Anomalies and Insights

Anomalies and Insights

At the bottom of the Summary tab, MatchLogic automatically analyzes your match results and surfaces two types of findings: Anomalies that flag unusual patterns, and Insights that provide actionable recommendations for improving match quality.

The Anomalies and Insights section at the bottom of the Summary tab, showing anomaly cards with warning icons on the left and insight cards with lightbulb icons on the right

Anomalies

Anomalies highlight statistical outliers or unexpected patterns in your match results. These are not necessarily problems, but they deserve attention. Common anomalies include:

  • Disproportionate source matching — one datasource accounts for a significantly larger share of matches than expected based on its record count. This could indicate data quality issues in that source.
  • Score clustering — an unusually high concentration of pairs at a specific score, which might suggest a systematic data pattern or an overly narrow definition.
  • Large groups — one or more groups contain far more records than typical, potentially indicating a blocking or grouping issue.
  • Unmatched source — a datasource has very few or no matches, which may mean its fields are not well-mapped or its data format differs significantly from other sources.

Each anomaly card includes a brief description and the relevant metric so you can quickly assess its significance.

Insights

Insights are recommendations generated by MatchLogic based on the match results. They suggest specific actions you can take to improve your matching outcome:

  • "Consider increasing the threshold" — if many pairs fall in the low-score bands, raising the threshold would reduce false positives.
  • "Review low-scoring pairs" — suggests switching to the #pairs-view and filtering by the Low or Poor confidence bands to check for false positives.
  • "Add more criteria" — when the score distribution is heavily concentrated in a narrow range, adding more fields can improve discrimination between true and false matches.
  • "Check field mappings" — if a datasource underperforms, the mappings may need adjustment. See https://help.matchlogic.io/article/397-field-mapping-between-datasources.

Tip

Treat anomalies and insights as a starting checklist after each match run. Address the highest-priority items, re-run the match if you make changes to definitions, and then check the summary again to see if the findings have resolved.

When No Anomalies Appear

If the anomalies section is empty, your results are within normal statistical bounds. This is a good sign, though it does not guarantee perfect results. You should still review a sample of pairs manually, especially those in the moderate confidence bands, before proceeding to #what-is-merge-and-survivorship.