Home/Blog/CATI vs CAPI vs PAPI vs CAWI: Survey Modes and CAPI Software Compared
Back to Blog

CATI vs CAPI vs PAPI vs CAWI: Survey Modes and CAPI Software Compared

May 25, 2026
25 min read
By DSTechnoverse
Survey Methods

CAPI, CATI, PAPI or CAWI? How the four survey modes compare on coverage, cost, interview length and data quality in India, how mode changes answers, and how ODK, KoboToolbox, SurveyCTO, CSPro and Survey Solutions compare as CAPI software.

CATI vs CAPI vs PAPI vs CAWI: Survey Modes and CAPI Software Compared

Quick answer: CAPI (computer-assisted personal interviewing) and CATI (computer-assisted telephone interviewing) are the two main ways professional surveys collect data today. CAPI means face-to-face interviews on a tablet or phone and is the standard for household, baseline and EIA surveys in India. CATI means interviews by telephone and suits short follow-ups and populations with reliable phone lists. PAPI (paper) and CAWI (web) fill narrower roles. Choose the mode by coverage first, then interview length, then cost and speed.

CAPI and CATI are the two data collection modes that most survey teams weigh first, and the choice between them shapes cost, data quality, reach and the time it takes to get from fieldwork to a clean dataset. Once you know what you want to ask, the next decision is how to ask it: face to face on a tablet, over the telephone, on paper, or through a web link the respondent fills in alone.

This guide compares CATI vs CAPI vs PAPI vs CAWI, explains how mode changes answers, and compares the main CAPI software options. Quality control, data cleaning and budgeting have their own guides, linked below.

The Four Survey Modes: CAPI, CATI, PAPI and CAWI Defined

Survey methodologists group data collection methods by two questions: is there an interviewer, and is the interviewer physically present? That gives four common modes.

ModeFull nameInterviewerChannelTypical use in India
CAPIComputer-Assisted Personal InterviewingYes, in personTablet or smartphone appHousehold, baseline, endline, EIA and R&R surveys
CATIComputer-Assisted Telephone InterviewingYes, remoteTelephone, with software on a computer or tabletFollow-ups, monitoring rounds, business and beneficiary lists
PAPIPaper-and-Pencil InterviewingYes, in personPrinted questionnaireSmall studies, device-restricted settings, backup
CAWIComputer-Assisted Web InterviewingNoWeb link on phone or computerEmployee, student, institutional and professional surveys

The "computer-assisted" part is what separates modern modes from older ones. In CAPI, CATI and CAWI the questionnaire is a program: it decides which question comes next, checks answers as they are entered, and stores data in a structured format with no separate entry step. In PAPI the questionnaire is a document, and every rule must be followed by a person.

CAPI Survey: How Computer-Assisted Personal Interviewing Works

In a CAPI survey, a trained enumerator visits the respondent and reads questions from a tablet or smartphone, entering answers directly into a survey app.

What CAPI Does Well

  • Enforced routing. Skip logic and relevance conditions run automatically, so an enumerator cannot ask a question about school fees in a household with no children.
  • Validation at the point of entry. Range and consistency checks catch errors while the respondent is still present.
  • Rosters. Household member lists, plots, livestock and loans can be repeated as many times as needed, with each row linked to its household.
  • Field verification data. GPS coordinates, start and end times, per-question timings and optional audio audits give the supervisor evidence that an interview happened and was done properly.
  • Offline operation. Interviews are stored on the device and synced later, which matters in villages with weak mobile data.

Where CAPI Is Weaker

  • Devices, power banks, cases and the server or platform account have to be budgeted and managed.
  • Programming takes time, and a badly programmed form can lock in an error across every interview.
  • Enumerators need training on the device as well as the questionnaire; see training enumerators for a field survey.
  • Charging, device theft and screen glare are real field problems, so every team should still carry printed backup forms.

The upfront programming effort is recovered on any study above a few hundred interviews, which is why CAPI is the default for baseline surveys for CSR and development projects and socio-economic surveys in EIA.

CATI Data Collection: How Telephone Surveys Work

In CATI data collection, an interviewer calls the respondent, reads questions from a screen and enters answers into software. Larger operations use a call centre with a dialler, call scheduling, disposition codes for every attempt and supervisor monitoring; smaller ones can use a CAPI app while calling from a mobile phone.

The Attempt Protocol

Telephone response depends heavily on persistence. A protocol should fix the maximum number of attempts per number and spread them across mornings, evenings and weekends, with callbacks at the respondent's chosen time. Apply it evenly: an interviewer who gives up early on hard-to-reach numbers biases the sample towards people who are easy to contact. Calculate outcome rates using standard definitions, such as those of the American Association for Public Opinion Research, so that rounds are comparable.

Where CATI Works and Where It Does Not

CATI is a strong choice for:

  • Follow-up rounds with households already interviewed in person, where a mobile number was collected at baseline.
  • Beneficiary lists from a programme database, farmer producer organisations or self-help groups, where phone numbers are recorded and reasonably current.
  • Business, institutional and professional surveys where every unit has a working phone line.
  • Rapid monitoring, where a short set of indicators is needed quickly.
  • Situations where travel is impossible.

It is a poor choice for a general-population rural survey. There is usually no complete list of phone numbers to sample from, the poorest households and many women are under-represented on phone lists, and the practical length of a telephone interview is much shorter than a household survey needs. Interviews longer than about 15 to 20 minutes see rising break-offs, so CATI questionnaires should be designed short from the start rather than cut down from a face-to-face instrument.

PAPI: When Paper Surveys Still Earn Their Place

PAPI, or paper-and-pencil interviewing, is the traditional method: enumerators fill in printed questionnaires by hand, and the forms are later keyed into a database. It needs no devices and no programming, which makes it look cheap and simple.

The hidden costs arrive after fieldwork. Every form has to be transported, stored, and entered, ideally twice by different operators so that the two entries can be compared. Skip errors made in the field cannot be fixed because the respondent is gone. Supervisors cannot see GPS or timing data, so verification relies on back-checks alone. The time from last interview to a clean dataset is typically measured in weeks rather than days.

Paper still makes sense in three cases:

  1. Device-restricted settings, where carrying tablets creates security or acceptability concerns, or where institutional rules do not allow them.
  2. Very small studies, where programming and device costs exceed the data entry savings.
  3. Backup, when a device fails or is lost mid-fieldwork. A CAPI study should always carry a stock of printed forms that mirror the programmed questionnaire.

If you use paper, pre-code every answer, print skip instructions beside each question and pre-print household IDs on every page.

CAWI: Web Surveys and Where They Fit

CAWI, or computer-assisted web interviewing, is self-completed: the respondent receives a link by SMS, WhatsApp, email or QR code and answers on their own device. There is no interviewer to explain questions or probe, which lowers cost and can increase honest reporting on sensitive topics, but also raises the risk of misunderstanding and careless answering.

CAWI works well when the population is defined, literate and online: employees of an organisation, students, members of a professional body, suppliers, or researchers. It is rarely suitable as the main mode for household or community surveys in India, because coverage of online self-completion falls sharply with income, age, education and gender in many districts. It is also harder to confirm who actually answered.

Keep it short, design for a phone screen and avoid large grids. For opinion scales in web and other modes, see how to analyse Likert scale survey data.

CATI vs CAPI vs PAPI vs CAWI: The Full Comparison

CAPI, CATI and paper compared
FactorCAPI (in person, device)CATI (telephone)PAPI (paper)CAWI (web)
Cost per completed interviewMediumLowMedium to highLowest
Rural household coverageExcellentPoor to fairExcellentPoor
Interview length tolerated60 minutes or more15 to 20 minutes60 minutes or more10 to 15 minutes
Skip logic errorsPrevented by the softwarePreventedCommonPrevented
Range and consistency checksAt entryAt entryOnly after data entryAt entry
GPS verificationYesNoNoNo
Audio auditYes, with consentYes, if recorded with consentNoNot applicable
Separate data entryNoneNoneDouble entry neededNone
Time from field to clean dataDaysDaysSeveral weeksDays
Photographs and documentsYesNoNoUploads possible
Works without connectivityYes, syncs laterNoYesNo
Disclosure of sensitive behaviourModerateOften higherModerateOften highest

Two rows drive most decisions in India. Rural coverage rules out CATI and CAWI as the primary mode for most general-population household surveys. Time from field to clean data is the reason paper has largely disappeared from professional survey work: data entry adds weeks and creates a new source of error.

Cost is relative: CATI saves travel but may need many more numbers dialled than interviews completed. For how mode feeds into a budget, see how much a socio-economic survey costs in India.

Coverage in India: Who Each Mode Actually Reaches

Coverage error happens when some members of the target population have no chance of being selected. It is the most important difference between modes, and it cannot be fixed by increasing sample size.

GroupCAPICATICAWI
Households in remote villagesReached, if accessibleOften missed where network is weakOften missed
Women in households where a man holds the phoneReached, with a female enumerator where neededFrequently missedFrequently missed
Older and less literate respondentsReachedReached if they have a phoneRarely reached
Migrant and seasonal workersMissed if away during fieldworkReachable if number is currentSometimes reachable
Businesses and institutionsReachedReachedReached

Phone ownership has grown widely, but ownership by a household is not the same as reaching the right respondent. Shared handsets and frequently changed SIM cards reduce effective coverage. If you must use CATI for a population with patchy coverage, compare the achieved sample with a known benchmark such as Census or programme data, and use weighting cautiously. For how sampling frames and selection work in field surveys, see sampling methods for household surveys.

Mode Effects: Why the Same Question Gets Different Answers

The same question asked face to face, by phone and on a screen produces systematically different answers.

EffectWhat happens
Social desirabilityRespondents report more socially approved behaviour when an interviewer is present, especially face to face
SatisficingTelephone respondents more often choose early or late options in long lists because they cannot see them
AcquiescenceAgreement bias tends to be stronger by telephone
RecallFace-to-face interviews allow probing, calendars and visual aids, which improve recall of events and amounts
Unit non-responseTelephone refusal and non-contact rates are usually much higher
Item non-responseHigher by phone for income and other sensitive items
Household compositionHarder to enumerate accurately without being in the household

The practical risk is largest in panel studies. If a baseline was face to face and the endline is by telephone, part of the measured change will be mode effect rather than programme effect, and it can be larger than the change you are trying to detect. A difference-in-differences design partly protects you only if treatment and comparison groups switch mode in the same way; see difference-in-differences with baseline and endline data and baseline vs endline surveys.

When a mode change is unavoidable, interview a random subsample in both modes, estimate the difference for each key indicator, and report it. For other sources of systematic error in field data, see how to reduce bias in field surveys.

CAPI Software Compared: ODK, KoboToolbox, SurveyCTO, CSPro and Survey Solutions

The five platforms below are the ones most often used for household and development surveys. All of them support offline data collection on Android devices, skip logic, validation and repeat groups. They differ in who maintains them, how forms are written, how data is hosted and how much quality-control tooling is built in. Pricing and plan limits change, so check each provider's current terms directly.

PlatformMaintained byLicence modelHow forms are writtenHostingStrongest at
ODK (ODK Collect and ODK Central)Get ODK and the ODK open-source communityOpen source; paid hosted option availableXLSForm in a spreadsheetSelf-hosted server or provider-hostedFull control of data, flexible deployments, longitudinal Entities
KoboToolboxKobo, a non-profit organisationFree tier with usage limits; paid plans for heavier useWeb form builder or XLSFormKobo's public servers or self-hostedQuick deployment for NGOs and humanitarian teams
SurveyCTODobilityCommercial subscriptionXLSForm-based, with an online designerProvider-hosted, with optional encryptionBuilt-in quality-control and monitoring tools
CSProUS Census BureauFreeCSPro designer with its own logic languageSync via CSWeb server or file-sharing optionsCensuses and large national household surveys, also paper data entry
Survey SolutionsWorld BankFreeWeb-based Designer with its own expression syntaxSelf-hosted or hosted Headquarters serverComplex rosters, supervisor review workflow, CAPI plus web mode

ODK

ODK is a family of open-source tools: ODK Collect is the Android app enumerators use, and ODK Central is the server that stores forms, submissions and user accounts. Forms are written in XLSForm, a spreadsheet format that many other platforms also use, so skills transfer easily. Teams can self-host, which matters when data must stay on infrastructure the client controls. Features include encrypted submissions, Entities for linking data across visits, web forms for browser-based entry, and an API for exporting data into analysis tools.

KoboToolbox

KoboToolbox is built on ODK technology and is widely used by NGOs and humanitarian agencies. It offers a drag-and-drop form builder alongside XLSForm upload, the KoboCollect Android app and Enketo web forms. Its main attraction is speed of set-up: a simple form can be live the same day. Public servers are run by Kobo, with usage limits on free accounts (check current terms), and organisations can also self-host.

SurveyCTO

SurveyCTO is a commercial platform that also uses XLSForm as its base, so forms written for ODK can often be adapted. It is widely used by research organisations running impact evaluations because of its quality-control tooling: audio audits, text audits that record time spent on each question, sensor meta-data, automated quality checks, a review-and-correction workflow before data is accepted, and case management for panel and follow-up studies. It can also be configured for phone-based interviewing.

CSPro

CSPro (Census and Survey Processing System) is free software developed by the US Census Bureau and used for many national censuses and large household surveys. Questionnaires are built in a Windows designer, interviews run in the CSEntry app on Android or Windows, and the same system handles paper-form data entry, editing and tabulation. It has a powerful logic language, which suits very complex questionnaires, but the learning curve is steeper than XLSForm-based tools.

Survey Solutions

Survey Solutions is a free platform developed by the World Bank. Questionnaires are designed in a web Designer, data is collected with the Interviewer app on Android, and a Headquarters server manages assignments, supervisors and data. Its strengths are nested rosters, large household questionnaires and a structured workflow where supervisors review, reject and return interviews to enumerators. It also supports a web mode for self-completion, which makes it a candidate for mixed CAPI and CAWI designs.

Features That Matter When Choosing CAPI Software

The serious platforms all handle the basics. The differences that matter for a particular project are usually in the following areas, which you should test on a pilot form before committing.

  1. Offline operation and sync. Interviews must save offline and sync without duplicates.
  2. GPS capture. Check accuracy settings and whether location can be required.
  3. Timestamps and per-question timing. These let you identify interviews completed implausibly fast.
  4. Audio audits. Random recording, with consent, is a strong check on enumerator behaviour.
  5. Rosters and nested repeats. Test the deepest nesting your questionnaire needs, such as members within households and plots within members.
  6. Language support. Hindi and regional scripts must display correctly on your devices.
  7. Preloading and case management. Panel surveys need baseline names and identifiers loaded into follow-up forms.
  8. Encryption and access control. Check encryption on device, in transit and on the server.
  9. Export formats. Check that exports work cleanly into the tools your analysts use, with labels intact.
  10. An audit trail. The platform should record edits and who made them, which supports the processes described in our guide to data quality control in large-scale surveys.

A short pilot with real enumerators on the actual devices will surface more problems than any feature list. For how to build the questionnaire itself before programming it, see how to design a baseline survey questionnaire.

Sample Size and Response Rates by Mode

Mode affects how many interviews you need and how many people you must approach to get them. The starting point is the standard sample size for estimating a proportion:

n0 = z² × p × (1 − p) / e²

where z is 1.96 for 95% confidence, p is the expected proportion (0.5 is the most conservative), and e is the margin of error. For p = 0.5 and e = 0.05, n0 is about 384.

Two adjustments then depend on mode.

  • Design effect (deff). CAPI household surveys usually sample villages first and households within villages. Households in the same village are similar, so each extra interview adds less information. The required sample becomes n = deff × n0. CATI from a list is often drawn without clustering, which lowers the design effect, though weighting for unequal coverage raises it again.
  • Response and contact rates. The number of units you must approach is n divided by the expected completion rate. Household CAPI surveys in rural India often achieve high completion when fieldwork is well managed, while telephone surveys have far lower contact and completion rates.

Full guidance on sample size, including finite population corrections and subgroups, is in how to determine sample size for a household survey.

Worked Example: Choosing Modes for a Three-Round CSR Evaluation

The numbers below are illustrative only, chosen to show the reasoning. They are not benchmarks for any specific project.

The brief. A company is funding a livelihoods programme in 40 villages, with 40 comparison villages. It needs a baseline, a midline after one year and an endline after three years. Key indicators are household income, women's participation in paid work, and adoption of three farming practices. The baseline questionnaire, including a household roster, runs to about 55 minutes.

Step 1: Baseline mode. The instrument is long, has a roster, and phone lists are incomplete. Baseline is CAPI.

Step 2: Baseline sample size (illustrative). Taking n0 = 384 per arm for a proportion at ±5%, an assumed design effect of 2.0 for village clustering gives 768 per arm. Allowing for 10% non-response gives about 854 per arm, or about 1,700 households across both arms, which is around 21 households per village.

Step 3: Collect follow-up contacts. At baseline, record one or two mobile numbers per household and the name of the person who usually answers, with consent for follow-up calls.

Step 4: Midline mode. The midline needs only a short set of indicators for programme monitoring. A 15-minute CATI module is designed using the same wording as the baseline for those items. Suppose, illustratively, that 60% of baseline numbers are reached and complete the call. That gives roughly 1,020 completed interviews; the team checks whether reached and unreached households differ at baseline.

Step 5: Mode effect check. Because the endline will be compared with the baseline, the team interviews a random subsample of about 150 households by both CATI and CAPI at midline. Suppose the share reporting women's paid work is 4 percentage points higher by phone than face to face in this subsample. That difference is recorded as an estimate of mode effect for that indicator.

Step 6: Endline mode. The endline is the impact estimate, so it returns to CAPI with the full baseline instrument for all indicators being compared.

Step 7: Analysis. The impact is estimated with difference-in-differences between treatment and comparison households, using baseline and endline data collected in the same mode. Midline CATI results are reported as monitoring, with the mode effect estimate noted.

What the example shows. Mixing modes saved effort at midline without contaminating the impact estimate, because every indicator used for the main comparison was measured in the same mode at both ends. For how this design feeds into analysis, see how baseline and endline surveys power impact evaluation.

Mixed-Mode Designs That Hold Together

Combining modes is often sensible. The common patterns are:

PatternHow it worksMain risk
CAPI baseline, CATI follow-upsIn-person enrolment, phone for light-touch roundsMode effects and attrition among households that change numbers
CATI first, CAPI for non-respondersPhone attempts first, field visits for those not reachedTwo modes within one round complicate comparison
CAWI with CATI reminderWeb link, then phone call for non-completersDifferent answers for those who self-complete

Three rules make mixed-mode work defensible:

  1. Keep the mode constant for any indicator you intend to compare across rounds or groups.
  2. Record the mode as a variable in every dataset, so it can be included in analysis and any effect examined rather than assumed away.
  3. Use the same question wording and response options across modes, adapting only what the channel requires, such as reading options aloud on the phone instead of showing a card.

Step-by-Step: Moving a Survey From Paper to CAPI

  1. Finalise the paper questionnaire first. Programming an unsettled questionnaire multiplies rework.
  2. Write a specification recording each question's variable name, type, codes, skip condition and validation rule.
  3. Choose the platform against the feature checklist above, including where data must be hosted.
  4. Program, translate and desk-test every path with dummy data, including extreme values.
  5. Pilot in the field with real enumerators on the actual devices, checking timings, language and layout.
  6. Train on the device and the content together, including the switch to backup paper forms, and set up daily monitoring as part of your data quality control plan.
  7. Freeze and version the form before fieldwork, and agree variable naming so the export flows straight into cleaning survey data before analysis.

Choosing Your Survey Mode: A Decision Guide

Most decisions follow from coverage, interview length and whether rounds will be compared.

SituationBest fit
Rural household baseline or endlineCAPI
EIA socio-economic or R&R surveyCAPI
Short follow-up of enrolled householdsCATI, same wording as baseline
Survey of businesses, FPOs or institutions with phone listsCATI, or CAWI plus CATI
Staff, student or member surveyCAWI
Very small study or device-restricted settingPAPI
Impact evaluation with comparison groupCAPI at baseline and endline

If you are commissioning fieldwork rather than running it, the mode an agency proposes says a lot about its approach; see how to choose a household survey agency and how to conduct a socio-economic survey step by step.

How DSTechnoverse Helps

DSTechnoverse plans and runs survey data collection across India for CSR teams, NGOs, EIA consultants, project developers and researchers. Through our socio-economic and baseline survey service we help you choose the right mode, design and translate the questionnaire, program CAPI forms in KoboToolbox, ODK or SurveyCTO-class tools, and run CAPI fieldwork, CATI follow-ups or paper surveys where they fit, with GPS checks, audio audits and daily monitoring. Our statistical and research data analysis service then cleans, analyses and reports the data, including mode-effect checks and baseline-endline comparisons. Contact us to discuss your survey.

Frequently Asked Questions

What is the difference between CAPI, CATI, PAPI and CAWI?

CAPI is face-to-face interviewing on a tablet or phone. CATI is interviewing by telephone with answers entered into software. PAPI is face-to-face interviewing on paper, which needs separate data entry. CAWI is a web survey the respondent completes alone. CAPI is the standard for household surveys in India, while CATI suits short follow-ups.

Which survey mode is best for rural India?

CAPI, in most cases. Telephone and web modes miss many of the poorest households, women in households where a man controls the phone, and areas with weak network coverage. CAPI also handles the long questionnaires, household rosters and observations that rural household surveys usually need, and records GPS and timing data for verification.

Is CATI cheaper than CAPI?

Usually, per completed interview, because there is no travel. But CATI often needs many more numbers dialled than interviews completed, it limits interview length to about 15 to 20 minutes, and it suffers coverage gaps in rural populations. For a general-population household survey, the lower cost rarely justifies the loss of representativeness.

What is the best CAPI software for a household survey?

There is no single best option. KoboToolbox and ODK are widely used and quick to set up; SurveyCTO offers extensive built-in quality-control tools; CSPro suits censuses and very complex questionnaires; Survey Solutions is strong for nested rosters and supervisor review. Choose by testing GPS, timing, audio audits, rosters, language display and export on a pilot form.

Can ODK or KoboToolbox be used for CATI?

Yes, for many studies. An interviewer can call the respondent from a phone and enter answers into the same form on a tablet or browser. What these tools do not provide by themselves is full call-centre functionality such as automated dialling and call scheduling, so attempt tracking must be designed into the form or managed separately.

Do people answer differently on the phone than in person?

Yes. By telephone, respondents tend to show more agreement bias, choose earlier or later options in long lists more often, and skip sensitive items such as income more often. Face to face, they often give more socially approved answers. These mode effects matter most when rounds are compared.

Can I switch from CAPI to CATI between baseline and endline?

You can, but not safely for indicators you intend to compare. The mode change alone can shift responses by more than the programme effect. If a switch is unavoidable, interview a random subsample in both modes, estimate the difference for each key indicator, and report it with the results.

Is paper (PAPI) still used for surveys?

Yes, in limited cases: very small studies, settings where devices are not allowed or not safe, and as backup when a device fails. Paper requires double data entry and reconciliation, and cannot prevent skip or range errors in the field, so time from fieldwork to clean data is much longer than with CAPI.

When is CAWI a good choice?

CAWI works when the population is literate, online and reachable through a known list, such as employees, students, members of a professional body or institutions. It is low cost and can encourage honest answers on sensitive questions, but it is rarely suitable as the main mode for household surveys in India because of coverage gaps.


How DSTechnoverse Runs Your Survey

DSTechnoverse is a full-service socio-economic and baseline survey agency based in Indore. We do not just advise — we run the whole study end to end and hand you evidence you can defend:

  • Design — objectives, indicators, and sampling (random, stratified, cluster, or multi-stage), with a piloted questionnaire translated into local languages.
  • Field data collection — trained enumerators using CAPI (tablet) tools such as KoBoToolbox, ODK, and SurveyCTO-style apps, with GPS stamping, audio audits, and built-in skip-logic validation.
  • Quality control — back-checks, spot-checks, and daily high-frequency data monitoring so problems are caught in the field, not after.
  • Analysis — SPSS, R, Python, and Excel for indicators, cross-tabulations, and impact estimates, including baseline–endline difference-in-differences.
  • A detailed report — cleaned datasets, indicator tables, charts, GIS maps, and clear interpretation, with interactive Power BI / Tableau dashboards on request.

Our survey experience in Indore includes Smart City citizen-feedback surveys, Startup India & MSME surveys, Swachh Survekshan (Swachh Bharat) cleanliness surveys, and skill-development impact evaluations — delivered for government, municipal, corporate, and NGO clients.

Planning a socio-economic, baseline, endline, or willingness-to-pay survey? Talk to our survey team for a scope and quote.

Sources and Further Reading

Official and primary sources for the standards, rules and data discussed on this page.

Share

Tags

CAPICATIdata collectionsurvey methodsfield survey

Explore More Articles

Discover more insights and updates from DSTechnoverse

View All Articles