the door but the person chosen by a rule independent of the interviewer: the next-birthday method or a selection grid. If the selected person is not at home, the interviewer returns, usually up to three times, at different hours. Replacing the household is allowed only after these call-backs and only with the next household on the route. Without these rules, the sample fills up with people who are at home during the day, and the result is representative of pensioners and mothers with small children, not of the population.
4. The instrument: programming, pilot, languages
The CAPI questionnaire is first written on paper, with every filter marked explicitly, then programmed in the tablet application. Programming is a verification step in itself: every skip must be tested along every possible path, and validations (age ranges, sums that must total 100%, mandatory answers) are defined question by question. In the Republic of Moldova, the instrument is programmed in Romanian and Russian, and the respondent chooses the interview language; the translation is checked through back-translation, so that a question measures the same thing in both languages.
The pilot is run on 20-30 real interviews, at sampling points that do not enter the final sample. Actual duration is measured, the questions on which respondents ask for clarification are identified, and the filters are checked. After the pilot, the questionnaire is adjusted and reprogrammed. A CAPI study without a pilot is a study in which errors are discovered after 400 interviews, when they can no longer be corrected.
5. Fieldwork: interviewers, training, supervision
The quality of a CAPI study lies in the interviewer network. A national survey needs 40-80 interviewers, distributed by region, each with 10-25 interviews. All of them go through training on the specific questionnaire, not only on interviewing technique in general: every question is discussed, its meaning explained, what to do when the respondent does not understand, how to record a refusal. Interviewers receive a list of sampling points, the starting address and the route rule, and the regional coordinator checks progress daily.
The response rate is reported and interpreted. In a doorstep survey, between 30% and 60% of contacted households accept the interview, depending on region, topic and season. The rate is not a mere efficiency indicator: if refusals concentrate in one group, for example young urban men, the achieved sample departs from the structure of the population, and the deviation is later corrected through weighting. The agency must report the structure of the achieved sample against the planned one, halfway through fieldwork and at the end.
6. Quality control: what a serious agency checks
The face-to-face interview carries a specific risk: the interviewer may fill in the questionnaire alone, shorten the interview or interview whoever is convenient. That is why quality control is not an extra but the part that justifies the price of the method.
| Check | What it detects | Reasonable standard |
|---|---|---|
| GPS coordinates | Interviews conducted somewhere other than the sampling point | 100% of interviews, checked automatically |
| Duration of the interview and of each question | Interviews completed too fast or with skipped sections | 100%, with an alert threshold below 60% of the median duration |
| Audio recording of some sections | Questions not read or reworded, invented answers | A random 10-20% of interviews, listened to |
| Telephone re-contact | The interview took place, with the right person, for the stated duration | 15-20% of interviews, randomly selected per interviewer |
| Analysis of distributions per interviewer | Abnormal patterns: identical answers, lack of variation, too few refusals | For all interviewers, before fieldwork closes |
Interviews that fail verification are removed and redone, and the interviewer concerned is taken off the project. Ask the agency for the quality control report as a separate deliverable: how many interviews were verified, by which method, how many were rejected.
7. The limits and the cost of face-to-face interviewing
CAPI has limits that an honest proposal mentions. The cost per interview is several times higher than CATI and many times higher than CAWI, because of travel, field time and verification. The duration of a national study is usually 4-8 weeks from questionnaire approval to the cleaned data set, against 1-2 weeks for a telephone survey. The interviewer effect exists: the presence of a person can influence answers to questions about socially disapproved behaviour, and different interviewers can obtain slightly different distributions; training and per-interviewer analysis reduce but do not eliminate this effect. Finally, physical access to some areas may be limited, and a proposal must state clearly which territories are excluded from coverage.
8. CAPI, CATI, CAWI: the choice in brief
| Criterion | CAPI (face to face) | CATI (telephone) | CAWI (online) |
|---|---|---|---|
| Population covered | The whole population, including rural, older, offline | People with a telephone, existing contact databases | Internet users, panels |
| Questionnaire length | 30-45 min, up to 60 | 10-20 min | 10-15 min |
| Visual materials | Yes | No | Yes |
| Sensitive topics | Yes, with a trained interviewer | Limited | Yes, through anonymity, but with no control over who answers |
| Study duration | 4-8 weeks | 1-2 weeks | 3-10 days |
| Cost per interview | High | Medium | Low |
| Typical use | National, social, KAP surveys, evaluations | Satisfaction, tracking, existing customers | Concept tests, campaigns, urban segments |
9. What to ask the agency for a CAPI study
- The number of sampling points and the number of interviews per point.
- The household and person selection rule, and the replacement rule.
- The number of call-backs before replacement.
- The territories excluded from coverage, and why.
- The number of interviewers, the length of training and who delivers it.
- The percentage of interviews verified through GPS, audio and re-contact, and what happens to those rejected.
- The interim field report: the structure of the achieved sample against the planned one.
- The weighting variables and the source of the reference data.
- The timetable by stage: programming, pilot, fieldwork, cleaning, delivery.
For the general population, the face-to-face interview remains the reference data collection method against which all others are judged. Not because it is old, but because it is the only one that reaches everyone. When the question requires it, the cost is justified; when it does not, the same money buys a larger sample or a second wave with a faster method. Details on sample size, margin of error and weighting are on the methodology FAQ page.
See also: CAPI and CATI Social and political surveys Quantitative research Telephone surveys (CATI) Online surveys (CAWI) Methodology FAQ
Do you have a decision to take?
Tell us the problem, not the method. We propose the right research design and explain what it can and cannot deliver.