If you are searching for how to translate barcodes, how to translate GTIN, EAN, UPC or SKU information, or how to preserve retail product identifiers across languages, the key distinction is between product language and product identity. A product description can be translated. A GTIN, EAN, UPC, SKU, catalogue number, model number or barcode value is usually an identifier that must remain exact. Changing one digit may point to a different product, package size, variant or inventory record.
Product-identifier translation matters in e-commerce, packaging, retail catalogues, inventory systems, marketplaces, logistics, customs, product manuals and multilingual product databases. Translators often work next to structured data: a title in one column, a GTIN in another, a brand name in a third, a size or colour code in a fourth. The linguistic fields should become natural and searchable in the target language, while the identity fields must stay aligned with the exact product they describe.
This guide explains how to translate product information around barcodes, GTINs, EANs, UPCs and SKUs without breaking retail identity. It shows how to distinguish global identifiers from retailer-specific codes, how to protect check digits and leading zeros, how to keep pack-level identifiers separate, how to handle QR codes and serialised labels, and how to review a multilingual product catalogue so translated text never becomes detached from the item it belongs to.
Why a product identifier is not a word to translate
Retail systems use identifiers to answer a basic question: which exact item is this? Human-readable names are unreliable for that job. Two products can have almost identical names but different flavours, pack sizes, colours, voltages or regional configurations. A machine-readable identifier lets databases, scanners and supply chains distinguish them.
Translation introduces risk because the text and identifier often appear together. If a translator duplicates a target title onto the wrong row, changes a digit while copying, drops a leading zero, or assumes that a master carton and a single unit share the same GTIN, the translated content can become operationally false even when every sentence is fluent.
Strong product translation therefore uses an identity-first rule: lock the identifier, identify the exact commercial item, translate only the fields intended for users, then verify that the target text still belongs to the same code. This is not merely database hygiene. It is translation fidelity at product level.
A reliable method for translating product identifiers and retail data
1. Classify every field before translating
Mark fields as translatable text, protected identifiers, numeric specifications, controlled vocabulary or media links. Product title and marketing description may be translated. GTIN, UPC, EAN, SKU, barcode data, model code and serial number generally should not.
2. Identify the product level
Confirm whether the identifier belongs to an individual consumer unit, multipack, case, carton, pallet or another packaging level. Similar products at different packaging levels can carry different identifiers. Translation should never merge those rows simply because the visible name looks the same.
3. Preserve digits and check characters exactly
Identifiers may contain check digits or structured prefixes. Do not recalculate or “correct” them from memory. Copy the source exactly and compare the final target data character by character.
4. Protect leading zeros
Spreadsheet software can interpret identifier fields as numbers and remove leading zeros. Treat product codes as text. A zero that looks mathematically insignificant may be part of the identity string expected by a scanner, API or marketplace.
5. Keep brand, model and variant relationships intact
Translate descriptive variant language such as flavour, colour or size where appropriate, but keep the corresponding model or SKU tied to the exact variant. A red product and a blue product may share most words while remaining different commercial items.
6. Distinguish global identifiers from internal codes
A GTIN can identify a trade item globally, while an SKU may be retailer- or seller-specific. Do not relabel them as equivalents. Preserve the source field name or explain its role clearly in the target.
7. Validate machine-readable and human-readable forms together
A printed barcode image and the digits shown beneath it should correspond to the same encoded value. Translators normally do not recreate the barcode symbol itself, but if artwork changes, final production QA should confirm that the printed symbol, visible number and product text still match.
8. Run a row-level identity audit
At the end, sample or systematically compare each target row against the source: identifier, product name, variant, size, unit and market. The goal is to prove that no translated description has drifted onto a neighbouring code.
Twenty product-identifier translation problems
1. GTIN
A GTIN identifies a trade item within a global identification system. The digits are protected data. Translate labels and product descriptions around the GTIN, not the GTIN itself. If the source uses a specific GTIN format, preserve it as supplied.
When several GTINs appear for different packaging levels, do not assume one is redundant. A consumer unit, case and multipack can legitimately require separate identifiers. Keep target descriptions aligned with the appropriate level.
2. EAN
EAN terminology is widely used in retail contexts. Preserve the identifier and recognised label. If the target audience uses GTIN terminology more commonly, a short explanatory connection may be useful, but do not alter the source code.
Do not strip digits because the number “looks too long.” Treat the entire string as product identity, not as a quantity.
3. UPC
UPC values should remain exact. A target-market product page can translate the product title and attributes while continuing to display the original UPC when that is the identifier used by the item and platform.
If a marketplace asks for “UPC/EAN/GTIN,” do not guess which field to populate. Use the identifier provided for that product and platform requirement.
4. SKU
An SKU is often an internal stock-keeping code. It may contain meaningful abbreviations for colour, size or department, but those embedded fragments do not automatically become translatable. Changing them can disconnect the product from inventory systems.
Translate the visible variant labels for customers while keeping the internal SKU stable unless the inventory owner explicitly maintains locale-specific SKU systems.
5. Model numbers
Model numbers identify product families or variants and generally remain unchanged. Do not translate letters that happen to resemble abbreviations unless the manufacturer officially uses a localised model convention.
Model names and model numbers are not always the same thing. A marketed model name may have a localised form while the alphanumeric model number remains fixed.
6. Catalogue numbers
Manufacturers, laboratories and industrial suppliers use catalogue or part numbers to identify exact items. Preserve them character for character and translate only the descriptive label or product text.
Catalogue numbers are especially important where products have tiny technical differences that are not visible in the name. The identifier can be more precise than the prose.
7. Check digits
Many standard identifiers use a check digit to detect common errors. A translator should not manually recompute it as part of ordinary translation. Preserve the source value and use validation tools only as QA, not as permission to silently alter source data.
If validation suggests the source code is invalid, flag the discrepancy to the content owner. Translators should not invent corrected product identity.
8. Leading zeros
Leading zeros are vulnerable when CSV or spreadsheet files are opened. Protect the field as text. A visually shortened code may no longer match a marketplace or warehouse record.
QA should compare character length as well as visible value. This simple check catches many spreadsheet-induced errors.
9. Product variants
Colour, flavour, scent, memory capacity, voltage and size variants can share a base title while using different product identifiers. Translate variant names naturally but never copy one target description across codes without checking the variant field.
A useful QA view sorts products by base model and compares variant identifiers side by side. This reveals duplicated or swapped translations quickly.
10. Multipacks and cases
A single bottle, six-pack and shipping case can have different identifiers even when the product inside is the same. Translate pack quantity and packaging level explicitly so users and systems do not confuse consumer and logistics units.
Do not “simplify” all pack names to one product title. Quantity is part of commercial identity.
11. Retailer-specific SKUs
The same physical product can carry different retailer or seller SKUs. Keep each marketplace’s internal code attached to the correct local listing. A global GTIN does not automatically replace seller-specific inventory identity.
When migrating multilingual listings between systems, map identifier fields explicitly rather than assuming similarly named columns mean the same thing.
12. Batch and lot numbers
Batch or lot numbers identify production groups and can be critical for recalls and traceability. Preserve them exactly. Translate the label “batch,” “lot,” or “production lot” according to target convention without changing the value.
If a source label combines lot number with production date, keep the fields distinct. Similar strings can serve different traceability roles.
13. Serial numbers
Serial numbers identify individual units rather than product classes. They are protected values and should never be translated. A serialised warranty or service document depends on exact one-to-one identity.
Do not confuse serial number with model number. The model describes a type; the serial usually identifies one physical item.
14. QR codes
A QR code can encode a URL, identifier, contact record, authentication token or other data. Translators should identify whether the encoded destination is locale-specific. The image itself should not be recreated unless the content owner intentionally changes the encoded value.
If a translated page uses a different localised URL, the QR code may need to be regenerated by the publishing workflow. That is a localisation-production decision requiring functional testing, not a casual translation edit.
15. Data Matrix and other machine-readable codes
Machine-readable symbols may encode identifiers that users rarely see. Preserve the underlying data and coordinate with production teams if artwork changes. Translators should not assume an unreadable symbol is irrelevant merely because it is not ordinary prose.
For regulated or traceability-sensitive products, verify that the final symbol still scans to the intended value after layout and export.
16. Marketplace identifiers
Online marketplaces may assign their own listing identifiers in addition to manufacturer and global trade-item codes. Keep platform identifiers stable while translating titles, bullets and descriptions.
A target-language listing should remain attached to the same underlying catalogue item. Duplicate listings can arise when translated content is imported as a new product instead of a localisation of the existing one.
17. Brand names beside identifiers
Brand names may have official local forms or may remain unchanged. Whatever naming strategy is used, keep the same brand-product-code relationship. Do not localise a brand into a form that makes target readers believe the code belongs to another company.
Check the manufacturer’s official target-market site where possible. Official naming improves both trust and searchability.
18. Size and colour codes
Internal size or colour codes can look translatable because they contain abbreviations such as BLK, RED, S or XL. Preserve the code unless the product system explicitly maintains locale-specific codes. Translate the customer-facing colour and size labels separately.
This separation avoids a common error: turning an operational SKU into a human-readable description and breaking inventory matching.
19. Customs and logistics references
Product identifiers can travel into customs declarations, shipping documents and warehouse labels. Translation should preserve the same identifier across all documents so logistics teams can reconcile the physical goods with invoices and product records.
Do not create a “translated code” for customs. Translate commodity descriptions and explanatory fields while keeping product identity stable.
20. Product recalls
Recall notices often identify affected products by model, lot, serial or GTIN. These values must remain exact. A mistranslated descriptive name is serious; an altered recall identifier can be worse because consumers may fail to recognise an affected unit.
Translate the safety message clearly and preserve every affected identifier. If the recall lists ranges or multiple variants, check each boundary and code independently.
Common failure modes
Opening identifier columns as numbers. This can strip leading zeros or apply scientific notation. Store codes as text.
Translating abbreviations inside SKUs. Internal codes are not automatically customer-facing language.
Copying one translation across variants. Products that differ only by colour or pack size can still require distinct target descriptions.
Treating GTIN, UPC, EAN and SKU as synonyms. They can play different roles. Preserve the source field identity.
Recomputing check digits without authority. Validation can flag a problem, but translators should not silently invent corrected product codes.
Ignoring packaging level. Consumer units, multipacks and cases can legitimately carry different identifiers.
Regenerating QR codes casually. A QR code is a data container; changing it requires controlled production and functional testing.
Failing to reconcile identifiers across channels. Website, marketplace, catalogue, packaging and recall data should point to the same product identity.
Worked practice
Practice 1: Three colour variants. The source has three rows with nearly identical titles but distinct SKUs and GTINs. Translate each colour naturally, then verify that every target row retains its original code pair.
Practice 2: Single unit and case. A beverage bottle and a case of twelve have different identifiers. Preserve pack quantity in the target title and keep the correct identifier with each packaging level.
Practice 3: Leading zero. A UPC-like value begins with zero. Format the spreadsheet column as text before editing and compare string length after export.
Practice 4: Internal colour code. An SKU contains BLK while the target customer-facing colour is translated. Keep BLK in the SKU and translate “black” only in the visible attribute field.
Practice 5: QR-linked manual. Packaging contains a QR code to an English manual. If the project later creates a target-language manual and local QR destination, regenerate the code through the publishing workflow and test it; do not alter it during ordinary text translation.
Practice 6: Recall notice. Translate the safety instructions, but preserve every model, lot and serial range exactly. Ask QA to compare identifiers independently from prose.
Practice 7: Marketplace migration. The source platform uses seller SKU plus GTIN; the target platform uses separate catalogue identifiers. Map fields explicitly before importing translated titles.
Practice 8: Branded shade. A cosmetic item has an official shade name and product code. Follow the brand’s target-market naming for the shade if available while keeping the code unchanged.
Using CAT tools, spreadsheets and AI safely
CAT tools can protect recurring identifiers and compare alphanumeric strings between source and target. Use that capability. Product datasets also benefit from terminology controls so the same attribute is translated consistently across thousands of rows without touching protected code fields.
Spreadsheets and CSV files need deliberate handling. Preserve column types, encoding and row order. Identifier fields should be text. Be careful with automatic conversion of long numbers, dates and scientific notation. A technically fluent translation is unusable if the import process has corrupted product identity.
AI can help draft product descriptions and classify attribute language, but it should not generate or “repair” GTINs, UPCs, SKUs or serial numbers from context. Treat identifiers as immutable inputs. If source data appears invalid, escalate it to the data owner rather than allowing the language model to invent a plausible replacement.
How this fits the wider eduKate translation system
This is a specialised product-identity owner within eduKate’s existing translation architecture. The broad translation method remains Master Art of Translation | The Complete System for Moving Meaning Between Languages. The Vocabulary Learning Hub supports precise attribute and product-language choices, while How English Works supports clear noun phrases, modifiers and labels. Product identifiers add the rule that translated language must remain anchored to the exact commercial item.
FAQ
Should a barcode number be translated?
No. Preserve the encoded identifier. Translate only surrounding labels and product information.
Are GTIN, EAN and UPC the same thing?
They are related retail-identification concepts and formats, but translators should preserve the source field terminology rather than treating every label as interchangeable.
Should SKUs be translated?
Usually not. An SKU is often an internal inventory identifier. Translate customer-facing attributes separately.
Why do leading zeros matter?
They can be part of the identifier. Removing them may make a code fail validation or match the wrong record.
Can a product name be translated while its model number stays unchanged?
Yes. That is common. Keep the identifier stable while adapting the human-readable title for the target market.
What about QR codes?
Identify what the QR code encodes. Change it only through a controlled localisation workflow when the encoded destination intentionally changes.
Should check digits be recalculated?
Use validation to detect possible source problems, but do not silently alter identifiers. Escalate discrepancies to the data owner.
How do I prevent row mismatches?
Keep a stable unique identifier with every row throughout translation and audit target rows against that identifier after sorting, filtering or import/export operations.
Can AI translate product catalogues safely?
AI can assist with product language, but protected identifiers, variant mapping and structured fields still require deterministic controls and QA.
What is the simplest rule?
Translate what the customer reads; preserve what the system uses to identify the product.
Final checklist
- Have I classified translatable and protected fields?
- Are GTIN, EAN, UPC, SKU, model and catalogue identifiers unchanged?
- Are leading zeros and full string lengths preserved?
- Does each target description remain attached to the correct product row?
- Are pack levels and quantities still distinct?
- Are colour, size and other variants mapped to the correct code?
- Have batch, lot and serial identifiers been preserved?
- If a QR or machine-readable symbol changed, was it regenerated and functionally tested?
- Do source and target marketplace fields map correctly?
- Could a retailer, warehouse or customer identify the same exact product from the target data?
Product translation succeeds when language becomes local without product identity becoming localised away. The title can change, the description can become more natural, and customer-facing attributes can follow target-market vocabulary. The identifiers that bind those words to a real item must remain stable. Protect the code, preserve the row, verify the variant and test the data flow. That is how multilingual retail content stays both readable and operationally true.