Mobile data collection is not “paper but on a phone.” In low-connectivity settings it is the front door of the research data system. If the form is wrong, the skip logic is broken, or the team cannot work offline, no amount of elegant analysis later will fix the week you already lost.
This is how I think about CAPI tools after building and supervising field systems across East African programmes.
Tools I actually use
| Tool | Where it fits |
|---|---|
| ODK / ODK Central | Open, offline-first surveys; strong for multi-country research ops |
| KoboToolbox | Humanitarian and NGO field work; familiar to many partners |
| REDCap | Clinical and longitudinal research; role-based access; audit trails |
| XLSForm | The shared design language behind many CAPI stacks |
I pick tools with partners, connectivity, IRB constraints, and handoff in mind — not brand preference.
Design rules that prevent expensive failures
- Design for offline first. If the team works without signal, the form must still submit later without drama.
- Validate at entry. Ranges, required fields, and constraints beat end-of-week cleaning arguments.
- Name everything. Consistent variable names and codebooks are how analysis starts on day two.
- Version the instrument. A silent form change mid-wave is a schema break.
- Train on the why. Enumerators who understand skip logic catch problems interviewers alone cannot.
- Pair forms with HFC. Digital capture without high-frequency checks is just faster ways to store bad data.
What “good” looks like in the first week of field
- Piloted form with real respondents and real devices
- Daily completeness and constraint-failure summaries
- A named person responsible for form versions
- A path from server extract → cleaning script → supervisor view
On diaries work, that path continued into live supervisor apps. On oncology chart abstraction, REDCap plus multi-language QC scripts played the same role: protect the record while collection is still open.