Avoiding False Matches

A false match is a group of records that look like duplicates but describe different people or companies: two colleagues who share info@acme.com, or every contact with the phone number 000-000-0000. In Dedupely, you avoid false matches by matching on at least three fields, choosing the right match option for each one, using filters and Safety Settings to keep weak data out, and rejecting the false matches that still appear. Treat every merge as permanent, so get this right before you bulk merge or turn on Auto Merge.

What do false matches usually look like?

How many fields should a search match on?

At least three. One or two fields create a lot of false matches, because two people can share a name, and one company can share an email domain. For example:

• Too loose: Last Name (Exact) + Email (Domain Root). Catches every colleague with the same surname.
• Better: First Name (Exact) + Last Name (Exact) + Email (Exact).

If a search has to match on fewer fields, add more Visible Fields, so you can compare other details (like Phone, Company or Created Date) before you merge. See Duplicate Searches.

Which match option is too loose for which field?

• Fuzzy matches names that sound alike ("Sean" and "Shawn"). Use it on names, not on emails or IDs.
• Similar catches typos. On phone numbers, it compares the core seven digits, so pair it with another field.
• Domain Root groups everyone at the same domain. Use it for companies, not to match individual people on its own.
• Similar Words (Any Order) can match unrelated names that share common words. Add those words to Ignored terms.
• Exact is the safest for emails, IDs and anything you can't afford to get wrong.

See Match Options.

How do Safety Settings help?

Safety Settings set risky matches aside in the Excluded tab, so they're left out of bulk selection and Auto Merge:

• Records per match (20 by default): a very large match usually means the search is too broad.
• Poor or example data (on by default): placeholder and test data.
• Very short values (off by default): turn it on if your data has initials or single characters.

Changes apply after your next scan. See Merge Safety.

How to filter out generic data

Use a filter to leave out values you know are shared. For example, to skip shared inboxes:

• Field: Email · Applies to: all · Operator: does not start with · Terms: info@, sales@, support@, hello@, admin@ · All

You can add up to 25 terms per filter. With a negative operator like this one, set the term setting next to the operator to All, so a record has to avoid every term. See Filters.

What to do with the false matches that still appear

• Reject matches you're sure aren't duplicates. They leave every search for that object, including Auto Match and Auto Merge. Rejected matches are saved in memory, so they never show up again.
• Send to Manual Review the matches you're not sure about, so the right person can decide.
• Fix the search if you keep rejecting the same kind of match. A pattern of false matches means a field or match option needs to change.

Example setup

A team keeps finding colleagues merged into one contact, because many leads come in through shared inboxes.

1. Search: First Name: Exact · Last Name: Exact · Email: Exact, instead of Last Name + Email domain.
2. Filter: Email does not start with info@, sales@, support@, office@ (term setting All), Applies to all.
3. Safety Settings: Records per match 10, Poor or example data on, Very short values on.
4. Visible Fields: Phone Number, Company and Create Date, to compare before merging.
5. Review: reject false matches as they appear, and send unclear ones to Manual Review.
6. Automate only after a few scans with no new false matches: Auto Match first, then Auto Merge.

Common mistakes

• Matching on one or two fields. It's the most common cause of false matches.
• Expecting blank fields to match. Two blank values never match, so they can't cause a false match, but records with blanks in a matched field won't show up in that search.
• Using Fuzzy on emails, or Domain Root alone to match people. Fuzzy is built for names. Domain Root on email groups everyone at the same company, so pair it with name fields.
• Turning off Poor or example data. Placeholder values then group thousands of unrelated records.
• Merging large matches without opening them. Check the Excluded tab and open matches with many records first.
• Rejecting matches one by one without fixing the search. If the same kind of false match keeps appearing, change the search.

FAQs about false matches

• What's a false positive in deduplication? A match of records that look like duplicates but aren't, like two people who share an email inbox.
• Can I stop a false match from coming back? Yes. Reject it, and it stays out of every search for that object.
• Can I restore a rejected match? Yes. Open the Rejected Matches tab, select it and click Remove from Rejected.
• Does Auto Merge merge excluded or rejected matches? No. Both are skipped.
• What if a false match was already merged? Undo it from Merge History within 30 days. See Undo a Merge. For older merges, see I Merged the Wrong Records: What Now?