Does CAPI mean the respondent completes the survey alone?

Usually no. The interviewer uses a computerized instrument.

Is a video interview the same mode?

It is a remote interview; report that distinction rather than assuming equivalence with in-person fieldwork.

Can face-to-face surveys be part of a mixed-mode design?

Yes, with attention to how mode affects responses.

AI Glossary / Statistics and probability / Intermediate

Face-to-face survey

Definition

A face-to-face survey collects responses through an interviewer and respondent meeting in person. The interviewer usually asks standardized questions and records the answers, on paper or using a digital questionnaire. Face-to-face describes the mode of data collection, not how respondents are selected. Its statistical quality depends on sampling, question design, fieldwork, response, and analysis as well as the personal interaction.

What it means

An interviewer can explain the survey process, present visual materials, and guide respondents through a structured questionnaire. This can be useful when questions are complex or other response modes are difficult to access.
Personal contact also changes the measurement situation. Respondents may answer differently in front of another person, and interviewers can unintentionally influence responses. A face-to-face interview is therefore neither automatically more accurate nor automatically more representative than an online or telephone survey.

How it works

Define the target population and select respondents using a documented sampling design. If interviewing households, specify how an eligible person is selected within each household. Interviewing whoever answers the door can produce a different sample from the intended design.
Develop and test the questionnaire. Standardized wording, clear response options, appropriate translations, and manageable length help make answers comparable. Distinguish permitted clarification from leading or suggesting an answer.
Train interviewers to introduce the study, identify eligible respondents, follow the question sequence, and record answers consistently. Plan contact attempts and callbacks so that people who are unavailable at one time are not systematically excluded.
Choose paper or computer-assisted personal interviewing, commonly called CAPI. A digital instrument can implement skip patterns, range checks, and consistency prompts. Those checks can catch some recording mistakes, but cannot establish that every answer is true or that every eligible person was reached.
Provide an appropriate setting for the interview and explain how information will be handled. Questions that are sensitive to another person's presence may require particular attention to privacy or a self-completion component. Any recording or AI-assisted transcription should follow the study's stated consent and data-handling arrangements.
Monitor fieldwork using contact outcomes, interview duration, missing responses, and appropriate verification procedures. Differences among interviewers can reflect respondent assignment as well as interviewer behavior, so investigate patterns before attributing causes.
Analyze with the sampling design and response process in mind. Weights may account for unequal selection and adjustments, but they do not automatically remove all nonresponse or measurement bias. If modes are combined, examine whether mode differences affect comparability.

Three practical examples

A household survey uses trained interviewers to administer a structured questionnaire at selected addresses.
A product study presents physical prototypes during in-person interviews, while keeping selection and questioning procedures consistent.
A community survey uses face-to-face interviews where digital access is limited, with translated materials and a documented respondent-selection process.

Think of it this way

Think of a guided questionnaire. The guide can help respondents navigate the questions, but must avoid steering their answers or choosing only the easiest people to interview.

Worked example

An invented fieldwork plan selects 200 addresses. Twenty are confirmed ineligible, leaving 180 eligible addresses. Interviewers complete surveys at 126 and obtain no completed survey at the remaining 54.
Under this deliberately simplified situation, with no unknown eligibility and all selected eligible addresses in the denominator, the completion proportion is 126/180 = 70%. If 63 completed respondents report using a service, the unweighted respondent proportion is 63/126 = 50%.
These calculations answer different questions. Seventy percent describes completion among eligible addresses; 50% describes the responses obtained. It does not follow that 50% of all eligible households use the service.
If users and nonusers differ in their likelihood of responding, the respondent proportion can be biased. Real survey response-rate standards may distinguish partial interviews, refusals, noncontacts, and estimated eligibility; this simplified arithmetic is not a universal reporting formula.

Illustrative case study

Fictional case: Lakeside Research surveys residents about access to local services. Interviewers initially visit only during weekday working hours, producing many noncontacts. The team adds evening and weekend callbacks and documents all outcomes. It also checks whether one household member is answering for others when individual responses are required. This invented case illustrates how fieldwork choices influence whose views enter the dataset.

Common misconceptions and mistakes

Do not confuse face-to-face collection with probability sampling. An in-person convenience sample remains a convenience sample.
Do not assume a high completion rate proves absence of bias or that a low rate quantifies the bias by itself. The relationship between response and the measured outcome matters.
An interviewer should not fill missing answers from personal impressions. Digital checks and AI transcription can assist recording, but they are not substitutes for the respondent's answer and proper quality review.

Business use

Face-to-face surveys can support research requiring visual materials, complex routing, or access to populations poorly served by other modes. Organizations should budget for travel, training, callbacks, and supervision, and compare the resulting information with the cost. Reports should state the sampling method and collection mode separately so that users can judge the scope of the findings.

Check your understanding

What is the denominator for the example's completion proportion?

Answer: The 180 confirmed eligible addresses.

Why is 50% not automatically a population estimate?

Answer: It describes respondents and may be affected by selection and nonresponse.

What should a digital skip rule accomplish?

Answer: Route the respondent to the appropriate next question consistently.

Related terms

Survey mode
CAPI
interviewer effect
nonresponse bias
sampling frame
mixed-mode survey
questionnaire
response rate

Learn next

Study probability sampling and questionnaire testing. Then examine interviewer training, survey response classifications, and how mixed-mode designs can affect measurement.

Sources

NIST/SEMATECH statistical glossary, “face-to-face survey”: https://www.itl.nist.gov/div898/handbook/glossary.htm U.S. Census Bureau, CAPI/CATI design-guideline research: https://www.census.gov/library/working-papers/2023/adrm/rsm2023-09.html U.S. Census Bureau, ACS response modes: https://www.census.gov/programs-surveys/acs/respond/other-ways-to-respond.html

Previous: Fabric computing Next: Facet theory