DataCleaner Pro: Duplicate Third Parties & Contacts (Detection, Preview and Safe Merge)

Your customer database has duplicates. You already know it.

The same customer entered twice, once as "DUPONT SARL" and once as "Dupont S.A.R.L.". Invoices on one side, quotes on the other. An outstanding balance that means nothing any more.

DataCleaner Pro does three things, and does them properly.

1. It finds duplicates, and explains why

The engine normalises what needs normalising, legal forms, dotted acronyms, accents, phone formats, email case, then compares. Every match is justified criterion by criterion: "same VAT number", "same phone", "company name 89 %". Never a percentage out of nowhere.

Three rules keep the noise out: two different VAT numbers dismiss the pair, a similar name with nothing else in common never reaches high priority, and the "very likely" level requires a strong identifier. On our labelled test sets, high-confidence precision is 100 % at 1,000 and 10,000 third parties.

2. It shows you what will happen before merging

How many invoices, orders, contacts and documents on each side. Which fields are kept, which ones conflict. Which record it recommends keeping, and why: you stay free to swap. Nothing is ever merged automatically, not even at 100 %.

3. It merges without losing anything: including what Dolibarr loses

For third parties the module relies on the core mechanism: Dolibarr itself relinks its 32 business classes. But from Dolibarr 21 to 24, the native merge deletes the absorbed third party's document directory before trying to move it: files are lost, and the merge still reports success. DataCleaner copies the documents beforehand.

And above all: it merges contacts. Dolibarr cannot: no method, no button, no extension point. DataCleaner relinks the attachments, deduplicates conflicting keys, and blocks the merge if a table it does not know about references the contact, rather than silently losing data.

The rest of the job

• Warning when creating a third party that looks like an existing one, with "open the existing one" or "create anyway". Optional blocking on an already used VAT number.
• Data health score across four axes and ten findings: duplicates, invalid emails, missing country or postal code, shared identifiers.
• Full audit trail of every merge: state before, plan, result. Written before the action, so it is usable even if something goes wrong.
• Configurable, automatically purged audit retention: those snapshots hold personal data (GDPR).
• The "Merge" permission is not granted by default, and additionally requires the Dolibarr delete permission for the object.

Measured performance

20,000 third parties: fingerprints in 15 s, scan in 47 s, 33 MB of memory. Blocking happens in the database, in batches: 178,520 pairs evaluated instead of 202 million comparisons.

No dependency

No external service, no AI, no data leaving your server. Everything runs locally, inside your Dolibarr. French and English.