Data Trust & Governance for Dolibarr

Data Quality, Integrity & Governance for Business-Critical ERP Data

Know which Dolibarr data you can trust — and what needs attention.

ERP decisions are only as reliable as the data behind them.

• A customer without payment terms.
• Two companies sharing the same VAT identifier.
• A product with no reference or cost basis.
• An invoice whose totals do not reconcile.
• A payment allocated beyond the invoice amount.
• A stock record linked to a missing warehouse.
• An invalid IBAN.
• A duplicate product or third party created during an import.

Individually, these problems can appear small. Together, they affect reporting, finance, stock management, customer service, integrations, automation and management decisions.

Data Trust & Governance checks the business data already stored in Dolibarr and turns data-quality problems into structured, reviewable issues. It currently provides 36 documented deterministic checks covering third parties, products, invoices, sales orders, payment allocations, stock and bank-account data across five quality dimensions: Completeness · Validity · Uniqueness · Consistency · Integrity.

More Than a Data-Quality Dashboard

Data Trust is designed around a complete control workflow:

Assess → Detect → Prove → Assign → Correct in Dolibarr → Re-check → Verify

It does not simply display a percentage and leave the user to work out what went wrong. Each failed rule becomes a managed issue that your team can:

• investigate;
• assign;
• annotate;
• snooze;
• dismiss with a reason;
• place under a controlled exception;
• open directly in Dolibarr;
• re-check after correction;
• keep as historical evidence.

Most importantly, a user cannot simply click “Resolved” and make a problem disappear. An issue becomes Verified Fixed only after a later complete assessment of the same policy scope confirms that the same record now passes the same rule version. That makes Data Trust a data-governance workflow, not just a scanner.

Why Businesses Need Data Trust

Businesses often discover poor ERP data only after it causes another problem: financial reports affected by inconsistent invoices; duplicate customers or suppliers fragmenting history; invalid e-mail addresses preventing communication; missing payment terms affecting receivables or payables; unreliable product records affecting purchasing and stock analysis; broken relationships between transactions and master records; imports creating duplicates or incomplete third parties; payments linked incorrectly; inventory totals becoming inconsistent; invalid or shared bank-account information creating additional review risk.

Data Trust moves those problems from “someone may eventually notice” to a repeatable control process. It answers: What is wrong? Where is it? Why does it matter? What evidence supports the finding? Who owns it? Has it actually been corrected?

36 Deterministic Data-Quality and Integrity Checks

Third-party data

Blank names; invalid e-mail syntax; missing customer or supplier codes; missing country; missing address; missing customer or supplier payment terms; possible shared VAT identifiers; possible duplicate name + country combinations; shared customer-code and supplier-code candidates. Potential duplicate identifiers are treated as candidates for review, not as proof that two records are duplicates.

Product master data

Missing reference or label; duplicate normalized references; duplicate barcodes; missing or questionable cost basis; missing unit; negative physical measurements; leading or trailing whitespace in references; product stock inconsistent with warehouse totals.

Customer and supplier invoice integrity

Invoices referring to a missing third party; due date earlier than invoice date; invoice date unexpectedly far in the future; net amount + tax not matching the gross total (customer and supplier invoices); optional comparison between invoice header totals and invoice-line totals. High-impact integrity problems are clearly distinguished from lower-priority completeness issues.

Sales orders, invoice lines, payments and stock

• Sales orders referencing a third party that no longer exists.
• Invoice lines referring to products that are no longer present.
• Payment allocations referring to a missing invoice, or exceeding the gross invoice amount. Data Trust does not alter payment records itself.
• Stock records referencing missing warehouses; product stock differing from the warehouse-total calculation. These checks identify structural inconsistencies; they are not stock forecasting or automatic correction.

Bank-account validation

IBAN checksum failure and the same bank account associated with several third parties. Bank evidence is masked rather than exposing complete account numbers. Shared accounts are a review signal, not automatic proof of bad data, because legitimate group-company situations exist.

Every Finding Becomes a Managed Issue

The issue workspace shows why the issue matters, the recommended action, business impact, the rule involved, owner, priority, first detected and last observed dates, the assessment in which it appeared, values captured when detected, current live values, expected values, evidence history, checksums, notes and the complete event history. For duplicate candidates, related records can be compared side by side. The issue links directly to the relevant standard Dolibarr record, so the correction takes place in the ERP rather than in a separate data-cleaning database.

See “At scan” vs “Now”

The comparison At scan → Current value → Expected lets a reviewer see immediately whether somebody has already corrected the record since the issue was detected. Re-check this rule now runs a targeted assessment and, when appropriate, closes the issue through verified evidence rather than a manual status change.

Verified correction — not manual closure

For automatic verified resolution, Data Trust requires: (1) the same record; (2) the same rule version; (3) the same policy scope; (4) a later complete assessment; (5) the record now passing the rule. If an earlier issue returns after correction, it is identified as Came back rather than silently treated as a new, unrelated problem.

Flexible Assessment Policies

Business-type starting points: Distribution, Retail, Services, Manufacturing, Small Business. General sets: Essential checks, Master data, Transactions & Finance, Everything. Individual checks can then be enabled or removed according to company policy. The selected rules and scan scope are stored as a policy fingerprint, so later assessments can tell whether two runs are genuinely comparable. Administrators can also skip closed third parties and products that are neither sold nor purchased.

Honest Coverage — Complete, Partial or Not Assessed

Data Trust does not pretend that an incomplete scan means the data is clean. Every check finishes in one of three states:

• Complete — the entire relevant population was evaluated.
• Partial — the assessment stopped because of a configured row ceiling, time limit or memory protection.
• Not assessed — the required source, field or schema was unavailable.

Partial and not-assessed results are never presented as clean data and are never extrapolated to the unexamined population.

No Artificial “Trust Score”

Data Trust deliberately avoids reducing ERP data quality to a single unexplained score. It reports records evaluated, records flagged, pass rate, checks completed, coverage, priority and active issues. Every percentage is accompanied by its underlying counts, which makes the result easier to challenge, verify and explain to management.

Dashboard

Active issues, high-priority issues, assessment coverage, unassigned issues, new issues this week, verified fixes in the last 30 days, approved exceptions, latest assessment status, a setup checklist, a flagged-vs-evaluated trend with policy-change markers, active issues by rule, priority mix, age distribution, workload by owner and “What needs attention” guidance. Partial runs are excluded from trend interpretation rather than plotted as if they were complete.

Review Queue Built for Daily Work

Quick views: Active, Unassigned, Assigned to me, High priority, New this week, Waiting more than 7 days, Came back, Snoozed, Exceptions, Verified fixed, Dismissed, Record not found, All. Filter by rule, business area, quality dimension, priority, owner, status, business impact, age, first detected, last observed and assessment. Search, sorting, configurable page size, saved views, one-click Assign to me, CSV export and keyboard navigation are included.

Bulk review without losing control

Select several issues and assign, snooze, dismiss or add notes, always with a preview and a confirmation step. There is deliberately no bulk “mark fixed” shortcut, because fixes must be verified against the ERP data.

Duplicate Management Without Dangerous Automatic Merging

Duplicate candidate groups are shown side by side with reference, name or label, VAT identifier, barcode, related-document counts, recent activity and an advisory “most used” record. A shared identifier remains a review signal, not proof of duplication. Data Trust never merges records automatically. If records are legitimate, the reviewer marks the group Not duplicates, dismissing the related issues with a recorded reason.

Controlled Exceptions

Some unusual data is intentional. Exceptions can be requested, reviewed, approved, limited to a defined period, shown as expiring soon and automatically expired. A different administrator must approve an exception from the one who requested it. When an exception expires, the issue becomes active again if the condition still exists — exceptions stay visible and temporary instead of becoming undocumented permanent exclusions.

Import Check — Validate Before Data Enters Dolibarr

A CSV pre-import checker for third parties accepts a UTF-8 file of up to 5 MB and 50,000 rows, detects common delimiters and reports missing names, invalid e-mail syntax, unknown countries, repeated VAT identifiers, names or codes inside the file, and VAT/name/code values that already exist in Dolibarr. Nothing is imported and the uploaded data is not stored.

Assessment History, Comparison, Coverage and Insights

• History: every assessment is retained with status, start time, duration, records evaluated and flagged, the user who started it, policy fingerprint, per-rule results, scope and coverage (Complete, Partial, Not assessed, Failed).
• Comparison: if the policy or scan scope differs, the comparison refuses to present the change as like-for-like, so a lower issue count is never shown as an improvement when the second assessment simply checked less data.
• Coverage and readiness: per business area, whether the source exists, record counts, selected checks, checks that can run, checks blocked by missing fields and partial-read reasons. A table being readable does not mean the data is good.
• Insights: weekly flow of new, fixed and reopened issues; concentration by rule, business area and impact; recurring issues; and, for authorised users, how flagged records entered the ERP (import-tool markers, creation month, creating account). These are associations, not causes.

Five Configurable Alerts — Inside Dolibarr or by E-mail

1. Open high-priority issues
2. Old open issues (30+ days)
3. Issues with no owner
4. Summary by check
5. Reminders to owners (each owner receives only their own active issues)

Each alert is configured independently: delivery inside Dolibarr (a bell in the top bar, unread count and a My alerts inbox; no mail server needed), by e-mail, or both. Schedules: every 12 hours, daily or weekly at a chosen hour and weekday, or when something changes (an assessment can trigger evaluation immediately, with a minimum interval to avoid repetitive messages). The Alerts page shows what would be reported now, last sent, next send and recipient warnings. Alerts run from a Dolibarr scheduled job, a server command (cron / Windows Task Scheduler) or lightweight checks while authorised users work in Dolibarr; locking prevents duplicate sends and failed sends are retried. A 180-day history shows when each alert ran, items reported, new items, trigger, recipients and read counts. All alerts are off by default.

Strong recipient protection

Alerts can reach only permitted active internal Dolibarr users. Customers, suppliers, contacts, members and external users are excluded; an address also used on a third party, contact or member is refused when saving, before sending and during delivery, and delivery fails closed if validation cannot be completed. E-mails contain counts, rule IDs, issue titles and links — never the underlying record values.

Reports, Exports and Audit Trail

A printable management report for a published assessment shows scope, source areas read, rows read, a scorecard by dimension, checks completed, evaluated and flagged records, pass rates, results by rule and active exceptions, and notes that an assessment is an observation across a scan interval rather than an atomic snapshot. CSV exports (issues, filtered views, selected issues, assessment summaries, audit history, diagnostics) include spreadsheet formula-injection protection. The audit trail records detection, review, notes, ownership changes, snoozing, dismissal, exceptions, verified correction and recurrence — with time, user, issue, change, previous and new value and reason — and can be filtered and exported. It is application-level history, not database-administrator-proof immutability.

Designed for Large Data Sets

Assessments read in keyset batches with a configurable row ceiling, time budget, memory protection, streaming evaluation, evidence limits and explicit partial-run reporting. Synthetic development benchmarks completed 200,000 third parties under a 128 MB PHP memory limit and 500,000 under 256 MB. These are development benchmarks on synthetic data, not guarantees; real performance depends on database size, hardware and other workloads.

Read-Only by Design, No AI, No Telemetry

Data Trust reads Dolibarr business data but does not edit customers or suppliers, merge duplicates, alter products, repair invoices, modify payments, change stock, delete ERP records or block transactions. Its workflow data lives in its own datatrust_* tables. It uses deterministic documented rules: no AI, no telemetry, no external analytics. Business data stays on your Dolibarr server; the only outbound path is e-mail when an administrator explicitly enables an e-mail alert.

Security and Permissions

Separate rights for View, Run, Review, Configure and Export (administrators have all rights). CSRF protection, GET requests that never change state, output escaping, CSV-injection protection, entity scoping, page-level rights, hostile-parameter handling, CLI-only administrative scripts and protected background-alert requests. Exception approvals require internal administrators and a second approver.

Integrated Directly Into Dolibarr

Own top-level menu and left navigation, home-page widget, a Data Trust tab on third parties, products, customer invoices, supplier invoices and orders, the alert bell in the top bar, a scheduled assessment job and a scheduled alert job (registered disabled).

Who Uses Data Trust

• Data stewards: work the issue queue, assign, annotate, review duplicates, re-check corrected records, use controlled bulk actions.
• Finance teams: invoice inconsistencies, missing relationships, unusual due dates, payment-allocation problems, supplier-invoice issues.
• Sales and CRM administrators: incomplete customer data, duplicate third parties, missing codes, invalid contact information, missing payment terms.
• Product and inventory administrators: missing references, duplicate references or barcodes, missing units, cost-basis gaps, stock-integrity issues.
• Data owners and governance managers: dashboard, coverage, insights, management reports, exception governance, audit trail.
• Dolibarr administrators: policy, presets, scan scope, scheduling, alerts, retention and diagnostics.

Practical Use Cases

1. Clean customer and supplier master data through a repeatable queue instead of an occasional spreadsheet cleanup.
2. Validate a CSV before importing it into Dolibarr.
3. Review duplicate master records without auto-merging.
4. Detect transactions pointing at missing parent records.
5. Improve invoice-data reliability (dates, totals, header/line differences).
6. Review payment-allocation integrity.
7. Improve product and inventory master quality.
8. Run an ongoing governance process with scheduled assessments, owners and recurring-issue monitoring.
9. Govern legitimate exceptions with a second approver and an expiry date.
10. Show real improvement over time, because only equivalent policies are compared and fixes must be verified.

Languages and Requirements

Interface in English, French, Spanish, German and Italian. The non-English translations are machine-assisted and have not been reviewed by native-speaking professionals. Requirements: Dolibarr 21; PHP 7.4 or 8.2 (both tested); MySQL / MariaDB; current entity only. Not validated for Dolibarr 23, PostgreSQL, Odoo or cross-entity MultiCompany analysis.

Test Evidence

Automated validation covers alerts (133 assertions), permissions (236), language consistency (37,007), rules, adapters, streaming scans, case storage, exceptions, imports, insights and output security, and the baseline-policy rule counts were reconciled against independently written SQL on development data. This is test evidence, not a claim that every possible real-world database has been validated.

Honest Current Limits

• MySQL / MariaDB only; current Dolibarr entity only; no MultiCompany sharing.
• Dolibarr 23 and PostgreSQL not validated.
• Multicurrency amounts are not validated.
• No automatic record repairs, no automatic merging, no transaction blocking, no ledger reconciliation.
• Assessment history is not tamper-proof against a database administrator.
• Stopped assessments restart from the beginning.
• Performance figures come from synthetic development data.
• Real e-mail delivery has not yet been validated.
• Non-English translations have not received native-speaker professional review.

What Data Trust does not pretend to be: no universal trust score, no AI-generated conclusions, no automatic root-cause proof, no guaranteed savings or ROI, no automatic repairs or merging, no tamper-proof audit storage. Instead: documented rules + evidence + coverage + accountable workflow + verified correction.

Why Businesses Use Data Trust & Governance

To identify incomplete master data, detect invalid values, find potential duplicates, uncover broken relationships, spot inconsistent invoice and payment data, review inventory integrity, validate data before imports, assign ownership, separate real problems from accepted exceptions, verify that corrections worked, monitor recurring issues, document who reviewed what and maintain an auditable history of data-quality work. The objective is simple: turn hidden ERP data-quality problems into visible, evidence-backed, owned and verifiably corrected issues inside Dolibarr.

▶ Watch the demo video · Product page · Silicon Blaze YouTube channel

Support, videos and custom ERP work

Need help? For further assistance, please contact us at support@siliconblaze.com.

Videos: watch demos on the Silicon Blaze YouTube channel.

Website: siliconblaze.com

Custom ERP assistance: need a custom Dolibarr module, integration or ERP implementation? Contact Silicon Blaze at siliconblaze.com or support@siliconblaze.com.