Product data governance that keeps catalogues reliable
Stop conflicting product edits reaching shoppers. Assign field owners, validation rules and approvals, with a clear way to correct and trace changes.
- Published
- Last updated
Product data governance sets who owns each catalogue field, which changes are allowed and what evidence must remain after publication. Give every field an authoritative source and a correction owner, then keep the approved version separate from incoming edits. You can then trace a wrong product answer back to the fact that caused it.
Assign authority at field level
Assign responsibility to individual fields because one supplier file rarely deserves authority over every part of a product record. Your manufacturer may know the dimensions, while your stock system knows what is available to sell.
Separate four roles. The owner decides the field’s meaning and rules. An editor proposes or enters changes, and an approver authorises changes that need judgement. A consumer uses the approved value, such as a storefront, marketplace feed or shopping assistant.
A small team can give one person several roles. Record each role separately so nobody mistakes permission to edit for authority to change a rule.
| Field | Authoritative evidence or system | Accountable owner | Approval focus |
|---|---|---|---|
| Dimensions | Manufacturer specification or documented physical measurement | Product-data lead | Product versus packaging dimensions and measurement method |
| Product claims | Evidence for the exact product and claim | Person responsible for compliance | Evidence scope, wording and expiry |
| Description | Approved product facts plus retailer editorial rules | Merchandising lead | Unsupported statements and changes of meaning |
| Selling price | Retailer pricing system | Pricing lead | Currency, tax basis and promotion rules |
| Available stock | Inventory system with a defined availability calculation | Inventory lead | Location, reservations and update failures |
Document which source wins when two disagree. A supplier’s recommended retail price must not overwrite your selling price merely because the file arrived later.
Define rules teams can apply
Define validation by product category, including what a value means and when it is required. A furniture category needs dimensions and units. An electronic accessory needs an explicit compatibility relationship to the supported model or variant.
Keep a field dictionary with the permitted format, allowed values and authoritative source. State whether dimensions describe an assembled item, a folded item or its packaging. For packaged consumer items, GS1’s measurement guidance explains how the default front determines height, width and depth.
Specify identifiers separately from descriptions. Preserve leading zeroes in identifier fields and distinguish the retailer’s SKU from the Global Trade Item Number (GTIN). A valid format or check digit does not prove that a code belongs to the intended item.
The GS1 GTIN Management Standard distinguishes a new product from changes to an existing product. Some changes require a new GTIN. Route those decisions to the responsible owner before overwriting an existing identity.
Check parent and variant relationships too. A compatible accessory must point to the relevant product version, and a pack must retain its quantity and unit. Resolve uncertain identities before merging their facts. The guide to product matching explains how to decide which records describe the same item.
Use explicit missing-value states. Unknown means the value has not been established. Inapplicable means the category rule does not require it. Expired means previously accepted evidence no longer supports using it.
Give inapplicable values a reason and expired values a review owner. Never convert an unknown weight into zero or unknown compatibility into “universal”.
Control incoming changes
Stage incoming files and staff edits before they replace approved values. Compare proposed changes field by field so a reviewer sees what will change and which source supplied it.
Use this sequence for each import:
- Save the source file with its supplier, import reference and receipt time. Keep its mapping to catalogue fields.
- Match each record to the correct product or variant. Hold ambiguous matches for a named person to resolve.
- Run category rules and compare accepted incoming fields with the current approved version. Show old and proposed values together.
- Route changes according to risk. Let permitted formatting corrections pass automatically, while unsupported claims and compatibility changes require an approver.
- Publish the approved version and record which channels receive it. Confirm each affected channel uses the corrected value before closing the change.
An urgent correction needs a shorter approval path with the same evidence trail. Consider a hypothetical adapter incorrectly marked compatible with model B. The product owner confirms that only model A is supported and approves the correction.
Pause the affected recommendation where the old answer could still reach shoppers. Publish the new relationship, refresh the storefront and feed, then refresh any assistant’s product information. Ask about model B in each affected interface. Close the incident only when the incorrect recommendation has stopped, with failed destinations assigned for follow-up.
A successful import receipt proves that the file arrived. It does not prove that every channel now uses the approved value.
Preserve evidence and versions
Keep enough history to reconstruct the exact product facts a channel used when it gave an answer. Retain the original source alongside the transformed value, approval decision and effective record version.
For a product claim, store the supporting document and its relevant passage. Record the product or variant it covers, who approved the wording and any expiry condition. A replacement document must not erase the evidence used for an earlier approval.
For each change, retain the old and new values, editor, reason and approver where required. Include the effective time and the version delivered to each channel. Protect published versions from silent edits and restrict access to evidence that contains confidential supplier information.
In the adapter example, start with the wrong compatibility answer and identify the product version used by the assistant. Follow its compatibility field to the import reference and supplier row. Compare that row with the original specification.
A supplier row for model A that becomes an approved field for model B points to mapping or editing. An assistant answering model B from an approved model A record needs a channel refresh or retrieval investigation. Those paths lead to different correction owners.
Resolve exceptions and measure quality
Send invalid records to a correction queue with a reason and a responsible person. Include the affected field, proposed value, breached rule and the next action needed to release it.
Keep valid records moving when failures are independent. Hold an entire batch when the problem affects shared identity mapping or would break product relationships. Report how many records were accepted and held so a partial import cannot look complete.
GS1’s Data Quality Framework covers management processes and physical inspection of product attributes. Automated checks therefore need a route for checking facts against the product or its supporting evidence.
Measure different defects separately:
| Measure | Practical calculation | What a good score still misses |
|---|---|---|
| Completeness | Required applicable fields populated divided by required applicable fields | A filled field can contain the wrong fact |
| Validity | Checked values passing defined rules divided by values checked | Plausible values can pass while describing another product |
| Consistency | Compared values or relationships that agree under the authority rules divided by comparisons made | Several channels can repeat the same error |
| Freshness | Records within their field-specific review or sync interval divided by records due for that check | A recent update can still be inaccurate |
Choose review intervals by the risk and rate of change. Stock needs different freshness rules from material composition. Report unresolved high-risk exceptions alongside percentages, with the oldest exception and its owner. Treat expiry as an actionable condition rather than a lower score alone.
Start with a manageable process
Start with a shared register for one category and the channels that use its data. Give the product-data lead authority to resolve conflicting sources and assign a backup for urgent corrections.
Use one register for field rules and another for changes. The field register needs the field definition, category requirement, source and owner. Add the editor, approval threshold and receiving channels. The change register needs the product identity, old and new values, evidence reference, decision and affected channel versions.
Keep evidence files and approved catalogue exports in controlled storage. Restrict who can publish and make the change register searchable by SKU and import reference. A spreadsheet can coordinate a small workload, but access controls and recoverable versions still need deliberate setup.
Move to a product information management (PIM) workflow when concurrent edits, category rules or missed approvals make the register unreliable. Evaluate the workflow against an actual disputed change. Require it to show the proposed difference, enforce the assigned approval and retrieve the effective version and evidence. Check those behaviours in the specific product and plan before replacing the register.
Remove a manual approval only when an automated rule controls the same risk. Converting a recognised unit can follow a tested rule that preserves the original value. An automated check for a populated claims field cannot judge whether the claim is supported.
For wider ecommerce automation, keep the same correction owners and channel confirmation steps. Pick one previously disputed product and reconstruct its source, approval and published version. If the team can correct it and confirm the result in every affected channel, expand the process to the next category.
Give your store its own shopping agent
Customers ask in their own words. Your agent finds, compares and checks out from your catalogue, right on your site.