Fingerprints

Findings tell you what was found. Fingerprints tell you where the same thing shows up again — the customer whose email appears in three systems, the record duplicated across two databases, the entity that connects a dozen findings. It’s how Classifyre turns a flat list of findings into a map of duplicates and similar items.

Fingerprints sit between detection and investigation:

You’ll find this in the app under Fingerprints, where a similarity graph shows how your assets connect.

How a fingerprint is built

A fingerprint is built from the values inside your findings — emails, names, phone numbers, account IDs, addresses, and so on. The approach is deliberately deterministic and explainable: there’s no black-box embedding. Two assets are linked because they demonstrably share the same values, and you can always see which ones.

1. Extract values from each asset’s findings and normalise them (so [email protected] and [email protected] match).
2. Build a signature for the asset — a compact summary of its values.
3. Compare signatures across every asset and score how much they overlap.

Two extra touches make matching smarter:

• Exact-duplicate bucketing — assets whose values are all identical are caught immediately as duplicates.
• Phonetic matching for name-like values — John Smith and Jon Smyth are recognised as likely the same person, not two different ones.

Similarity, related, and duplicate

Every link between two assets carries a similarity score from 0 to 1 — the share of important values they have in common. Two thresholds turn that score into plain language:

In the Fingerprints graph, a similarity slider lets you raise or lower the bar: drag it up to see only strong duplicates, down to explore looser relationships.

Clusters: the same entity, everywhere

When several assets are all linked by strong evidence, Classifyre groups them into a cluster — a set of assets treated as one entity, even when they live in different sources. A cluster might be “all the records about customer Jane Doe” scattered across a CRM, a support inbox, and a data warehouse.

Each cluster shows how many assets it spans, how many sources it touches, and the values its members have in common — the shared evidence that binds it together. Every asset belongs to at most one cluster, so the picture stays clean.

From cluster to case in one step. When a cluster looks worth investigating, you can create a case straight from it — its members come in as evidence, ready to work. This is one of the fastest ways to start an investigation. See Cases.

Tuning what counts (advanced)

Fingerprinting has sensible defaults, but you can shape it to your domain:

Changing any of these recomputes the fingerprints across your data so the graph and clusters reflect the new rules.

Where Fingerprints fit

• Inquiries watch findings by rule; fingerprints connect them by shared identity. They’re complementary lenses on the same findings.
• Cases are where a cluster becomes a real investigation, with evidence and hypotheses.
• Autopilot keeps fingerprints up to date automatically: a deterministic correlation step refreshes duplicates and clusters after each scan — before the AI agents reason over them — and the agents can open a case from a notable cluster on their own.

In short

Next: turn a cluster or a set of findings into a real investigation in Cases.