Match
The Match module finds duplicates within systems and links records across systems to build a complete view of each entity.
Why Matching Matters
The same people and accounts live in several systems. Without matching, records stay fragmented, duplicates cause confusion, and analytics miscount unique individuals.
What You Can Do
Choose fields to compare and the weight each carries.
Use exact or fuzzy matching to suit your data quality.
Compare potential matches side by side before accepting.
Set thresholds for auto-match, manual review, and non-match.
Match Types
Each attribute is compared with one of nine match types, so you can be strict on identifiers and forgiving on names. Identifier labels without a match type are matched exactly.
String equality for IDs, SSNs, and emails
Tolerates minor typos and transpositions
Tolerates stronger variation
Broad fuzziness for badly degraded data
Dictionary of nicknames, so "Bob" matches "Robert"
Tolerates year flips, transpositions, and month/day swaps
Similarity weighted toward the start of a value, good for short names
Phonetic matching, so "Smith" equals "Smyth"
Edit-distance similarity between two values
Two Ways to Configure Matching
Pick labels, a match type and priority for each, and one confidence threshold applied across every pair of tables.
Build multi-step strategies for Member 360, Patient 360, and Resident 360, such as an exact ID match, then a fuzzy name match on what remains.
Match Attributes
Nicknames, misspellings, and name changes
Formatting differences and abbreviations
Exact matching on unique identifiers
Different formats and entry errors
Normalized contact information
Any field in your data
Key Capabilities
Match Scoring
Each pair gets a 0–100 percent confidence score, a weighted average showing which attributes agreed and how strongly.
Null-Safe Scoring
Attributes missing on either side are left out of the average, so incomplete records are not penalized.
The Review Band
Pairs above the auto-merge threshold merge automatically, pairs in an optional review band below it go to the Review Queue for a data steward, and the rest are discarded.
Suffix-Aware Matching
Records with conflicting generational suffixes, such as Jr. and Sr., are never merged automatically, so relatives who share a name and address stay separate.
Configuration Checks
Saving a configuration flags errors, such as duplicate step names, before the job runs.
Transitive Matching
If A matches B and B matches C, all three are recognized as possibly the same entity.
Block and Compare
Records are blocked into candidate groups first, then compared in detail only within blocks.
Persistent Cluster Identity
Clusters and their golden records keep the same identifier across re-runs; a full-rebuild mode is available when you want to start clean.
Progress, Cancel, and Recovery
Jobs report progress, can be cancelled, and keep a run history. Stalled runs are healed automatically.
Runs Where Your Data Lives
Matching runs natively on Databricks, Snowflake, or Microsoft Fabric with the same results everywhere. On Snowflake, numeric IDs match exactly, oversized blocking fails fast, and Cancel run also cancels the query.
Match Audit Trail
Every decision is logged with score, contributing fields, and timestamp.
Business Impact
See each individual or entity across systems.
Know how many unique individuals you serve.
Give teams complete information.
Who Benefits
- Operations Teams: Complete records for coordination
- Quality Teams: Accurate attribution for quality measures
- Finance Teams: Proper counting for revenue analysis
- Compliance Teams: Accurate regulatory records