Perferens Identity · Rule-based person matching Spelled differently.
How similar
is the data?
Our library compares person and address data and assesses possible matches against business-defined rules. The result: a match score as the basis for your further checks and decisions.
Reusable, functionally tested and in production for datasets in the millions.
Business rules. Differentiated assessment.
Not every field is the same.
Your matching should reflect that.
Names, place details and numbers follow different rules. Our library combines targeted preparation of your data with specialised comparison methods and a field-by-field assessment. The result is a match score that reflects your business requirements.
Prepare the data for comparison
A purpose-built processing pipeline is a central part of the comparison: it normalises the input, removes stop words, resolves abbreviations and breaks the data down into comparable parts with analyzers and tokenizers. Matching starts before the actual similarity assessment.
Combine suitable methods
Jaro-Winkler and Levenshtein assess similarities in spelling. Double Metaphone and Beider-Morse add phonetic methods. Business-defined rules determine how names, place details and numbers are compared.
Assess each field specifically
Comparison rules and assessment are defined for every field. Following the business rule set, the individual results feed into a match score – as the basis for further checks and decisions.
Built for enterprise use.
Reusable, functionally tested and developed with performance in mind. The library is in production for datasets in the millions.
< 10 msto compare two persons
Fields of use
Where person matching
makes the difference.
Wherever the same person can appear in different records, reliable matching decides data quality, legal certainty and costs.
Duplicates and fraud detection
Persons created more than once distort analyses and open the door to fraud patterns. Matching finds records that denote the same person despite different spellings – as the basis for clean-up and for reviewing suspicious cases.
Migration into a central register
When datasets are moved into a central register, the same person must be created only once. Matching assigns incoming records to persons already known and makes every assignment traceable.
One-time benefits at banks and insurers
New-customer offers such as an instant-access savings account with a higher interest rate should be used only once per person. Matching recognises whether a new application comes from a person already known – even with differing details.
Achieving more together
Let's talk about your project.
Would you like to develop software, modernise existing systems or settle a technical question? Together we discuss your goals and how Perferens can support you.
Further features
Florenz or Firenze?
One place, many names.
Data from different countries and systems spell the same place differently. The library takes that into account – across language boundaries.
Compare place names across languages
Place names come in their own form per language: Florenz in German, Firenze in Italian, Florence in English. The library maps such variants to the same place, so the comparison recognises the match – even when the records come from different countries.
- FlorenzDEFirenzeITFlorenceENone place
- KölnDECologneENColoniaITone place
- WienDEViennaENVienneFRone place
Resolve abbreviations and stop words
The processing pipeline brings the input into a common form before the comparison: it normalises spellings, resolves common abbreviations and removes stop words. Variants that mean the same value meet – instead of failing on formalities.
- Frankfurt a. M.Frankfurt am Mainsame value
- Hauptstr. 12Hauptstraße 12same value
- Mara von LindénMara Lindensame value