ODK Field Data Quality Toolkit

ODK field data quality starts before the first interview. Use this toolkit to check the form, monitor collection, review audit evidence, and document decisions before analysis.

The checklist is a starting point for supervision. It does not decide whether an interview is valid. A short interview, an unusual GPS point, or a high edit rate is a reason to review the record and its field context.

Before fieldwork

Build the checks into the form while changes are still cheap:

• Add constraints to numeric, date, and range questions. An impossible answer should be rejected at collection time.
• Use relevant expressions for questions that do not apply to every respondent. This reduces accidental answers in hidden sections.
• Mark consent, identifiers, and other required fields explicitly.
• Add an audit row before the pilot. Audit logging cannot be added retroactively to interviews that are already complete.
• Set identify-user=true when the team needs to connect interviews to enumerators.
• Enable location tracking when the field protocol requires location evidence, and explain the collection protocol to enumerators.
• Test translations, repeats, calculations, media, and choice filters on the device used in the field.
• Run a complete interview in airplane mode, reconnect, submit it, and compare the exported record with the answers entered.

Use the existing XLSForm constraint examples, skip logic guide, and ODK Collect setup guide while preparing the pilot.

For a practical audit configuration, start with this row on the survey sheet:

type	name	parameters
audit	audit	track-changes=true identify-user=true location-priority=balanced location-min-interval=60 location-max-age=120

◀ ▶surveychoicessettings+

The exact parameters depend on the study protocol, device settings, expected interview length, and battery constraints. Review them during the pilot instead of treating one configuration as universal.

During fieldwork

Review collection data on the same day when possible. A daily check gives the supervisor a chance to clarify a form question, retrain an enumerator, or repeat a small number of interviews while the field team is still available.

Signal	Question to ask	Follow-up
Interview duration	Are interviews far shorter than the form and protocol allow?	Review the interview, form path, and enumerator context.
GPS evidence	Do recorded locations fit the sample area and visit plan?	Check accuracy, device behavior, and assignment records.
Submission volume	Is one enumerator submitting an unusual number of records?	Compare output with workload, dates, and travel time.
Answer changes	Are important answers being edited unusually often?	Review the affected records and the interview protocol.
Sync behavior	Are finalized forms staying on devices or failing to send?	Check app-user access, connectivity, storage, and form version.

These signals are prompts for review, not automatic findings. Enumerator performance monitoring explains how to compare them with assignment records and field context.

After fieldwork

Before analysis, preserve the evidence and record what the team decided:

Use the ODK submission review checklist for the Central-side receipt, review-state, export, repeat, and media checks that precede this quality review.

• Export submissions and media from ODK Central.
• Keep audit.csv with the dataset when audit logging was enabled.
• Reconcile server submissions with the sample frame and enumerator records.
• Review the most unusual duration, location, edit, and volume patterns.
• Record the reason for every excluded interview.
• Keep the original exports before applying cleaning or recoding steps.
• Note which quality checks were unavailable because the form did not collect the required evidence.

The last item matters. A report cannot recover GPS, user identity, or answer-change history that the form never recorded.

What the audit log can support

The ODK audit log records events such as question visits, timing, answer changes, user identity, navigation, and available device locations when the relevant parameters are enabled.

Evidence	Possible investigation
Timing and interview duration	Rushed interviews, unusual pauses, or a form path that needs clarification
User identity	Which enumerator submitted the interview
GPS readings	Whether the recorded location fits the assignment and sample area
Answer changes	Whether key fields were edited repeatedly or late in the interview
Navigation events	Whether the form was completed in a way that matches the protocol

The evidence still needs context. A weak GPS fix may reflect a device or environment problem. A short interview may be valid for a simple respondent. A high edit rate may reflect a confusing question. Document the interpretation rather than turning one signal into a verdict.

Review the evidence in DataSnap

You can analyze an exported audit.csv manually or use a DataSnap report to review the available evidence in one place. The current workflow includes overview, activity and time-use summaries, question friction, flagged interviews, interview details, and a GPS view when the source contains the required data.

For the upload steps, see How to Analyze Audit Log Data. For specific investigations, continue to curbstoning detection, enumerator monitoring, or the survey data quality checklist.

A simple supervision record

For each review, record:

1. The interview or enumerator pattern that triggered review.
2. The source evidence used, such as duration, GPS, audit events, or assignment records.
3. The field context supplied by the supervisor or enumerator.
4. The decision, including whether the record was retained, corrected, repeated, or excluded.
5. The person and date responsible for the decision.

This makes the cleaning process easier to explain to a research lead, reviewer, or donor later.

Check ODK data quality for free

Read the ODK audit log guide