Snapshot Analysis

sz_snapshot takes a point-in-time snapshot of the entity resolution results and generates summary reports. These reports answer high-level questions about the data: how many entities exist, how records are distributed across data sources, and where cross-source matches occur.

Viewing snapshot results in sz_explorer

To view the snapshot reports interactively, load the snapshot file when starting sz_explorer:

sz_explorer -s truthset_snapshot.json

Or load it after sz_explorer is already running:

load truthset_snapshot.json

This unlocks the snapshot report commands: data_source_summary, cross_source_summary, entity_source_summary, entity_size_breakdown, and principles_used. Each command displays a table with a prompt to drill into specific rows for more detail.

data_source_summary

The data_source_summary command shows how records from each data source resolved into entities.

An entity count significantly lower than the record count indicates many records belong to the same real-world person or organization. The Compression column shows the percentage of records that were duplicates within each data source.

Selecting a data source drills into its match level breakdown:

This shows how the matched CUSTOMERS records break down by match level. Most matches are high-confidence Matches, with smaller counts for Possible Matches and Possible Relationships.

Selecting a match level drills into the match keys that fired:

Each match key shows which combination of attributes caused the match. For example, +NAME+DOB means a name and date of birth matched, while +NAME+ADDRESS means a name and address matched.

Selecting a match key drills into individual entities:

The entity detail view shows all records resolved to this entity, including the specific attribute values that drove the match.

cross_source_summary

The cross_source_summary command shows matches between records from different data sources, revealing connections that span data silos.

Cross-source matches drive compliance screening, fraud detection, and risk assessment workflows. For example, CUSTOMERS-to-WATCHLIST matches represent customer records that resolved to the same entity as a known risk entry.

Selecting a data source pair drills into its match level breakdown:

The CUSTOMERS-to-WATCHLIST pair shows the count of matches at each confidence level. High-confidence Matches indicate records that Senzing is confident belong to the same real-world entity across these two data sources.

Selecting a match level drills into the match keys:

The match keys show which attributes drove each cross-source match and what types of identifying information connect records across data sources.

Selecting a match key drills into individual entities:

This entity resolved records from both CUSTOMERS and WATCHLIST, meaning the same real-world person appeared in both data sources.

entity_source_summary

The entity_source_summary command groups entities by which combination of data sources contributed records.

Selecting a row drills into the entities for that source combination, showing a paginated entity list with full entity detail:

entity_size_breakdown

The entity_size_breakdown command shows the distribution of how many records make up each entity.

The Entity Count column shows how many entities exist at each size. Entities in Size Group 1 are “singletons,” records that did not match any other record in the system. Large entities (those with many records) should be reviewed. They may represent:

• Legitimate matches: A long-time customer with records across multiple systems and name/address changes over the years.
• Over-resolution: Records that Senzing matched but that belong to different people. Use the how command in sz_explorer to review how records joined an entity step by step.
• Data quality issues: Duplicate submissions, test records, or data entry errors inflating entity size.

The Review Count and Review Features columns flag entities that contain more of a specific attribute than expected. For example, 5 entities in Size Group 2 have conflicting GENDER values, 1 entity in Size Group 4 has conflicting DOB values, and 1 entity in Size Group 5 has conflicting ADDRESS values.

Entities flagged in the Review Features column typically indicate one of three situations:

• Data quality problems: Typos, misspellings, or bad data causing attribute conflicts within a legitimately resolved entity.
• Intentional obfuscation: Someone altered identifying information to avoid detection, resulting in conflicting attributes.
• Overmatching: Records that belong to separate people were incorrectly resolved into the same entity.

Selecting a row drills into the entities of that size, showing a paginated entity list with full entity detail:

From the entity detail view, press H to run how and see the step-by-step resolution path:

The how decision tree shows how each record entered the entity, including the MATCH_KEY and principle that fired at each step. This is useful for large or flagged entities to see which attributes drove the resolution. For a full walkthrough of the how command and its different views, see Using how .

principles_used

The principles_used report shows how many entity relationships were established at each match level. Senzing performs Principle Based Entity Resolution where each match level represents a different level of confidence.

Selecting a match level drills into the specific principles, then into match keys, and finally into individual entity detail:

Using snapshot analysis effectively

Snapshot reports are most valuable when compared over time or against known baselines:

• Initial assessment: Run a snapshot after loading data to establish a baseline.
• After changes: After updating data mapping or adjusting entity resolution configuration with Senzing Support , take a new snapshot to measure the impact.
• Ongoing monitoring: Periodic snapshots track data quality trends as new records are added and existing records are updated or deleted.

Next steps