Why the right survey method depends on the target population
Choosing CAPI, CATI or CAWI in Kenya requires more than comparing the price per interview. The core issue is whether each mode reaches the people whose experiences matter, allows them to understand the questions and protects their privacy. Research in Nairobi and Mombasa can reveal different recruitment conditions because one represents a diverse inland metropolis with technology and services and the other a coastal port city with tourism and trading activity. A method that succeeds with urban customers may fail for households with limited contact information.
A survey mode is a methodological choice, not a competition among devices. Research teams should map the intended population, expected contact information, survey complexity, preferred language and location constraints before buying fieldwork. For every population subgroup, ask whether a tablet-based visit, a telephone call or an online questionnaire offers a credible invitation to participate. The final proposal should identify who each method would systematically miss. Mode selection is defensible when those exclusions are visible before the sample is drawn.
CAPI: when in-person digital interviewing is strongest
CAPI works when observation and face-to-face selection are central to the design. Enumerators can follow preselected area segments, locate dwellings, display consent statements and show visual materials that are awkward by telephone. Digital logic limits some manual entry errors, but it cannot prove that the correct respondent was selected or that the interviewer was neutral. GPS points and duration checks are supplementary quality signals, never substitutes for a transparent sample. Device security, safe routing and interviewer training must be budgeted from the beginning.
CATI: when interviewer-led telephone surveys fit
CATI is often useful for tracking people already enrolled in a programme or recorded in a customer database. It can collect structured answers across dispersed locations quickly and reduce transport requirements. Yet the quality of those records determines whom the team can reach. Disconnected lines, wrong numbers and selective answering can change the composition of the achieved sample. Monitor contact rates by subgroup and location, keep the interview conversational and schedule follow-up calls at different times. Interpret survey results as estimates for the reachable frame unless the design justifies broader inference.
CAWI: when self-completed online surveys are credible
CAWI is attractive because responses flow directly into a dataset and participants can answer privately at a convenient time. However, digital access varies by age, resources, geography and user familiarity. The survey must be designed primarily for mobile screens where that reflects the audience, with simple navigation and response options. Verify the panel or invitation list, block duplicate participation where feasible and check unusual answer patterns without collecting unnecessary personal data. Findings should be described as applying to the recruited online population when broader coverage cannot be established.
Use the administrative research geography of Kenya
A location frame in Kenya should use the relevant geographic structures, such as counties, sub-counties and wards, rather than importing another country’s local government labels. The frame should distinguish where people live, work and access services from the administrative boundary used for supervision. A district or municipality may help organise field assignments without supplying a complete list of eligible respondents. Verify any boundaries and contact sources with field teams before sampling. These decisions affect who can enter the CAPI sample, which calling lists may be credible for CATI, and whether a CAWI panel covers the intended area.
Build a defensible sample before selecting devices
A questionnaire may be identical across modes while the achieved populations are not. Face-to-face routes can underrepresent workers absent during the day; phone frames may miss those without working numbers; web invitations can overrepresent connected or enthusiastic users. The design must identify these risks beforehand. Record initial selection probabilities where known, contact outcomes and subgroup gaps. Weighting can correct some measured imbalances only under defensible assumptions; it cannot manufacture coverage for an excluded population. Study conclusions should clearly name the actual survey universe.
Applying CAPI, CATI and CAWI in Nairobi
In Nairobi, the local research environment includes a diverse inland metropolis with technology and services. A household livelihood or service-access survey might use mapped clusters and CAPI where the team must visit selected addresses, verify assets or record in-person observations. A CATI follow-up could be considered when respondents have willingly provided functioning numbers, while CAWI may suit a separately defined digitally reachable subgroup. These methods must not be substituted casually: each can select a different type of participant. Pilot questions and contact procedures within selected neighbourhoods before calculating the city estimate.
Applying CAPI, CATI and CAWI in Mombasa
In Mombasa, the team should independently assess who can be recruited and how, taking account of its character as a coastal port city with tourism and trading activity. Check whether the listed households or users can be visited safely, whether sufficient valid numbers exist and whether online recruitment would exclude relevant communities. If a mode used in Nairobi proves weak here, an adapted design may be justified, but preserve core indicator definitions. Note that two selected centres do not represent Kenya as a whole; a claim about the entire country would require a wider sampling strategy.
Adapt questionnaire design for each survey mode
A single questionnaire copied unchanged into three systems is rarely good design. CAPI can support observations and visual exercises; CATI depends on spoken clarity and manageable call length; CAWI requires self-explanatory layouts and light pages. The team should maintain an indicator master and then create mode-specific presentation versions under change control. Pilot difficult recall questions, response scales and translated terms with the intended audience. In the final dataset, retain a mode variable so analysts can test whether responses differ systematically by collection channel.
Test the actual languages respondents use
Interviewers and respondents may prefer different languages across the same city. Plan local screening, consent and questionnaire testing in Swahili, English and relevant local languages, choosing versions based on the actual target group rather than a single presumed national language. CATI needs phrases that are easy to understand by voice and not overly long. CAWI needs instructions that remain clear without an interviewer. CAPI can support spoken clarification, but staff must not explain questions inconsistently. A short glossary, cognitive pilot and version-control register help maintain concept equivalence across languages and modes.
Plan access, travel and connectivity around local conditions
For this country, a practical design consideration is that county and ward boundaries provide meaningful local sampling geography. Evaluate whether this affects road access for CAPI, call completion for CATI, or digital coverage and completion for CAWI. Conduct a reconnaissance and pilot in each selected centre. Estimate realistic interview duration, interviewer workloads, callback needs, connectivity and secure data transfer rather than applying an unchanged budget from another country. If teams change modes mid-project, document why and test whether the change affects the measured indicators.
Know when mixing modes improves coverage
A mixed-mode design can be sensible if the modes cover complementary groups by design. For example, CAPI might reach selected households lacking reliable numbers, while CATI covers an existing beneficiary list; CAWI might be reserved for digitally active professionals. Do not recruit convenient respondents from all three channels and call the total nationally representative. Agree how invitations will be tracked to prevent duplicates, which questions may be mode-sensitive, and whether final results will be shown separately or weighted into a defined combined population. The reasons for mode allocation should appear in the methods section before results are interpreted.
Apply mode-specific data quality checks
Field quality is strongest when it is monitored daily rather than reconstructed after the target has been met. CAPI dashboards may show interview locations and durations, but anomalies need supervisor investigation. CATI supervision can compare contact rates and call outcomes between interviewers. CAWI requires fraud screening and inspection of device or response patterns within privacy limits. A quality flag is a reason to examine a case, not automatic proof of fabrication. Document the decision, resulting exclusions and any replacement interviews so final totals are auditable.
Treat consent and privacy as part of research design
Ethical practice varies by contact situation even when the same core principles apply. At a household doorstep, consent cannot be assumed because an enumerator has arrived with official-looking identification. On the telephone, the respondent must be able to verify the purpose of an unexpected call. Online, participation should be genuinely voluntary and the invitation should not disguise commercial or research use. Plan safe storage, retention limits and procedures for sensitive responses before any device is deployed. Where legal requirements are uncertain, obtain competent local review rather than guessing.
Analyse mode effects and report coverage limits
Results should be interpreted against the intended sample rather than against the size of the dataset. A two-city online survey of professionally connected adults cannot describe households lacking internet. An area-based CAPI sample may support more general claims for its defined areas if selection and response are properly handled. A CATI sample drawn from an NGO register speaks primarily about those on that register. Show denominators, disaggregation limits and uncertainty honestly. Methodological transparency is part of the insight delivered to decision-makers.
An illustrative survey decision for Kenya
Imagine a research organisation assessing Kenya residents’ experience of health or education services. One component may require direct observation of a facility, which favours CAPI; another may involve checking in with beneficiaries already enrolled on a verified contact list, which can favour CATI. A third component may solicit structured feedback from employees or registered online users, which may fit CAWI. The question is not which technology wins overall, but which population and evidence requirement each technology can legitimately serve. For Nairobi and Mombasa, set different logistics assumptions while applying the same rules for respondent eligibility and data quality.
A method-selection decision for Kenya should explicitly compare: population coverage; sample-frame defensibility; language and questionnaire fit; respondent privacy; requirements for direct observation; connectivity and travel constraints; consent and data protection; interviewer verification; nonresponse; and cost per valid, usable interview. A simple choice table can rank CAPI, CATI and CAWI against the particular study, rather than using a universal hierarchy. A cheap method that systematically misses core respondents is usually a poor value, while an expensive method with no extra evidence benefit may also be hard to justify.
What a credible survey report should contain
The study protocol should specify sampling units, city boundaries, contact or visit schedules, translated questionnaires, mode-specific recruitment procedures, quality checks, data storage and the rules for excluding or replacing cases. Reporting should provide separate denominator and nonresponse information for Nairobi and Mombasa when they are compared, with a clear limitation on inference beyond those populations. Good survey research gives programme managers and business leaders evidence they can use, plus enough operational detail to understand its strengths and gaps. It does not disguise coverage problems behind a high response count.
How Surveysphere Africa can support
Surveysphere Africa supports the design and delivery of CAPI, CATI, CAWI and carefully planned mixed-mode studies. For projects in Kenya, including Nairobi and Mombasa, the team can help define a credible sample, localise instruments, supervise fieldwork, verify data quality and explain which populations the evidence represents. The best method is the one that reaches the intended respondents and yields defensible decisions.
Related guidance
For the continent-wide view, read our guide to CAPI, CATI and CAWI.
Our services for research in Kenya
We run research across African markets. Start with these services for work in Kenya:
- Outsourced fieldwork in Kenya: CAPI, CATI and CAWI data collection with live quality control.
- Market and consumer research in Kenya: Market-entry, consumer and retail research.
We also offer: Monitoring, evaluation and learning, Social and development research, Public health research, Policy, economic and institutional research, Data analytics and insights visualisation.
