- CATI means a telephone survey with a programmed questionnaire: dialling, filters, call-back scheduling and recording are managed by a central system.
- It is the right method for existing customers, repeated monitoring, satisfaction studies and short opinion polls.
- The questionnaire must fit into 10-20 minutes; beyond that, drop-out rises and answer quality falls.
- Telephone sampling depends on the number frame: a customer database allows random selection, the general population requires random digit dialling and careful weighting.
- A telephone survey cannot show visual materials, finds it harder to obtain answers on sensitive topics, and under-represents groups that are hard to reach by phone.
1. What CATI is and how a research call centre works
CATI (Computer-Assisted Telephone Interviewing) is the telephone survey in which interviewers work from a central system that manages the list of numbers, dials the calls, displays the programmed questionnaire and records the answers. Unlike an ordinary call, the system decides which number is dialled, when a busy or unanswered number is called again, and how many attempts are made before a number is closed. Filters and validations work as in CAPI: the interviewer cannot skip a question or record an impossible answer.
A CATI study has three operational advantages that no other method combines. The supervisor listens to interviews in real time and steps in immediately when an interviewer misreads a question. Data are available daily, so the distribution of the sample is tracked as it is collected. And fieldwork can be stopped, corrected and resumed the same day, which makes the telephone survey the standard instrument for repeated monitoring.
2. What the telephone survey is suited to
- Existing customers. A company holds its customers' telephone numbers, so the sampling frame is complete and selection can be random. Satisfaction, loyalty and post-service evaluation studies are almost always conducted by CATI. The guide on customer satisfaction and NPS shows how such a study is built.
- Repeated monitoring (tracking). Brand awareness or trust in an institution, measured monthly or quarterly with the same short questionnaire. Speed and low cost allow frequent waves.
- Short opinion polls on current topics, where the result is needed within days.
- B2B studies with managers or specialists, where the list of companies exists and a scheduled 15-minute call is easier to obtain than a visit.
- The second wave of a CAPI study: respondents who agreed to be re-contacted can be re-interviewed by telephone, to measure change over time at a much lower cost.
The common criterion: the population can be reached by telephone from a known list, and the questionnaire is short. When either condition is missing, the telephone survey starts to lose ground to the face-to-face interview or, for connected urban populations, to the online survey.
3. Telephone sampling: everything depends on the number frame
The quality of a telephone survey is decided at the sampling frame, before the first call. There are three situations, with three levels of confidence.
A customer or member database. The list is complete and known, every person has an equal probability of selection, and the sample is drawn at random, possibly stratified by region, product type or tenure. This is the ideal situation, and the results generalise to the whole customer base with the calculated margin of error.
Random digit dialling (RDD). For the general population there is no list of numbers. Valid numbers are generated at random from the mobile operators' number ranges and all of them are dialled, including non-existent ones, which means a large share of calls reach nobody. The method produces a probability sample of people who own a mobile phone and answer unknown numbers, but it systematically under-represents older people, rural residents in areas with poor coverage, and those who do not answer calls from unknown numbers. Weighting partly corrects these deviations, but it cannot correct what is missing entirely.
Purchased lists or telephone panels. Lists of numbers aggregated from various sources, with unknown coverage. They are useful for quick studies, but the results cannot be presented as representative of the population, however large the sample.
4. Questionnaire length: the 15-minute rule
Attention on the telephone is limited. In practice, a CATI questionnaire works well up to 12-15 minutes and acceptably up to 20; beyond that, respondents start to abandon the interview or answer mechanically, and the quality of data in the last sections visibly declines. A 30-minute questionnaire designed for CAPI is not "adapted" to the telephone by reading faster; it is rebuilt.
A few construction rules specific to the telephone survey:
- Scales have at most five points and are read out in full only once; 7- or 10-point scales are hard to retain without visual support, although the 0-10 scale works because respondents know it.
- Answer lists have at most four or five options; long brand lists become spontaneous questions ("which brands of X do you know?") followed by individual confirmations.
- The order of options is rotated automatically, because on the telephone respondents tend to pick the last option they heard.
- Open questions are limited to one or two, because the interviewer types them in real time.
- The first question is simple and relevant to the respondent; socio-demographics come at the end.
5. Fieldwork: calls, call-backs, response rate
A number is not dialled only once. The system schedules call-backs at different hours, usually 5-7 attempts over several days, including evenings and weekends, before the number is closed as "not contacted". Without call-backs, the sample fills up with people who answer at the first attempt, and they are not the same as busy people, shift workers or people who travel.
Response rates to telephone surveys have fallen steadily in recent years as commercial calls have increased. A cooperation rate of 20-40% of contacted people is common for the general population, higher for the customer base of a well-known company. The agency must report the number of numbers dialled, contacts, refusals and completed interviews, so that the rate can be interpreted. A low rate does not invalidate the study, but it calls for weighting and caution in interpreting subgroups with very different response rates.
6. Quality control in a telephone survey
CATI offers the most direct quality control of all methods, because everything happens in one place. The standard checks:
| Check | How it is done | Reasonable standard |
|---|---|---|
| Real-time monitoring | The supervisor listens to interviews in progress without the interviewer knowing which | Every interviewer, daily, at least one interview |
| Audio recording | All interviews are recorded, with the respondent's consent | 100%, with a 10-15% sample reviewed |
| Duration per interview and per question | Automatic; abnormally short interviews are flagged | 100% |
| Drop-out rate by section | Where respondents give up; indicates problematic questions | Analysed after the first 50 interviews |
| Distributions per interviewer | Refusal rate, average duration, variation in answers | For all interviewers, halfway through fieldwork |
7. Where the telephone survey stops
The method has four limits that cannot be removed by better organisation. It cannot show anything: packaging, advertisements, maps, long lists; everything that must be seen requires CAPI or CAWI. It obtains sincere answers less easily on sensitive questions, because the respondent is talking to a stranger, sometimes with other people present, without the time to build trust. Coverage of the general population is incomplete and uneven, and the missing groups cannot be recovered through weighting. And duration limits depth: a segmentation study with 60 attributes does not fit into 15 minutes.
These limits do not make the method weaker, but more specialised. For the study it is made for, the telephone survey is faster, cheaper and better controlled than any alternative. For the study it is not made for, it is a false economy.
8. When CATI, when something else
| Research question | Suitable method | Why |
|---|---|---|
| How satisfied are our customers with the service they received last month? | CATI | The contact base exists, the questionnaire is short, the result is needed quickly |
| How has brand awareness evolved over the last three months? | CATI or CAWI | Short questionnaire, repeated identically; the choice depends on the target population |
| What do rural residents think about a reform? | CAPI | Incomplete telephone coverage, a topic that requires trust |
| Which of the three packaging designs is more attractive? | CAWI or CAPI | They have to be seen |
| What are household incomes and expenditures? | CAPI | Long questionnaire, sensitive topic, general population |
| What do 200 managers think about a public policy? | CATI | The list exists, scheduling a call is simpler than a visit |
9. What to ask the agency for a telephone survey
- The source of the telephone numbers and its coverage of the target population.
- The number of dialling attempts per number and the time slots.
- The estimated questionnaire length, confirmed by a pilot of 20-30 interviews.
- The call report: numbers dialled, contacts, refusals, completed interviews.
- The percentage of interviews monitored and checked against recordings.
- The weighting variables and the reference source.
- The timetable: programming, pilot, fieldwork, delivery, usually 7-14 days in total.
The telephone survey is the data collection method with the best balance of speed, cost and control, provided the population can be reached by phone and the questions fit into a quarter of an hour. When those two conditions are met, it is the right choice; when they are not, the saving is paid for in data that do not describe the people they were meant to describe. Sample size and margin of error are explained on the methodology FAQ page.
See also: CAPI and CATI Quantitative research Face-to-face interviews (CAPI) Online surveys (CAWI) Customer satisfaction and NPS 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.