Relait
See how your data is connected – without it leaving your infrastructure.
Relait discovers how data relates across your organisation's databases, finding the joins, lineage, and shared identifiers that traditional cataloguing misses – all of it inside your own infrastructure.
Patent-pending deterministic matching, from DataJPS.
Where else does this data live?
Most organisations cannot answer that simple question – or how the copies are connected. Relait answers it automatically, turning sprawling, undocumented databases into a navigable relationship graph.
What you get
Evidence, not guesswork
Relationships come from what the data actually contains, not names, comments or documentation that may be wrong or missing.
A confidence score on every relationship
Filtered at query time, nothing pre-filtered away – the same scan answers a high-bar migration question and a low-bar sensitive-data hunt.
Direction matters
Two columns can overlap without being equivalent; one may sit wholly inside the other, or both may be partial views of the same population. Relait tells those cases apart.
An explorable relationship graph
How every column connects across your databases, with history on every edge – Relait keeps score over time.
How it works
Relait starts from the data itself, not the column names and documentation that describe it – which is how it finds the relationships nobody wrote down.
Runs in your infrastructure
Relait is deployed inside your own environment and profiles your databases in place. Nothing is sent to DataJPS or any third party.
Fingerprint every column
Each column is reduced to a fingerprint of what it contains, and the relationships are established from those alone. Your documentation is then laid over the result – to corroborate it, or to show where it has drifted.
Weighs the evidence
Relait works out which columns genuinely hold the same information, how strongly, and in which direction.
Explore & decide
Query joins, lineage, and identifiers at whatever confidence level suits the question you are asking.
Privacy is the architecture, not a feature
Relait runs entirely inside your own infrastructure – nothing is sent to DataJPS or any third party. It works on column-level aggregates and never joins back to individual source records. Relationships between columns become visible; the records behind them are never reconstructed.
Kindred – when close counts
Real-world spellings drift – Jon, John, Jonn in one column; Smith, Smyth in another. Exact matching treats every variant as a different value, so a column full of drifted spellings looks less related than it really is.
Kindred works the way everything in Relait works: at the column level. It compares what columns contain phonetically, so column pairs whose values agree in sound stand out even when the spellings disagree. It never compares individual people or records – first names in one column and surnames in another stay exactly where they are.
What you get is candidates: pairs of columns worth analysing when you build identity-resolution rules – which columns should take part, and how much phonetic drift they carry.
The comparison draws on established phonetic algorithm families, among them Soundex, Double Metaphone and Eudex.
Kindred's candidates are kept separate from Relait's deterministic matches and are included per analysis at a threshold you choose, so the deterministic graph stays pristine.
Kindred ships today as an optional add-on to Relait.
- Identity-resolution rule analysis – deciding which columns should take part in a matching rule, and how much phonetic drift to tolerate.
- Master data reconciliation planning – prioritising which record sources are worth reconciling first.
- Data quality auditing – columns whose values look like phonetic variants of each other are a leading indicator of upstream quality issues.
Where Relait earns its keep
- Migration and system replacement – discover cross-system relationships before migration begins; validate assumed join keys, and find alternatives where links are weak.
- Governance and sensitive-data discovery – surface potential undeclared copies of sensitive data; trace which columns share values with sensitive columns for blast-radius analysis.
- AI readiness – structured, machine-readable relationship context for AI initiatives.
- Validating documentation against reality – audit schema documentation against what the data actually shows.
Relait's outputs feed catalogues and governance platforms such as Atlan, Collibra, Alation and Purview – it discovers what those platforms then manage.
What it changes
Based on DataJPS analysis of typical data projects, Relait cuts elapsed delivery time by roughly a third – worth approximately AUD 220,000 on a typical project. These are modelled figures, an expected order of magnitude for a typical project rather than a promise.
In one engagement, a national children's health organisation mapped an undocumented multi-system estate spanning close to two decades of program data – with clinical data never leaving their on-premises environment.
See it run on a whole estate
Start with one table you already recognise. Then every table in four systems, and the relationships nobody wrote down – the 79-second film and the full technical detail live at relait.io.