Hi everyone,
I'm Mehrab Ali. I've been using ODK for field surveys for many years at ARCED Foundation.
One challenge we've repeatedly faced is managing data quality during fieldwork rather than waiting until data collection ends.
Our earlier workflow relied on Stata, including an open-source tool called arceddataflow that we developed to organize project data-processing scripts.
We've recently built a desktop application called DataFlow around this experience. It connects directly to ODK Central and helps research teams review incoming submissions.
Here's how it works with ODK Central:
- Sync: Connects to a Central project and retrieves new submissions without modifying anything on the server.
- Quality checks: Runs checks for duplicates, interview duration, outliers, ranges, skipped and don't-know responses, other-specify answers, logic inconsistencies, GPS, and back-check differences.
- Fieldwork monitoring: Identifies the interviews and enumerators associated with flagged issues so supervisors can investigate while fieldwork is ongoing.
- Corrections: Records each correction, supports undoing changes, and exports a Stata do-file to reproduce the cleaned dataset from the raw submissions.
The AI processing runs locally, without sending survey records to external AI services.
One current limitation: If a submission is edited in ODK Central after DataFlow has downloaded it, that edit isn't picked up by subsequent syncs yet.
I'm particularly interested in how others handle this last issue. When submissions are edited in Central, how do you reconcile those changes with quality-control decisions already made outside Central?
For transparency, DataFlow is a commercial application developed by our team. I'm sharing our approach because this has been a recurring challenge in our own ODK projects.
Thank you.