Same Entity.
Proven Match.
Deduplicate and link customer, supplier and product records across systems, where they live. Every merge names the rule that fired and the exact records it used, so match quality is something you check, not a score you trust.
Proven in production
Black-Box Scores.
Or Rules You Read.
When the same customer appears five ways in five systems, every report, model and agent inherits the mess. ML scores, vector similarity and LLM judgements hide the decision. Prometheux makes it explicit logic.
Rules That Say
Why Two Match
Write matching criteria as rules over your sources: shared tax IDs, emails, addresses or any combination. Prometheux joins across systems, links records that match, and groups them into resolved entities.
Cross-System Joins
Compare CRM, ERP, mainframe files and spreadsheets in one query, without staging them in a hub.
Transitive Clusters
Recursion links chains of matches so every record of the same entity ends up together.
Evidence for Every Merge
The chase graph records which rule linked two records and which facts it used.
AI for the Hard Cases
Route ambiguous pairs to an LLM concept that returns a typed decision, logged in lineage like any other step.
Trust the customer count in every report. Defend each merge to auditors and data owners. Change a matching rule once and re-run everywhere.
From Duplicates to
Resolved Entities
Connect the Systems
Connect each source that holds the entity: CRM, billing, ERP, files. Data stays in place.
Write Match Rules
Encode when two records are the same, from strict ID matches to combinations of attributes.
Resolve and Group
Recursive rules turn pairwise matches into one resolved entity per real-world thing.
Review and Publish
Inspect the evidence for each match, then write the resolved view back or serve it via API.
One Real-World
Thing, One Record
Customer and Counterparty
Link clients across core banking, lending and onboarding systems before computing exposures.
HCPs and Organisations
Resolve healthcare professionals and institutions across CRM, events and third-party lists.
Supplier Master
Find the same supplier under different names and IDs across ERPs and spend files.
Cross-System Consistency
Check that an identifier in one system matches downstream systems, and explain every mismatch.
“The ability to query heterogeneous databases through a flexible ontology makes it possible to infer deep knowledge and automate complex tasks, delivering trustworthy insights that far exceed the capabilities of traditional systems.”
Anna FavaFormer Senior Manager, Data & AI · Dompé
Questions,
Answered
What is entity resolution?
It is deciding when records in different systems refer to the same real-world customer, supplier, product or organisation, and grouping them into one resolved entity. It is also called record linkage, deduplication or match/merge.
Do I have to load all sources into one place first?
No. Prometheux joins across your databases, warehouses and files where they are, and can write the resolved view back to the system you choose.
How do I know why two records were merged?
Every match is derived by a rule, and the chase graph records which rule fired and which records it used. You can trace any resolved entity back to its evidence.
How accurate is the matching?
Matches follow the rules you approve, and every one can be checked: each merge shows the rule and records behind it. When a data owner spots a false match, you fix the rule that caused it and re-run, instead of retuning a score.
How is this different from master data management?
MDM tools usually copy records into a central hub and curate a golden record there. Prometheux resolves entities over your sources in place, keeps the match logic as readable rules, and writes the resolved view, your golden record, back to whichever system you choose.
Can AI help with ambiguous matches?
Yes. An LLM concept can review uncertain pairs and return a structured decision, while clear-cut matches stay deterministic. Both appear in lineage.
Related Capabilities
Resolve Entities
You Can Defend
Bring two systems that disagree about your customers. We'll match them with rules you can read.
