Overview
Deduplication is now powered by Bolt Analysis, Element451's AI engine for contact data quality. Instead of showing you a match score and leaving the interpretation to you, Bolt Analysis evaluates each potential duplicate pair, weighs the evidence on both records, reaches a conclusion, and explains that conclusion in plain language.
This article covers when Bolt Analysis runs, what it produces for each flagged pair, and how to use confidence levels to prioritize your review work.
When Bolt Analysis Runs
Bolt Analysis works in the background automatically. There is nothing to configure or trigger.
Change detection: When a contact record is created or updated, Element451 watches key identity fields, such as name, email address, phone number, and date of birth.
Candidate matching: A fast first-pass comparison checks whether the record now looks similar to any other record in your database.
Deep analysis: When the first pass finds a likely match, Bolt Analysis examines the full pair in depth and saves its findings.
Pairs flagged by Bolt Analysis appear in the review queue under Contacts > Deduplication, ready for a person on your team to review.
What Bolt Analysis Tells You
Every analyzed pair includes the following, visible from the review queue and the merge view:
Confidence level: A color-coded rating (High, Medium, or Low) showing how certain Bolt is that the two records belong to the same person.
Reasoning: A plain-language explanation of why the pair was flagged and how Bolt reached its conclusion. No raw scores to decode.
Key signals: The specific data points that drove the conclusion, such as an exact email match or a matching name plus date of birth.
Concerns: Evidence that argues against the match, so you can see both sides before deciding.
Recommended record: Which of the two records Bolt suggests keeping as the primary record if you merge.
Working with Confidence Levels
High confidence: The records agree strongly across identifying fields. These are usually quick confirmations. Skim the reasoning, confirm, and merge.
Medium confidence: There is real evidence for the match, but something needs human judgment, like conflicting addresses or partial identity overlap. Read the reasoning and concerns closely before acting.
Low confidence: The records share some surface similarity, but the evidence is weak. Many of these are not true duplicates and can be ignored after a quick look.
You Always Make the Final Call
Bolt Analysis identifies, explains, and recommends. It does not merge records on its own. Every merge in Bolt Deduplication is reviewed and confirmed by a member of your team, and ignored suggestions remain available if you want to revisit them later.
❗If you use Bulk Merge, and decide to to "Confirm and Merge" without reviewing the selection first, you will not have the ability to make the final call and are relying on the AI analysis to do the merge for you. Please be aware that AI can make mistakes, and therefore it is highly recommended that you review the selections manually to prevent false merges. Confidence level can help you make that decision, but you take responsibility for any issues that may occur.
How This Differs from Legacy Matching
The legacy Deduplication module compares individual fields and produces a weighted match percentage. You can read about that approach in Deduplication Logic.
Bolt Analysis starts from the same identity signals but adds a reasoning layer on top.
Rather than a percentage alone, you get an explained conclusion, the evidence behind it, and the concerns against it. Both experiences remain available, and you can switch between them at any time from the Deduplication page header.