Drop a Kobo, ODK, Excel or CSV export. In seconds, Dalili computes a trust score, flags anomalies and tells you what this data does — and does not — let you conclude.
No structural issue detected. Solid base.
No anomaly detected by the 11 rules.
The Description and Unit columns are blank in the export: fill them in to get a real data dictionary.
Ranked by priority. Fix the urgent items before reporting or targeting with this data.
Dalili Check is the diagnostic. To track quality over time, audit several datasets, share with your team or get a signed report to show a donor, move to Dalili Studio.
Decision support, not a statistical certification. The score reflects what the rules detect in this file, not the representativeness of the collection.
A Kobo/ODK export (CSV, JSON), an Excel workbook or any CSV. The file is read by your browser, never sent.
Trust Score 0–100, 6 sub-scores, A/B/C/D level, row-by-row anomalies and structural issues.
Report, anomaly list, data dictionary and recommendations — to share with your team or your donor.
The Field Data Trust Score and its 6 sub-scores are computed by published rules, not by an opaque AI. The same engine runs in your browser, in Dalili Studio and in the local toolkit: a given file always produces the same result.
Share of rows with every critical field (ID, date, enumerator, GPS, zone).
Duplicates, impossible values, non-conforming types and formats.
Forms filled too fast, abnormal daily volumes, identical answers.
Missing, out-of-zone or frozen GPS; under-covered zones.
End before start, submission before start, outside the collection window.
Numeric values outside the interquartile fence or with a high robust z-score.
missing_critical_fieldduplicate_respondentimpossible_gpssurvey_duration_too_shortenumerator_volume_anomalyidentical_response_patternbusiness_logic_violationoutlier_numeric_valuetemporal_inconsistencycoverage_gapschema_validationA ≥ 85 · B ≥ 70 · C ≥ 50 · D below. Any critical anomaly (GPS out of country, negative count, end before start) caps the level at C; if systemic, at D.
Dalili Check runs entirely in your browser: no account, no upload, no trace. When you move to Dalili Studio, you pick the confidentiality level that suits you — it is the same engine everywhere.
Dalili Check: nothing leaves, not even a fragment.
Studio on data without names or phones, or on 50–500 rows.
The same engine as a command line on your laptop, zero upload.
Dalili Studio runs on your servers (Docker).
Per-organisation isolation, audit logs, encryption, DPA.
Findings point to a row and a column, never to a personal value.
Dalili Check stays free, with no limit. Dalili Studio adds history, teamwork, connectors and signed reports.
Methodological honesty. Dalili provides decision support, not a certification. Every deliverable documents its usage limits. A Starter tier (20,000 FCFA/month, 3 projects) exists for small teams — ask for it.
No. Dalili Check runs in your browser. You can go offline once the page has loaded: the analysis still works. Nothing is stored.
No. Dalili sits on top of your tools: you keep collecting as you do today, Dalili audits what they produce.
The tool is built for everyday datasets, up to a few tens of thousands of rows. Beyond that, use a sample or Dalili Studio.
By 11 deterministic rules and 6 weighted sub-scores, described in the Method section. No opaque model: the same input always gives the same output.
Dalili Check saves nothing, by design. Download the folder before closing the tab. To keep a history, create a Dalili Studio account.
Dalili Studio tracks quality over time and produces signed reports. The team also delivers full audits (Field Data Trust Audit) and M&E support — write to us.
A full audit, a self-hosted deployment, a question about the method: write to us.