Multi-column deduplication finds duplicate records by comparing several properties in the same check.
Useful combinations include:
- First name + last name + company domain
- Company name + city + country
- Product name + SKU + supplier
- LinkedIn URL + company domain
Why compare several columns
One property is not always enough. Two contacts can share a name, and two companies can have similar names. Adding a stable identifier reduces false matches.
| First name | Last name | Company domain |
|---|---|---|
| John | Smith | example.com |
| Jon | Smith | example.com |
The first names differ, but the last name and domain agree. Whether a check returns this pair depends on its algorithms and threshold. Smart text normalization does not automatically turn Jon into John; use a distance comparison when you want to admit spelling differences.
📌 Tradeoff
More properties can reduce false matches, but they can also hide real duplicates when one property is empty or inconsistent.
Datablist workflow
Select several properties and configure each comparison independently. For example:
- Exact on company domain
- Distance on first name when spelling variations should be considered
- Smart on last name
- URL processor on LinkedIn profile URL
All selected properties contribute equally to the final similarity score. This is an average of the per-property comparison scores, rather than a requirement that every column match exactly.
For example, two exact matches score 1 each. A third populated text field that does not match scores 0. Their average is about 67 percent, so that pair does not meet an 80 percent overall threshold. A field passing its distance threshold can contribute 0.90 instead, giving about 97 percent with the two exact matches.
Averaging also means agreement on several weak fields can outweigh disagreement on another field. If a domain must agree, constrain candidates to the same domain before relying on a broader comparison. Test known duplicate and non-duplicate records before bulk processing. Check the empty-value rules separately: a missing identifier is not evidence of a match.
After the check, use the Duplicates Finder to review groups, merge Ready groups, or resolve conflicts manually.
For separate files, read cross-list deduplication. For several emails or URLs inside one cell, read multi-value deduplication.
Follow the list deduplication guide for a tested workflow and the data matching guide for algorithm selection.