TL;DR
- A higher duplicate count isn't a better result. It often means a looser search with more false matches.
- Loosening every field at once, matching on one weak field or skipping filters all inflate the count.
- The useful number is how many matches are real, which only a checked sample can show.
- Compare tools and searches by accuracy on the same records, not by the size of the number.
- In Dedupely, each search shows the fields and match options it uses, so you can see where a count comes from.
The first time a team runs a duplicate search, the number gets everyone's attention. Then someone loosens a setting and the number doubles. Then another tool shows a bigger one. It's tempting to treat the biggest number as the most thorough result.
It usually isn't. A duplicate count is only as good as the matches inside it. This post covers why counts grow, when a bigger count is a warning sign and what to measure instead.
Why the count goes up
Duplicate counts grow for good reasons and bad ones.
Good reasons: a search starts catching real duplicates it missed, like typos, nicknames or phone numbers in different formats.
Bad reasons:
- Every field is loosened at once. Similar or Fuzzy matching on the name, the company and the city together pairs up people who only look alike.
- The search relies on one weak field. A first name, a city or a generic company name matches thousands of unrelated records.
- Placeholder values match each other. "N/A", "test@test.com" or "Unknown" turn into huge groups.
- No filters. Old test records, internal users and partners all end up in the results.
From the outside, both kinds of growth look the same: a bigger number.
When a bigger count is a warning sign
Look closer when:
- The count jumps after one setting changes, much more than you expected.
- Groups get very large. A set with many records in it is rarely one person or one company.
- A sample is mostly wrong. If more than two or three in a sample of 20 aren't duplicates, the search is too loose for bulk merging.
- The same few values show up everywhere, like a shared inbox or a default phone number.
What to measure instead
The useful questions are:
| Instead of | Ask |
|---|---|
| How many duplicates did it find? | How many of a sample of 20 are real? |
| Which tool finds the most? | Which one is right more often on the same records? |
| How big can we get the number? | How many can we merge without checking each one? |
| What's the total? | What's the total for each object and each search? |
A search that finds fewer duplicates but is right almost every time is the one you can merge in bulk. A big, noisy search is better used for manual review, a few sets at a time.
How Dedupely shows you where the count comes from
In Dedupely, each Duplicate Search shows the fields and match options it uses, so you can see why a count is high: Exact, Similar, Fuzzy, Similar Words or Domain Root, field by field. Filters narrow which records are searched, ignored terms leave out placeholder words, and two blank values never match. Safety Settings hold back sets with more than 20 records and placeholder data from bulk merges, and you can reject sets that aren't duplicates so they stay out of future results.
The right number is the one you trust
A bigger duplicate count feels like progress, but it often means a looser search. Check a sample, compare searches on the same records, and use the count you can merge with confidence. The rest belongs in manual review.
Frequently asked questions
Why did my duplicate count double after changing one setting?
Usually because a field was loosened, for example from Exact to Similar or Fuzzy. Check a sample of the new matches to see whether they're real.
Is a tool that finds more duplicates better?
Not by that alone. Compare how many of its matches are real on the same set of records.
How do I know if a search is too loose?
Check a sample of 20 sets. If more than two or three aren't duplicates, or groups are very large, tighten the search before bulk merging.
What inflates a duplicate count?
Loose matching on several fields, matching on one weak field, placeholder values and missing filters.
What should I report instead of the total count?
The count per object and per search, and how many of a checked sample were real duplicates.



