Good product data practices

Overview

Clean product master data makes it easier to find parts, keep stock accurate, and connect PartLogic to your ERP, WMS, spreadsheets, and automation tools. The same fields you prepare for import are what our free data quality checker evaluates on a sample of up to 25 rows—entirely in your browser.

This guide explains what to fill in, what to avoid, and how those choices affect your score. For GTIN validation and standard category taxonomies, see the GTIN check & categories guide.

Supported CSV formats

The data quality checker accepts two CSV formats. Choose Auto-detect on the checker page unless you know which export you uploaded.

Sage exports are mapped into the same scoring model as the PartLogic template:

• Code or Item Code → Item Code
• Description → Description*
• Sales Price or Sales Price 1 Value → Sales Price
• Rate → Sales Price (when Sales Price is blank, e.g. service lines)
• Usual Vendor → Supplier
• Vendor Code → Supplier Part Number
• Quantity in Stock → Quantity In Stock
• Barcode → Barcode

Sage stock exports typically do not include GTIN, manufacturer, or storeroom fields, so those checks may score lower until you enrich the data in PartLogic or your ERP.

Use the template

The free data quality checker also accepts the PartLogic stock import template below (see supported CSV formats). Product Match AI in the portal uses a different import file focused on descriptions and matching—see Product Match import columns. Download the checker template from the data quality checker or directly:

Download import template (.csv)

For scoring, these columns must be present (additional template columns are optional):

• Item Code, Description*, Category, Sub-Category
• Sales Price, Cost Price
• GTIN, Barcode
• Manufacturer, Manufacturer Part Number
• Supplier, Supplier Part Number
• Quantity In Stock, Storeroom, Location

Fill one row per part. Avoid blank rows; the checker may remove empty lines and pad missing cells with blanks before scoring.

Identifiers

• Item Code — your stable internal SKU or part code. Use a consistent format (no duplicate codes for different products).
• GTIN — the global trade identifier on the product or packaging (see GTIN check).
• Barcode — useful when you have a scannable code but not yet a confirmed GTIN.
• Manufacturer Part Number and Supplier Part Number — capture both where possible so procurement and matching workflows can align records across systems.

Use the portal GTIN check flow to validate structure and speed up category selection when you have a barcode number.

When the same product arrives under different names or codes from ERP, WMS, and supplier files, see Product identifier fragmentation for how PartLogic stores multiple identifiers against one governed master record.

Descriptions

Write descriptions for people—what the product is, size, material, or use—not as a dump of codes. Good descriptions are several words long and readable in a pick list or purchase order.

• Use more than one word — single-word descriptions incur a large penalty in the checker.
• Avoid repeated words — e.g. "valve valve brass" reduces clarity and costs points.
• Keep identifiers out of the description — do not paste GTIN, barcode, manufacturer or supplier part numbers, or category names into the description field; those belong in their own columns.

Categories

Provide both a Category (primary) and Sub-Category so reporting and filters stay consistent. During PartLogic onboarding you can align to GS1 GPC, UNSPSC, or a custom taxonomy—see GTIN check & categories for how to choose and apply a standard.

Use the same naming conventions across your catalogue (fixed spelling, no mixed synonyms for the same group).

Manufacturer & supplier

Record the Manufacturer name and Manufacturer Part Number together, and the Supplier name with Supplier Part Number. Partial data makes it harder to match duplicates, alternate suppliers, and external catalogues.

Prefer official manufacturer names over internal nicknames where you can, and keep supplier names stable so historical orders and stock movements still join correctly.

Pricing & stock

Cost Price and Sales Price should be numeric and greater than zero when the part is active in your catalogue. Use consistent decimal formatting in the CSV.

Quantity In Stock should be a valid number (zero is acceptable for out-of-stock items). Avoid text placeholders such as "TBC" in quantity fields.

Locations

Fill in either Storeroom or Location (or both) so warehouse and integration views know where stock sits. Use the same location labels your team uses on the floor to avoid split records for the same bin.

In the portal, storeroom and location values typically map to the Device level under a Site, while shared product fields live on the master record. See the Portal data model guide for the full hierarchy.

How the sample checker scores

Top score: only rows with a valid GTIN can reach 100. Without a GTIN, the score is capped at 99 even when other fields are complete. A barcode alone does not unlock 100.

The data quality checker applies the rules below to up to 25 rows in your browser. Your file is not uploaded to PartLogic servers during the demo.

Completeness checks

The score reflects how many checks pass, then description penalties are subtracted:

• −10 — Description is a single word
• −2 — Description contains repeated words
• −1 — Description embeds identifiers (GTIN, barcode, part numbers, category names)

Banding in the results: 80–100 strong, 50–79 needs attention, 0–49 poor. You can download a scored CSV to share fixes with your team.

Product Match AI

The sample checker scores field completeness on up to 25 rows in your browser. Product Match AI runs in the PartLogic portal after you sign in. It takes low-quality product descriptions and attempts to improve them, then presents recommendations for your team to review—nothing is applied automatically.

GTIN-first, GS1-aligned

PartLogic is a GTIN-first product data platform and a GS1 partner. We prioritise global trade identifiers and use the GS1 API where appropriate to validate and enrich product records. Product Match can suggest or assign a GTIN when one was missing, so your catalogue moves toward consistent, standards-based identifiers—not ad-hoc codes invented in isolation. For portal GTIN validation and category workflows, see GTIN check & categories.

How descriptions are improved

Improvements are driven by a bespoke vector store built from many years of supply-chain and product-mastering work—not a generic rewrite. The model compares your line against that curated knowledge to infer likely product type, brand, and attributes, then proposes a clearer description you can accept or edit.

Example: if a row only says "Bearing 12345" and the model recognises the pattern for that item, it may expand the description to something like "Ball Bearing 12345 : Grey steel, lubricated"—adding detail your team can confirm before it is saved.

Confidence scores

Each recommendation includes an AI-generated confidence percentage (0–100). The score reflects how likely the suggested match and improved description refer to the same real-world item, so reviewers can focus on high-confidence lines first and spend more time on ambiguous ones.

Human-in-the-loop review

Product Match is not an unattended automation tool. Users must review outputs—cleaned descriptions, match candidates, and suggested GTIN values—before anything is merged or published into your governed catalogue.

Product Match import file (not the checker template)

In the portal, Product Match expects an import parts CSV export—for example example_import_parts_YYYYMMDD_HHMMSS.csv. It is not the same file as partlogic-import-template.csv used by the sample checker: there are no stock, pricing, VAT, reorder, or location columns. Instead you supply identification and classification fields (plus optional image filenames for batch rows).

Download example import parts (.csv)

When packaging text is wrong

Shelf labels and supplier lines are not always trustworthy. Product Match can use photos—either per row via Image File in a batch, or through ad-hoc photo upload—to infer what the physical product actually is, even when the written description disagrees.

Batch files and ad-hoc identification

• File upload — upload an import-parts CSV or Excel file to process many lines in one run (supplier lists, legacy exports, onboarding).
• Photo identification — upload a photo of a physical product on its own; Product Match attempts to identify it when descriptions are missing or unreliable (for example old warehouse stock).

Trial accounts vs subscribed access

Free trial accounts can use Product Match in the PartLogic portal on a limited basis—the same review workflow, with caps on AI-assisted calls, parts stored, and users (see the pricing page). Paid plans raise those limits so you can run matching and description improvement across larger catalogues; Scale includes unlimited AI calls for high-volume work.

More detail and examples are on the ProductMatch overview. For a demo or help scoping a rollout, contact us.

Next steps

Use the checker on a representative sample before import. Sign in to the portal to upload files or photos through Product Match AI, review suggestions with confidence scores, and publish only what your team approves—or book a guided assessment for larger catalogue work.

Try the data quality checkerContact us