Blog · Procurement and supply chain · Procurement
Create a dated supplier alias crosswalk without combining separate legal entities. Reconcile AP totals before and after normalization.
The same supplier can appear under several ERP identifiers, while two different suppliers can have almost the same trading name. A text-cleaning formula cannot reliably distinguish these situations. An approved crosswalk can, provided it records what was matched and why.
This method owns supplier identifiers for AP spend. It does not replace the customer-focused account hierarchy guide. Its output is a supplier mapping that the spend cube can use without silently changing the purchase population.
Trim spaces, standardize case and normalize punctuation in a candidate-name column. Preserve the original name. These steps help find possible duplicates; they do not establish that two records identify the same legal supplier.
Do not automatically remove every corporate suffix, combine suppliers sharing an address or treat a common payment destination as decisive proof. Groups, branches, agents and shared-service arrangements can explain similarity. Where the reporting question also needs a supplier-group roll-up, store that as a separate relationship rather than merging legal supplier identities.
Keep company and source system in the source key. Vendor V-017 in ERP A may have no relationship to V-017 in ERP B. A supplier record reused after a migration needs dated mapping, not a permanent lookup based on its current display name.
One row should identify source system, source company, source supplier ID, canonical supplier ID, valid-from/to dates, evidence reference, match status, reviewer and review date. Use restricted evidence references where necessary instead of copying sensitive documents into a broadly shared workbook.
Acceptable evidence depends on the company's supplier-master controls. A documented migration map or a reviewed supplier record may be stronger than a name resemblance. Procurement and AP should agree who can approve a merge and when an ambiguous case stays unresolved.
Microsoft documents multi-column merges and fuzzy matching. Treat fuzzy matches as candidates for review; a similarity score is not an entity-verification result. Merge overview, Fuzzy merge.
These fictional purchases use one USD net-spend basis.
| Source record | Display name | Spend | Reviewed canonical outcome |
|---|---|---|---|
| ERP A / V-017 | Northstar Supply LLC | $4,000 | S-100 |
| ERP A / V-091 | NORTHSTAR SUPPLY | $2,500 | S-100 |
| ERP B / A-210 | Northstar Supplies | $1,500 | S-100 |
| ERP A / V-033 | Beacon Industrial East Inc. | $3,000 | S-200 |
| ERP B / B-770 | Beacon Industrial West Inc. | $5,000 | S-300 |
| ERP A / V-099 | Acme | $1,200 | Unresolved V-099 |
A reviewed migration map supports the three Northstar records being one supplier. The Beacon records identify two separate businesses and remain separate despite similar names. Acme has insufficient evidence and stays on the exception list.
Canonical totals are S-100 $8,000, S-200 $3,000, S-300 $5,000 and unresolved $1,200. They sum to the original $17,200. The source has six supplier keys; the output has three confirmed suppliers plus one unresolved key. Calling the output four confirmed legal suppliers would overstate the evidence.
Left-join the approved crosswalk to invoice lines on source-system/company/supplier keys and applicable dates. Before expanding results, count matched mapping rows for each source key and date. Zero matches create an unresolved exception; multiple matches create a crosswalk-control failure.
Reconcile pre-join and post-join row count and signed amount. Normalization should change labels and aggregation, not spend. Keep the original supplier key in the result so a category manager can open the invoices behind a canonical total.
Test effective-date boundaries, migrated IDs, supplier acquisitions and expired approvals. If a group acquires a supplier, decide separately whether historical group reporting should use the parent in force then or a current-group comparison. Do not restate legal supplier identity solely to produce a convenient group total.
In the example, unresolved spend is $1,200/$17,200 = approximately 7.0%. One unresolved supplier out of six source keys is 16.7% by count, a different measure. Show both with their denominators.
Prioritize unresolved records that materially affect supplier concentration, contract matching or category decisions. A large ambiguous record can matter more than dozens of low-value spelling variants. Maintain an approval log and rerun controls after every mapping update.
The contract-match exception guide explains the next join. An authorized sample and reviewed master can support a scoped procurement analysis, but this guide does not promise automatic supplier verification or an available ERP connection.
Once supplier identities are reconciled, review invoice candidates, comparable purchase units and onboarding stages as separate operational questions.
No. It identifies a candidate for review. Approve a canonical mapping using evidence and preserve separate entities when the evidence differs.
A label mapping should preserve the source line count and signed amount. Changed totals indicate a join, perimeter or separate correction issue.