Understanding Match Score Thresholds

  • Updated

Match score ranges from 1 to 100 and indicates how likely a Demandbase match is correct.

Important:  A match score is not a percentage. For example, a score of 60 corresponds to roughly 90% accuracy, not 60%.

See Set Match Score Thresholds and Common Causes of Low Match Rates.

How Match Scores Are Calculated

Match scores depend on additive and subtractive factors:

  • Number of input fields: More fields improve accuracy. Fewer fields lower confidence.
  • Number of potential matches: More possible matches reduce confidence and lower scores.
  • Similarity across companies: Many similar records dilute accuracy and reduce scores.

Why Use Match Thresholds?

A match score threshold sets the minimum score required for records to pass into downstream systems.

Threshold selection reflects tolerance for false positives:

  • Lower threshold: more matches, higher false positives
  • Higher threshold: fewer matches, higher accuracy

How to Set Match Score Thresholds

 Demandbase groups results into the following categories:

  • Acceptable/Good: Meets or exceeds the threshold
  • Rejected: Below the threshold
  • Uncertain: Multiple possible matches, may require review

Best Practices for Setting Match Score Thresholds

Thresholds balance precision and recall:

  • Higher precision reduces record volume
  • Higher recall increases volume but lowers precision

Recommended starting points:

  • 0.7: Very high precision, decent recall
  • 0.5: High precision, high recall
  • 0.3: Moderate precision, very high recall

Match Score Benchmarks

This table shows the match score (also known as "confidence score") based on a sample size of 100,000 records:

Match Score  True Positive (TP) True Negative (TN) False Positive (FP) False Negative (FN) FP_FN Precision Recall F1 Score
0.3 62,182 13,273 15,909 2,727 5,909 74.03% 87.80% 80.33%
0.5 62,000 18,454 10,727 3,455 5,364 79.39% 87.55% 83.27%
0.7 54,000 28,091 1,091 14,818 2,000 94.59% 76.25% 84.43%
0.9 14,455 28,726 455 56,273 91 96.36% 20.41% 33.69%

Example

A Demandbase customer needed a match score threshold that balanced accuracy and enrichment. Demandbase provided three options, and the customer selected a match score of 60 to meet their target balance.

Accuracy versus Enrichment Balance Match Score Threshold Percentage of Actual Match Accuracy Percentage of False Positives
High Accuracy & Low Enrichment 70 94% 5%
Medium Accuracy & Medium Enrichment 60 90% 10%
Low Accuracy & High Enrichment 50 85% 15%

Was this article helpful?

0 out of 0 found this helpful