A consistent file name helps a person recognize a document; a well-designed index helps a system retrieve it. This guide provides a reusable standard for naming scanned files, defining metadata, assigning unique IDs, handling versions and exceptions, and reviewing search performance over time.
Overview
Document scanning projects often fail after the scanning is complete, when staff cannot tell which file is authoritative, search results are too broad, or important context exists only in someone’s memory. A naming convention and metadata dictionary prevent those problems by making every scanned record predictable.
Use the file name for information that should remain visible when a document is downloaded, attached to an email, or viewed outside the document management system. Use metadata for information that users need to filter, sort, report on, or apply through workflow rules. Do not force every possible detail into the file name.
A practical standard usually has four layers:
- Stable identity: a record, matter, customer, property, invoice, or employee ID.
- Date: the document date, using a consistent format such as
YYYY-MM-DD. - Document type: a controlled term such as
Invoice,Contract, orInspectionReport. - Version or status: a value such as
v01,Executed, orFinal, used only when it adds useful meaning.
For example: ACME-00482_2026-03-18_Invoice_v01.pdf. The exact fields can differ by department, but the order, spelling, separators, and date format should not change casually.
What to track
1. The naming convention
Write the naming rule as a formula before scanning begins:
[UniqueID]_[DocumentDate]_[DocumentType]_[VersionOrStatus].[Extension]
Use letters and numbers that work across the systems likely to receive the file. Avoid slashes, quotation marks, ampersands, long strings of spaces, and characters that may be treated as commands or separators. Keep document types short and controlled. Choose either PersonnelFile or EmployeeRecord; do not permit both unless they have different meanings.
Use leading zeroes for sequential numbers when they improve sorting, such as INV-000482. Preserve the original identifier when it is needed for reconciliation, but consider assigning an internal ID if external numbers can repeat across customers, departments, or systems.
2. A reusable metadata dictionary
Document each field in a table or system configuration record. At minimum, specify its name, purpose, format, whether it is required, who supplies it, and how it is validated.
- RecordID: unique internal identifier; required; assigned by the system or indexing team.
- DocumentDate: date shown on the document; required where available; format
YYYY-MM-DD. - DocumentType: controlled vocabulary; required; for example, Invoice, Contract, Deed, PersonnelRecord, or InspectionReport.
- EntityID: customer, matter, employee, property, project, or vendor identifier; required when applicable.
- Department: owning business unit; controlled list rather than free text.
- Status: Draft, Signed, Executed, Superseded, Void, or another approved value.
- Version: sequential value such as
v01; use for documents that can be revised. - Confidentiality: access classification, if the repository supports it; never use sensitive personal information as a substitute for an access-control rule.
- RetentionCategory: the approved records category or schedule reference; do not invent retention periods during indexing.
- SourceLocation: box, shelf, branch, file room, or batch reference where physical recovery may matter.
- ScanBatchID: batch or job identifier used for quality review and exception tracking.
- OCRStatus: NotRun, Passed, NeedsReview, or another defined value.
Keep sensitive fields out of file names when they could expose private information through downloads, browser tabs, backups, or email attachments. A neutral ID is usually safer than a full name, medical detail, financial account number, or government identifier.
3. Record-specific examples
Invoice: VEND-0187_INV-000482_2026-03-18_Invoice.pdf. Index VendorID, InvoiceNumber, InvoiceDate, Amount if needed for business search, Department, and PaymentStatus.
Contract: MAT-00214_2026-04-01_ServiceAgreement_Executed.pdf. Index MatterID, Counterparty, EffectiveDate, ExpirationDate, DocumentType, Status, and Version. Keep signed and draft copies distinguishable.
Personnel file: EMP-00482_2026-02-10_BenefitsElection.pdf. Index EmployeeID, DocumentType, DocumentDate, Department, and access classification. Avoid placing the employee’s name or sensitive benefit information in the file name unless the organization has expressly approved that practice.
Property record: PROP-7712_2025-11-06_InspectionReport.pdf. Index PropertyID, Address or site reference where appropriate, InspectionDate, Project, and DocumentType.
Customer folder: use a stable CustomerID rather than relying on a customer name that may change or be spelled differently. Index the customer-facing document type, date, owner, and relationship or case ID as applicable.
4. OCR, barcodes, and exceptions
OCR can populate fields from printed text, but extracted values should be checked against the source document for fields that drive billing, legal status, identity, or retention. A barcode index can improve consistency in bulk document digitization by carrying a record ID, document type, or separator instruction. Define what happens when a barcode is missing, unreadable, duplicated, or inconsistent with the paper record.
Create an exception code rather than silently guessing. Useful examples include DATE-MISSING, ID-UNREADABLE, DUPLICATE-SUSPECTED, TYPE-UNCLEAR, and QUALITY-REVIEW. Maintain an exception log containing the batch ID, page or file reference, issue, decision, person responsible, and resolution date.
Cadence and checkpoints
Review the standard at three points: before a scanning batch, during quality control, and after users begin searching the repository.
Before each batch
- Confirm the destination repository and permitted file formats.
- Confirm required fields, controlled vocabulary, and the responsible indexer.
- Test the naming formula with representative documents, including multi-page records and documents with missing dates.
- Check whether the batch needs barcodes, separators, OCR, or manual indexing.
- Record the BatchID, source location, expected record range, and exception process.
During quality control
Sample files from the beginning, middle, and end of a batch rather than checking only the first records. Verify that files open, page order is correct, names match the index, OCR text is usable, and required fields are populated. Track defects by category so recurring problems can be corrected at their source.
Monthly or quarterly review
Maintain a simple dashboard or worksheet with the number of records processed, missing required fields, duplicate IDs, unresolved exceptions, failed OCR checks, and retrieval requests that required manual assistance. The purpose is not to create a complicated reporting system; it is to reveal whether the standard is working in daily use.
How to interpret changes
A rise in missing dates may indicate that a source collection does not reliably display dates, not that indexers are careless. A rise in duplicate records may point to repeated rescanning, inconsistent identifiers, or a migration that imported the same files twice. More manual searches may mean that the metadata model is missing a field users naturally use, such as project number, claim number, or property ID.
Separate data-quality problems from process problems. If files are correctly named but difficult to find, improve metadata, filters, or search guidance. If names vary between batches, revise the controlled vocabulary and add validation. If OCR is unreliable on a particular document type, route that type for targeted review rather than applying the same confidence threshold to every collection.
Do not solve every problem by adding another field. Each required field increases indexing effort and creates another opportunity for inconsistent values. Add a field when it supports a clear retrieval, workflow, compliance, or audit need. Remove or make optional fields that are rarely used and difficult to verify.
When to revisit
Revisit this standard monthly or quarterly while a new scanning program is settling in, then at least annually once the process is stable. Review it sooner when the organization adds a department, changes its repository, migrates records, revises retention categories, introduces a new OCR or barcode workflow, or receives repeated complaints about search and retrieval.
Use this action list for the next review:
- Export a sample of recently scanned files and inspect names, metadata, OCR, and access classification.
- Ask several users to find common records without being shown the answer.
- List the fields they searched, the failed searches, and the information they expected to see.
- Compare exceptions and duplicates with the previous review period.
- Update the naming formula, dictionary, controlled terms, and examples only after testing the proposed change.
- Version the standard itself and communicate the effective date to scanning, indexing, records, and IT teams.
For related workflow planning, see the bulk scanning preparation checklist and the OCR and searchable PDF quality checklist. If scanned records will move into a cloud repository, the scan-to-cloud workflow and migration checklist can help connect naming rules with access controls and migration testing.