Fuzzy Matching

Written by Uday Kumar Updated

Overview

Fuzzy Matching in SmartVendor AP Invoices allows the system to treat “close but not exact” text values as potential matches when identifying:

  • Purchase Orders (POs) and PO lines

  • Vendors, entities, and other SoR‑backed dropdown values (via Templates)

This is especially useful when:

  • There are human errors or formatting differences in the invoice (e.g., spaces, dashes, OCR noise in PO numbers).

  • Vendor names, entities, or identifiers in the invoice don’t exactly match the way they are stored in the System of Record (SoR).

Where enabled, fuzzy matching increases match rates and reduces manual corrections, while still requiring a sufficiently high confidence match to avoid mis‑association.

Where Fuzzy Matching Is Used

In AP Invoices, fuzzy matching can be used in two main places:

  1. SmartFlow Settings – Allow Fuzzy Matching (PO Configuration)

    • Exposed in the SmartFlow wizard as Allow Fuzzy Matching = Yes/No.

    • Controls whether the system attempts fuzzy matching for certain PO‑related identifiers (e.g., PO Number, Item ID, Vehicle Registration Number) when linking invoice content to SoR records.

  2. SmartVendor Templates – Field‑Level Fuzzy Matching

    • In SmartVendor > Settings > SmartVendor Templates, you can enable Fuzzy Matching for specific text or dropdown fields.

    • This governs how SmartVendor interprets free‑text extracted values when attempting to match them to SoR‑backed lists (e.g., Entity, Worktags, custom dropdowns).

1. Fuzzy Matching for PO and Line Matching

When Allow Fuzzy Matching is turned on in the SmartFlow Settings (Step 2: PO Configuration):

  • SmartVendor can interpret slightly malformed PO numbers on invoices as matches to actual PO numbers in the SoR.

  • Example:

    • SoR PO Number: PO-4281

    • Invoice text: P0 4281, PO4281, or P42 81 (with spaces or OCR noise)

  • With fuzzy matching enabled, SmartVendor can recognize these as likely the same PO.

This capability is tied to:

  • PO number extraction from the invoice

  • Fuzzy matching of the extracted PO number against SoR POs

Important constraints (implementation notes from Jira):

  • Fuzzy matching is not a replacement for the primary “exact” match logic; it’s used after failing to find an exact match.

  • Release notes make clear that in some configurations “PO number takes priority; fuzzy matching is not used for PO match” (e.g., when Allow PO Override is the main behavior). That is customer/tenant‑specific:

    • For some tenants, fuzzy PO matching is disabled by design even if extraction is imperfect.

    • For others, it can be enabled via Allow Fuzzy Matching + template settings.

If fuzzy matching does not find a high‑confidence candidate:

  • The PO field remains unmatched,

  • The document will require human review (e.g., user chooses PO from dropdown).

2. Fuzzy Matching via SmartVendor Templates

At field level, you can configure fuzzy matching in SmartVendor Templates:

  • Navigate to SmartVendor → Settings → SmartVendor Templates → Review and Edit your template.

  • For supported fields (typically text and dropdown fields), you can toggle Fuzzy Matching ON.

When enabled:

  • SmartVendor treats extracted field values as approximate queries against the SoR value list.

  • It uses fuzzy matching techniques (partial matches, normalized spacing, case‑insensitive comparisons, and other heuristics) to pick the best candidate.

Examples:

  • Vendor / Entity names:

    • Extracted: Amazon Web Servces

    • SoR value: Amazon Web Services

    • With Fuzzy Matching ON, SmartVendor can still pick Amazon Web Services.

  • Worktag / custom dropdowns:

    • Extracted: IT Ops

    • SoR values: IT Operations, IT Support

    • Fuzzy Matching may prefer IT Operations if that is the closest string and historically chosen for similar patterns.

This is complementary to:

  • Vendor Hints (explicit mappings such as AWS → Amazon Web Services)

  • Predictive Coding (using historical selections to suggest values)

Template‑level Fuzzy Matching affects how extracted values are resolved to dropdown options, while Hints and Predictive Coding help decide which candidate is most likely correct.

3. Vendor / Entity Fuzzy Matching

Fuzzy matching is also used internally to improve vendor and entity identification:

  • Combined with Vendor Hints and entity restrictions, SmartVendor:

    • Uses PO entity and hints as higher‑priority sources.

    • May use lookup‑based fuzzy matching to guess the best entity when:

      • No PO entity is present, and

      • No hint‑based entity is available.

Priority for entity population:

  1. PO Entity

  2. Hint Entity

  3. Lookup / Last Bill / Predicted Entity (one configurable fallback)

  4. Manual entry if none found

In step 3, “Lookup assigned entity” uses fuzzy matching between the bill content and the list of allowed entities for that vendor.

4. UI Indicators – Info Icon

When fuzzy matching is influencing a field (especially PO‑related fields), SmartVendor can display an Info icon next to the field:

  • Indicates that the value may have come from fuzzy matching (e.g., PO Number, Vehicle Reg Number, Item ID).

  • Hovering or clicking can provide more context (for example, that the value was matched fuzzily, not exact).

This visual indicator:

  • Helps AP analysts understand why a value was chosen,

  • Encourages them to validate fuzzy‑matched identifiers carefully before writing to SoR.

5. Limitations and Edge Cases

Current and historical behavior notes:

  • Fuzzy matching does not always apply to PO Number in some tenant configurations.

  • Certain edge cases (e.g., heavy OCR corruption, very short IDs, or too many similar candidates) will cause the system to decline fuzzy matching and leave the field unmatched.

  • Fuzzy matching is generally not used where ambiguity would introduce high risk of mis‑posting (for example, posting against the wrong PO or wrong entity).

Where fuzzy matching fails or is not enabled:

  • The system relies on exact matching, hints, predictive coding, or manual user selection.

  • AP analysts can always override and correct values directly on the Document Information page.

6. Configuration Guidance

SmartFlow Settings (PO Configuration)

  • Allow Fuzzy Matching = Yes

    • Use when:

      • Your invoices frequently contain minor PO formatting errors (spaces, missing dashes, etc.).

      • You want higher auto‑match rates for PO Number and related identifiers.

    • Be sure users understand to check Info icons and PO linkage carefully.

  • Allow Fuzzy Matching = No

    • Use when:

      • Your environment is highly controlled; vendors almost always include exact PO numbers.

      • The risk of incorrect PO matches outweighs the benefit of automatic fuzzy linking.

SmartVendor Templates – Field‑Level Fuzzy Matching

  • Enable Fuzzy Matching for:

    • Vendor‑like names (where you don’t already rely fully on Vendor Hints).

    • Entity, custom dropdowns, or worktags where human‑entered names on invoices are often “close but not exact.”

  • Keep Fuzzy Matching OFF for:

    • Highly sensitive identifiers where only exact matches are allowed (e.g., specific contract IDs, critical GL codes if they are numeric and strict).

7. Best Practices

  • Combine Fuzzy Matching + Vendor Hints for strong vendor and entity resolution:

    • Use hints to map known abbreviations or alternate labels.

    • Let fuzzy matching catch new, yet similar variations.

  • Use Info icons and audit history to build user trust:

    • Train AP analysts to review fields with Info icons when invoices look atypical.

    • Use document history to see if a value came from:

      • Exact match

      • Fuzzy match

      • Hint

      • Predictive coding

      • Manual edit

  • Test Fuzzy Matching with a sample set:

    • Run representative invoices with known PO typos through AP Invoices.

    • Verify that matches are correct and that any problematic patterns can be mitigated via Hints or stricter settings.