Data Quality
Duplicate Grouping
Groups rows that share their match keys — exactly, or fuzzily — and keeps one survivor per group: the most complete row, the longest, or the first.
When to use it
Use it to merge duplicate records such as three accounts sharing an EIN, when Deduplicate by Key's exact match misses "12-3456789" vs "123456789" or you want the best-filled row kept rather than the first.
In a pipeline
File InputStandardizeDuplicate GroupingExport CSV
Settings
| Setting | Type | Required | Default or hint |
|---|---|---|---|
| Match keys | select | Yes | Pick at least one column |
| Match | select | — | exact |
| Similarity (0–1) | number | — | 0.85 |
| Keep | select | — | most-complete |
| Max comparisons | number | — | 5000000 |
Example
Three accounts, one EIN
- Match keys=EIN
- Match=Exact (ignoring case, spaces and punctuation)
- Keep=Most complete row (fewest blanks)
| Account | EIN | Phone |
|---|---|---|
| Acme Inc | 12-3456789 | — |
| ACME Incorporated | 12-3456789 | 555-0100 |
| Acme | 123456789 | — |
| Beta LLC | 98-7654321 | 555-0199 |
| Account | EIN | Phone |
|---|---|---|
| ACME Incorporated | 12-3456789 | 555-0100 |
| Beta LLC | 98-7654321 | 555-0199 |
Worth knowing
- Each survivor carries its group number in the hidden __group_id column; the note says how many groups were merged and how many rows were removed.
- Pick Levenshtein, Jaro-Winkler or Soundex under Match to group near-identical keys; the same Max comparisons safety limit as Fuzzy Match applies.
- Rows whose keys are all blank are never grouped with each other.
- Ties between equally complete rows go to the earliest one.