Understanding Match Score Thresholds

A match score is a number from 1 to100 that indicates the likelihood that a match obtained by Demandbase is accurate. Match scores provide insight into the quality of a matched record.

The match score is also by used Data Integrity, Enrich, Refresh, and Diagnose products to return a quality matched record based on your input values.

Note: Match scores do not exactly correspond with raw probability; that is, a match score of 60 does not mean that a match is 60% likely to be accurate. In fact, a score of 60 corresponds with accuracy rates of about 90%.

How are Match Scores calculated?

Match scores are composed of additive and subtractive components. These additives (adding to the score) components affect the match score thresholds based on:

• The number of data fields provided as input by you—lesser the data fields lesser the match score is because the accuracy of data depends on the number of fields for which Demandbase has the information that can be matched.
• The number of potential matches in the pipeline—higher the number of potential matches in the pipeline lower the match score is because if the number of alternatives is many, then matching and choosing from multiple alternatives is a difficult choice.

Any company with a lot of potential matches in the database would be subtracted from the score proportionately to the number of companies that could potentially dilute the match results.

Why you should use Match Thresholds?

When configuring Enrich, account administrators may set a match score threshold—i.e. a minimum match score below which enriched leads will not be sent into the marketing automation system. The appropriate match threshold for your use case will depend on your business requirements.

Because match scores are correlated to levels of data accuracy, the choice of where to set the match score threshold is, in essence, a question of how many false positives can be tolerated in the lead supply.

Generally speaking, a lower threshold will conduce to a greater number of leads with a higher proportion of false positives, whereas a higher threshold will yield better accuracy but at a lower volume.

How to Set Match Score Thresholds

Data Integrity, Enrich, and Refresh provide an easy mechanism for system administrators to set match score thresholds, which gives you control over the quality of data you receive from Demandbase for updating and enriching your records. We use these thresholds to assign matches into these categories:

Best Practices for Setting Match Score Thresholds

When setting match score thresholds, it’s important to note that precision and recall work in an inverse relationship. If you are focused on precision, fewer records will be returned, meaning recall will be reduced. If you’re focused on a high match rate (high recall), precision is likely to be sacrificed. 

Here are guidelines to help you decide what fits your need:

• Very high precision with decent recall: Set the confidence score to 0.7
• High precision with high recall: Set the confidence score to 0.5
• Decent precision with very high recall: Set Confidence score to 0.3

Match Score  Benchmarks

This table shows the overall match score (aka confidence score) for various data matching results, based on a sample size of 100,000 records:

Use Case Example

An Demandbase customer from the tech industry wanted to fine-tune Enrich with a match score, which would enable them to have the right balance of accuracy versus enrichment.

Demandbase's approach was to provide this customer with three options. The following table illustrates these options, showing levels of accuracy, false positives, and enrichment for match score thresholds of 70, 60, and 50 respectively: