When a market research company lands a door-to-door survey project, the first thing it usually does is not rush to hire interviewers — it answers one question: what system will we use to collect the data?
Building a CAPI (Computer-Assisted Personal Interviewing) system comes down to four things: which system to use (selection) → how to deploy it (go-live) → how interviewers use it (training) → whether the data is trustworthy (quality control). These four are tightly linked; a weak link in any one of them will eventually show up in data quality.
Below, we walk through a practical, deployable method for building an interviewing system in these four steps.
Step 1: Decide First — Build or Buy?
This is the most easily skipped yet most cost-impacting step. Technical leads at market research companies often fall into two traps: thinking "it's just an app — our IT team can write one," or "the project is small, so pen and paper will do."
The real criterion for whether you need a CAPI system is not scale but the level of data-quality requirements:
- High data-quality requirements (the client demands audio recording / GPS / re-checks / end-to-end quality control) → use CAPI regardless of project size. Small projects work too — a mature system like TLSUN CAPI supports low-cost entry for small projects and can go live quickly even for a few hundred questionnaires, depending mainly on questionnaire length;
- Low data-quality requirements (internal exploration, no quality control) → paper questionnaires or a lightweight online tool are enough;
- Build in-house: for organizations with a long-term product strategy, sufficient budget, and an R&D team. But be clear-eyed — a CAPI that can withstand a national-level project hides a pile of "invisible engineering" behind it: offline fault tolerance, sample selection, audio quality control, and quality-control feedback loops. The hidden cost of building is often far higher than buying;
- Buy a mature system: for most market research companies — ready to use out of the box, backed by proven case studies, maintained and iterated by the vendor, with costs that scale elastically with project size.
In one sentence: what matters isn't the project's size but its data requirements — if your data must withstand scrutiny, use CAPI.
Step 2: Selecting a System — the 6 Key Capabilities
The value of an interviewing system is not "it can record questionnaires" but "it can keep data quality in check." When selecting, focus on these six — every one of them can be verified with a live demo from the vendor:
- 1. Offline data collection: door-to-door and on-site work is the core scenario, and unreliable networks are the norm. The system must let interviewers open questionnaires, enter answers, and run skip logic with no network, then sync once online. This is the essential dividing line between CAPI and online survey tools (see Offline Data Collection).
- 2. Quality-control checks — audio + GPS: what clients and supervisors care about most is "did the interviewer really go, and really ask?" The system must record full interview audio and GPS location so every interview is traceable and auditable.
- 3. Quality-control loop — review, send back, re-visit: a professional CAPI system should support it — supervisors review responses, flag issues, and send them back to the interviewer with comments; interviewers can receive, respond, correct, or re-visit once online, instead of redoing everything after the full data collection. When selecting, be sure to confirm: can supervisors leave per-question comments and send them back, and can interviewers correct and resubmit online?
- 4. Sample management: large surveys are not "random door-knocking" — they follow strict rules for household sampling, household-member sampling (such as the common Kish grid method), and backup sample replacement. The system should support these sampling logics natively rather than relying on manual Excel work.
- 5. Low operational barrier: interviewers are mostly temporarily recruited part-timers with little training time. A paper-like questionnaire interface significantly lowers onboarding and reduces entry errors.
- 6. Data security and backup: offline data encrypted locally, with multi-tier backup to prevent data loss if a device is lost or damaged.
Turn these six into a checklist and tick them off one by one during selection, and you'll avoid most of the pitfalls.
Step 3: Deployment and Go-Live
Once the system is chosen, deployment is a relatively "standard" step, but three details are easy to overlook:
- Device selection: tablets first (battery life, portability, paper-like feel), laptops as backup; buy 10–20% spare devices beyond the number of interviewers.
- Account and permission tiers: a mature system should use three tiers — administrator, supervisor, and interviewer. Supervisors are further split into field supervisors (assign samples and review questionnaires) and quality-control supervisors (review the response data within their scope of authority, with two-level review). If permissions aren't layered properly, you get chaos like samples getting crossed or data being altered.
- Server deployment model: SaaS cloud service (out of the box, low ops burden) or on-premises deployment (data stays inside your network, for government or classified projects). Choose based on the client's data-compliance requirements.
Step 4: Interviewer Training and Pilot
A system's ceiling is "easy to use," but a project's floor is "whether the interviewers can actually use it." Training is best done in three layers:
- Basic training: device operation, login, downloading task packages, data entry, sync — half a day to a day;
- Mock interviews: run 2–3 mock households with the real questionnaire, walking through details like skip logic, required-field validation, and starting audio recording;
- Pilot (pre-test): before the formal project, run a small pilot of 10–30 households to surface problems before rolling out at scale. This round's value far exceeds its cost.
Step 5: Quality Control and Supervision — Who Is Accountable for Data Quality
This is the "last mile" of building an interviewing system, and where a system shows its true value. During formal fieldwork, supervisors do four things every day:
- Watch progress: completion counts and sample recovery rates per interviewer, catching interviewers who are "abnormally slow" early;
- Listen to recordings: sample interview audio to check whether questions were asked properly and without leading;
- Verify GPS: whether the interview location matches the sample address, preventing "sitting at home filling out questionnaires";
- Watch for conflicts: whether samples were interviewed twice or data looks abnormal, with automatic system alerts.
What matters even more is "what happens after a problem is found." A mature system relies on a quality-control loop: after interviewers sync data, quality-control supervisors review the responses, flag issues by leaving per-question comments, and send them back to the interviewer; interviewers can receive the comments once online and respond, correct, or re-visit — instead of redoing everything after all data is collected. TLSUN CAPI is designed exactly this way: quality-control supervisors review the response data within their scope of authority, leave per-question comments, and support two-level review; interviewers receive the comments online and correct or re-visit.
These actions are unrealistic to do by manually checking every questionnaire — they must be supported by a "supervisor dashboard" in the system. This is why a mature CAPI system's supervisor-dashboard and quality-control-feedback capabilities matter more than the interviewer app itself.
Build vs. Buy: At a Glance
| Dimension | Build in-house | Buy a mature system (e.g., TLSUN CAPI) |
|---|---|---|
| Time to launch | 6–18 months | Days (a few hundred questionnaires can go live quickly) |
| Cost | Continuous R&D + operations investment | Per-project / annual subscription or on-premises; low-cost entry for small projects |
| Quality-control capability | Build audio / GPS / sampling yourself | Audio, GPS, and sample management out of the box |
| Risk | Heavy trial-and-error, no case-study backing | Backed by large projects such as CSS / CRRS |
| Fit | Organizations with a long-term product strategy | Most market research companies |
FAQ
Do market research companies always need a CAPI interviewing system?
It depends on data-quality requirements. As long as the data must withstand scrutiny (audio / GPS / re-checks / end-to-end quality control), use CAPI regardless of project size; small projects can also launch at low cost and quickly. For internal exploration with no quality-control requirements, paper questionnaires or a lightweight online tool suffice.
What matters most when selecting an interviewing system?
Data-quality control: audio recording, GPS location checks, sample management, and whether quality-control issues can be sent back to interviewers to form a re-visit loop. Second comes the interviewer-side data-entry experience.
Is CAPI suitable for small-scale surveys?
Yes. CAPI's value is data-quality control, not project scale. TLSUN CAPI supports low-cost entry for small projects and can go live quickly even for a few hundred questionnaires (depending on question length) — as long as data quality matters, it is worth adopting.
After quality control finds an issue, how does it get back to the interviewer?
After interviewers sync data, quality-control supervisors review the responses and can leave per-question comments to send issues back; interviewers receive the comments once online and can respond, correct, or re-visit, forming a quality-control loop that avoids redoing everything after collection.
How do I choose between on-premises deployment and SaaS?
For classified or government projects requiring data to stay on your intranet, choose on-premises; for general commercial research, SaaS is faster and cheaper to launch.
Which large projects has TLSUN CAPI served?
TLSUN CAPI has long served the Chinese Social Survey (CSS), the China Rural Revitalization Survey (CRRS), and the CAS longitudinal household survey of ethnic minority populations in southwest China, completing 100,000+ interview samples.
Final Thoughts
Building an interviewing system is a systems-engineering task: choose the right system, deploy it solidly, train well, and let quality control hold the line. Rather than stumbling from scratch, it's worth first seeing how a mature system that has withstood national-level projects actually works.
To get ready for your next door-to-door or on-site project, feel free to contact us for a demo and a proposal of the TLSUN Computer-Assisted Personal Interviewing (CAPI) System.