Your CRM likely has a duplicate problem.

Native duplicate detection only works when values are identical​

Out of the box, Dynamics compares values character by character. "Robert Whitfield" and "Bob Whitfield" are two different people to it. So are "Acme Inc." and "Acme, Incorporated", and an email in emailaddress1 on one record and emailaddress2 on another (we call this cross-field matching).

You are also capped: five published duplicate detection rules per table, no matching on fields of related records, and no way to say "these two are probably the same, tell me how confident you are" (we call this precision).

The result is not that detection fails loudly. It is that it reports clean and your data is not.

Manual deduplication costs hours​

The native merge dialog handles one pair at a time. Working a real backlog of duplicates means opening a record, hunting for its twin, choosing a master, comparing fields side by side, and merging. Then repeating it a few thousand times.

AutoMerge can merge many duplicate sets of 2+ with a single click, not just one set of only 2 records as the OOB merge does, and AutoMerge preserves every field along the way: any value present on a losing record is copied onto the winner wherever the winner's field is blank. Nothing needs manual reconciliation before the merge runs.

The full field-by-field comparison against native merge →

Duplicates in your CRM affect downstream processes​

A duplicate contact is one row until it reaches everything downstream:

• Marketing emails the same person twice, or splits their engagement history across two records so neither one qualifies for the campaign.
• Reporting counts one account as two, inflating pipeline coverage and account counts that leadership is making decisions against.
• Forecasting double-counts opportunities attached to sibling accounts, or loses them behind the record nobody is looking at.
• Service opens a case against the record with none of the history.

Deduplication is not data hygiene for its own sake. It is the precondition for trusting anything your CRM tells you.

AutoMerge tags every set with a precision score, making bulk AutoMerging easier​

During an analysis, AutoMerge works through three steps in order: match records into candidate sets, rank each set's members to determine the winner-designate, and score each set's precision. Then it tags the results in your CRM.

From there the methodology is straightforward. Work through the tagged sets from the highest precision score downward. As you go, you'll find the point where valid duplicate sets give way to false positives: that's your bulk-merge threshold. Everything above it AutoMerges in bulk with minimal review; everything below it needs a closer look before you AutoMerge it by hand. Still quick and easy, but it takes more human validation.

There is no simple formula behind this, and that is deliberate: a single threshold applied to a single field is exactly what native detection already does badly. What matters for your work is that the score is comparable across sets, so "everything above this line, merge it" becomes a decision you can actually defend.

How precision scoring is configured →

What a match might look like​

Three contacts, all the same person, none of them identical. Native detection matches none of these pairs.

Three things happened there that native merge cannot do. The nickname and the misspelling were matched to the full name; the email that lived in emailaddress3 on one record was recognized as the same address in emailaddress1 on another; and every field with a value on a losing record was preserved onto the winner.

The ranking rules in your AutoMerge Management App analysis rules decide which record survives: oldest, newest, most complete, or a field of your choosing. That's the right call in the vast majority of cases. When it isn't, use AutoMerge Set Primary in your CRM to override the winner for that one set.

How field preservation works →

Matching duplicates with native Dynamics versus AutoMerge​

Matching​

See the full matching comparison →

Merging​

See the full merging comparison →