Fuzzy Matching

Learn how to use fuzzy matching to link audit samples when identifiers are missing or inconsistent.

Use fuzzy matching to link documents when exact identifiers like invoice numbers are missing or inconsistent.

Fuzzy matching uses AI to identify similar values that are not exact matches, like vendor name variations or abbreviations.

When to use fuzzy matching

Missing identifiers

Bank statements lack invoice numbers but include vendor names and amounts.

Name variations

Vendors appear differently across systems ("Microsoft Corp" vs "Microsoft Azure").

Abbreviations

Documents use shortened forms ("ProTrain Solutions" vs "P-Train Solutions").

Name reversals

Payroll lists show "Last, First" in one system and "First Last" in another.

Using a combination of fields reduces false matches. Match on vendor name AND amount, not just one non-unique value.

Set up fuzzy matching

Select the match field

Open match settings for your document group
Choose the field for secondary matching (e.g., Vendor Identification)
Verify the field contains text that might have variations

Enable fuzzy logic

Click the three dots on the right side of the match column
Select Variance from the menu
Toggle Fuzzy Matching on

This deploys AI to identify similar values rather than requiring exact string matches.

Add custom instructions

Provide context to help the AI understand potential variations:

Example instructions:

"Look for abbreviations like Inc., Corp., LLC, Ltd."
"Consider name reversals (first name, last name vs. last name, first name)"
"Match shortened vendor names to full legal names"
"Treat hyphenated and non-hyphenated versions as the same"

Clear, descriptive instructions are critical for accurate AI identification, especially with complex datasets like payroll registers.

Review fuzzy match results

Run the matching process

Click Run matching from the workflow toolbar
The AI processes documents using fuzzy logic rules
Results appear in the workpaper grid

Verify matches

Check the matching column for results:

"Yes" indicates successful fuzzy matches
Review matched pairs to verify accuracy
Look for examples like:
- "Microsoft Azure" matched to "Microsoft Corp"
- "ProTrain Solutions" matched to "P-Train Solutions"
- "John Smith" matched to "Smith, John"

Click matched cells to see the source document and verify the AI made the correct connection.

Best practices

Combine with exact matches

Use fuzzy matching for one column (vendor name) alongside exact matches for others (amount, date).

Test instructions

Run on a small sample first to verify your instructions produce accurate matches.

Review all matches

Always review fuzzy matches manually before approving—false positives can occur.

Refine iteratively

Adjust instructions based on results, then rerun to improve accuracy.

Document Matching

Learn the basics of automatic document matching before adding fuzzy logic.

Manual Matching

Override fuzzy matches by dragging files to specific rows manually.

Request Validation

Set up validation rules to ensure uploaded documents meet criteria.

How is this guide?