Finance operations

Detect Duplicate Records in back office administration

A duplicate check for back office administration that uses two identifiers, holds both records, and leaves the merge decision with the owner.

Operational guidance only. Qualified owners retain accounting, payroll, HR, legal, security, insurance, and policy decisions.

A team presentation with laptops and written notes

Treat a match as a review candidate

Two records that look alike are not automatically the same entity. A customer may share a name with another customer; two locations may use the same billing address. Prepare a candidate list with the reason each pair was flagged. Preserve both records until an authorized owner decides what to do.

Define the population and the comparison

Choose a bounded record set and identify its authoritative system. Agree which fields can indicate a match and which differences require a stop. Compare stable identifiers where available; names, addresses and dates can help review but may be incomplete or change over time.

EvidenceWhat the assistant recordsWhat remains uncertain
Record identifiersBoth original referencesWhether they represent the same entity
Matching fieldsExact values and source locationsWhether a shared value is unique
Conflicting fieldsBoth values without overwritingWhich source governs
Linked activityOrders, invoices or other relationshipsHow consolidation affects history
Proposed next actionReview question and ownerPermission to merge, void or retain

Example: one customer name, two branches

Two customer records have the same trading name and head-office address, but different branch identifiers and active orders. The assistant records the apparent match and the conflicting branch information. The owner may decide that the accounts must remain separate. A matching name should not erase that relationship.

Keep detection separate from merging

Use approved comparison rules to flag candidates. Do not run an irreversible bulk merge merely because a similarity score is high. The reviewer needs to decide the surviving identifier, field values, linked-record treatment and any required communication. A tidy-looking list is not evidence that a merge is safe.

Prepare a traceable decision packet

Return the candidate pair, source links, match reason, conflicts and related activity. Ask one clear question: retain both, correct a field, investigate further or authorize a defined merge. Record the dated decision and its scope. An unanswered item stays open with an owner.

Verify an authorized change

Where the system permits an approved change, retain the before-and-after identifiers and the execution result. Check that intended relationships remain accessible and unrelated records were not changed. Follow the system owner's recovery process for an incorrect result; do not improvise a second merge to conceal the first error.

Improve the intake rule

Review false positives and recurring causes, such as inconsistent identifiers or imports that bypass existing-record checks. Test a proposed rule on a representative sample before expanding it. Data entry support can maintain the candidate queue and evidence while your team retains merge authority.

Questions to settle before starting

Is a matching email address sufficient?

It may be useful evidence, but shared inboxes, recycled addresses and imported errors can create false matches. Use the approved identity rule for the record type and preserve conflicting evidence.

What should the reviewer receive?

Both references, matching and conflicting fields, linked activity, the proposed action and its limitations. Include source locations so the reviewer can inspect the original records.

What does completed mean?

The authorized decision was applied or explicitly deferred, the result was checked and the evidence is retrievable. Counting flagged records alone does not establish a successful cleanup.

Useful internal handoffs

Sources