Master data: SKU, SKU, barcode
Style, color-way, size and code: at which level each field lives, how many records to keep and how GTIN and marking work.
About the topic
Master data are the reference records that must be identical in every company system: what a model, colorway, SKU, barcode, category, season and supplier are. It is a dull topic until one day stock on the site differs from the warehouse, and a factory delivery note does not match the order only because black is called black in one place, schwarz in another and blk in a third.
Why it matters to a brand or retailer. Every report, every stock exchange, every labeling code upload and every marketplace match rests on the same records. If a record is single and accurate, everything runs automatically; if there are several and they diverge, hours go to manual reconciliation and customers find the errors. The decisions here: at which level to store each attribute, which system is the master for the card, who may change a record and which code every product variant receives.
The logic is built on levels. The model (style) describes the cut: name, category, composition, fit. The colorway adds color and often price and photos. The SKU is the lowest unit, color and size together; it is what is sold, counted in the warehouse and given a barcode. A set or carton has its own code and its link to the SKUs inside. If you assign a barcode to the style, a dress in three colors and six sizes gets one code for 18 different items, and neither the warehouse nor the till can tell what was sold.
The second rule: reference lists live apart from text. Color, size, season and category come from a list with a stable code, and customer text is assembled from it. Then black does not multiply into three spellings, and a site filter and a color report see one value. The third: every record has one system of record. It is changed only there, and elsewhere it arrives as a copy on a schedule. Otherwise two systems edit one field differently and nobody knows which version is right (the golden record the owner protects).
Where to start: draw the levels and the mandatory attributes at each, choose the card’s master, set up lists for colors, sizes and seasons, and give each SKU one GTIN with a check-digit test. Then count the SKUs and the labor for cards: this calculation is often sobering, because adding another color and size multiplies the work. Typical traps: confusing the style SKU with the SKU, keeping color and size as free text, holding a card in several systems with no master, losing the link of carry-over goods to the ancestor model and creating a new SKU instead of fixing an error in an existing one.
Key ideas
Three levels: model, colorway, SKU
Dress D-100 in 3 colors and 6 sizes: one model, three colorways, 3 × 6 = 18 SKUs. Composition and category live on the model, color and photos on the colorway, barcode, weight and stock on the SKU. If everything is kept on one level you copy composition 18 times, and a composition fix becomes 18 fixes.
A reference list instead of free text
Color black is recorded in three forms: black, Black and blk. The site filter shows three values, and the color report splits sales into three lines. If you store color code 01 and a label from the list, there is one value, and translations and spelling change in one place. The same applies to size, season and category.
The barcode belongs to the SKU and has a check digit
In a GTIN-13 code the last digit is a check digit. For an illustrative 12 digits 460123456789 we compute: digits from the right multiplied by 3, 1, 3, 1…, sum 117, 3 short of the next ten, so the code is 4601234567893. If someone types 4601234567894, the check catches it before printing. One code for two sizes the check will not catch; that is a reference-level error.
The SKU count grows by multiplication
180 models × 3 colors × 6 sizes × 85% of combinations actually made = 2,754 SKUs, plus 120 sets with their own code, 2,874 codes in all. A card at 6 minutes and 450 roubles an hour gives 2,754 × 6 ÷ 60 ≈ 275 hours and about 124 thousand roubles. Add a fourth color and a seventh size: 180 × 4 × 7 × 0.85 = 4,284 SKUs, 56% more. You pay for assortment width in work as well.
The system of record and the golden record
In the ERP the dress is called Dress D-100 blk, on the site Dress D100 black. By name the records cannot be matched, by the SKU key they match instantly. The rule: a record is edited in one system, the others receive a copy and reconcile by key. Then every report and exchange rests on one version of truth.
How to work through it
- Describe the levels: model, colorway, SKU, pack, set. For each, write what lives on it.
- Define the mandatory attributes on each level and name who fills them in and when.
- Fix the system of record for the card and the order of passing it to other systems.
- Set up lists of categories, colors, sizes, seasons and suppliers with stable codes.
- Give each SKU one GTIN and check the check digit and the absence of duplicates.
- Count SKUs, codes and the labor for cards and compare with the assortment plan.
- Keep the link of carry-over goods to the ancestor model and the rules for when to create a new SKU.
- Load the collection from a master-data template and validate the file before loading, not after.
Where to start
1) Describe the levels: style, color-way, SKU, pack. 2) Define the mandatory attributes at each level. 3) Fix the system of record of the card. 4) Set up the reference books of categories, colors, sizes and seasons. 5) Give each SKU one GTIN and check the check digit. 6) Count the SKUs and the labor of cards. 7) Load a new collection from the master data template. Calculations of the topic in order: “SKUs, barcodes and the labor of product cards”. See also (their primary topic is another one): “The marking code chain: from factory to till”.
Calculations in this topic
SKUs, barcodes and the labor of product cards
From the number of styles, colors and sizes calculates the number of SKUs and GTIN codes, how many cards must be created in a season, how many hours and roubles it costs and what one more color across all styles adds.
How many SKUs and codes the collection will create and what entering them costs.
The marking code chain: from factory to till
Shows the steps of a marking code for your role: what to do, what most often breaks and where costs build up. The steps follow the general code workflow; details for your group: check the rules of the marking system.
Which steps a code passes and what most often breaks.
Self-check questions
- At which level (model, colorway, SKU) does each attribute live and are there no duplicates across levels?
- How many SKUs and GTINs are in the collection and how many hours will cards take?
- Are there duplicate barcodes and one code for two sizes or colors?
- Which system is the master of the card and who may change it?
- Are color and size taken from a list or typed as text?
- Does every code pass the check-digit test before printing and loading?
- How is the link of carry-over goods to the ancestor model stored?
- Which attributes are needed for labeling and for platforms and are they filled?
- When do you create a new SKU and when do you fix an existing one?
Knowledge check
Multiple-choice questions. The correct answer and the explanation follow each question.
At which product level does a GTIN barcode live?
- A. At the style (SKU) level
- B. At the color-way level
- C. At the SKU level: each color and size has its own code
- D. At the season or collection level
Correct answer: C. At the SKU level: each color and size has its own code. A barcode belongs to the SKU, that is, to a style in a specific color and size. A code on a style or color-way would glue the sizes together: stock and sales by size would become indistinguishable, and one code for two sizes breaks the records. A season code does not tell goods apart at all. A set or a carton has its own separate code.
The collection has 40 styles, on average 3 colors and 5 sizes, 80% of combinations are actually made, plus 30 set variants with their own code. How many GTIN codes are needed?
- A. 480
- B. 510
- C. 630
- D. 150
Correct answer: B. 510. SKUs = 40 × 3 × 5 × 0.8 = 480; GTIN = SKUs + sets = 480 + 30 = 510. 480 is the SKUs alone without the sets, 630 comes without the share of combinations made (600 + 30), and 150 is the color-ways (40 × 3 = 120) plus the sets: the sizes are not counted in the codes.
Of the 480 SKUs in the collection 25% are carried over and needing no new card. A card takes 6 minutes, an employee hour costs RUB 450. What does entering the cards cost per season?
- A. RUB 16,200
- B. RUB 21,600
- C. RUB 12,150
- D. RUB 972,000
Correct answer: A. RUB 16,200. New cards 480 × (1 − 0.25) = 360; hours 360 × 6 / 60 = 36; cost 36 × 450 = RUB 16,200. RUB 21,600 comes from creating cards for all 480 SKUs (48 hours); RUB 12,150 from applying the carried-over share twice (270 cards); RUB 972,000 from not dividing the minutes by 60.
The designer proposes adding one more color to all 40 styles; there are 5 sizes, 80% of combinations are made. How many SKUs are added?
- A. 200
- B. 480
- C. 40
- D. 160
Correct answer: D. 160. 40 × 5 × 0.8 = 160 new SKUs: each has its own GTIN and its own card. 200 comes without the share of combinations made; 480 is all the SKUs of the current collection (40 × 3 × 5 × 0.8); 40 is just the number of styles, the sizes are not counted. Besides cards, a color multiplies stock and leftovers, so its depth is counted separately.
The supplier clarified that the composition in the card of an existing SKU is wrong. What should be done?
- A. Create a new SKU with the right composition and close the old one
- B. Fix the attribute in the card of the existing SKU in the owning system and let the change spread to the other systems
- C. Fix the composition in the system where the error was noticed
- D. Leave the card as it is and write the right composition in the order comment
Correct answer: B. Fix the attribute in the card of the existing SKU in the owning system and let the change spread to the other systems. An error in an attribute is a correction of a record, not a new product: the SKU keeps its code, sales history and stock. A new SKU would break the history and split the stock; an edit in any copy gives two truths until the owner is fixed; an order comment does not change the card from which documents and marking are produced.
Common mistakes
Mistake: Confusing the style SKU with the SKU and putting the barcode on the SKU
What it costs: Till and warehouse cannot tell sizes apart
How to avoid: Code at SKU level.
Mistake: Keeping color and size as text
What it costs: One value multiplies and reports and filters fall apart
How to avoid: A list with a stable code.
Mistake: Keeping a card in several systems without a master
What it costs: Versions diverge and edits are lost
How to avoid: One master and scheduled exchange.
Mistake: Not storing the link of carry-over goods to the ancestor model
What it costs: Sales history is lost
How to avoid: An ancestor field and an inheritance rule.
Mistake: Creating a new SKU instead of fixing an attribute error
What it costs: Duplicates and stock mismatches
How to avoid: Edit the record, new SKU only when color or size changes.
Russia and the EAEU
In Russia GTIN codes are issued by GS1 Rus, prefix range 460–469; an imported GTIN is also allowed if legally issued. For groups under mandatory labeling each product variant is described in the National Catalog, and the Data Matrix label code holds the GTIN and a unique serial number; so color and size must be set up as separate cards, otherwise codes will not match documents. In electronic documents (UPD) the same GTINs and names are used, and a master-data mismatch becomes a rejection at receiving.
For import a customs nomenclature code (EAEU CN FEA, ten digits) is needed per model: it determines duty and import VAT (standard rate 22% from 01.01.2026). Fill composition and category so that they match the declaration of conformity. Labeling rules and dates have changed, so check with the system operator and the current text on the upload date (checked 14.11.2026). For export to the EU the digital product passport for textiles is expected by the European Commission’s plan in the fourth quarter of 2027, dates may change; add composition, origin and care fields to the card in advance.
The principle is the same: one SKU, one code, one system of record. The differences are in who issues codes and in which system the product is described for labeling. Reference, check: national GS1 organizations exist in Belarus, Kazakhstan, Armenia and Kyrgyzstan, and country prefixes differ (about 481 for Belarus, 487 for Kazakhstan, 485 for Armenia, 470 for Kyrgyzstan, indicative as of 14.11.2026, check). For deliveries to Russia the customs nomenclature of the union is common, but national catalogs and labeling dates are each country’s own; in Kazakhstan labeling rules are in force from 14.11.2026 and the first groups from 01.12.2026, reference, check the text of the order. The card in the master system should be able to store several catalogs and a description language.
Categories and audiences
- Apparel
- Here the most common matter is model, color, size. The size chart should be a list with a mapping of RU, EU and international labels (XS–XL), otherwise filters and platforms show different sizes. Composition and care are better held at model level, color and photos at colorway level.
- Footwear
- In footwear the size is set not only by a number but by the scale (EU, RU, Mondopoint) and width, and often a pair counts as the unit. One size in different scales must not become two SKUs. Last and materials of upper, sole and lining matter and are stored as separate attributes.
- Accessories
- Many accessories have no size (one size), so the SKU equals the colorway; but there are belt length, bag volume and contents. A gift or assembled set needs its own code and a link to its parts. Materials and customs codes are varied, and storing them carefully matters especially.
- Kids
- For children’s goods size follows height and age, and the child grows, so the chart must store height, age and their mapping. Safety requirements and the certificate are tied to the model and the batch. Labeling and certification both need exact age and composition attributes.
- Men
- The men’s master data needs two things: a link to the predecessor model and a size pair (collar and height, waist and length). Without the link the sales history of a basic shirt breaks when the SKU changes, and without the pair trousers of different lengths are counted as different items or, conversely, merged. The reference must know how a size is stored.
- Women
- In the women’s line the master data needs a size-conversion table: Russian numeric 42–52, letter and European. Without it one fit lands in reports under different sizes, and stock on the site and in the store diverges. Card discipline (who edits, when it is archived) decides whether season reports are comparable.
Questions this topic answers
- At which level (style, color-way, SKU) does each attribute live?
- How many SKUs and GTIN codes are in the collection and what does their entry cost?
- Are there duplicated barcodes or one code on two sizes?
- Which system is the system of record of the product card?
- How to maintain color-way and size so that attributes do not drift between systems?
Key measures and terms
Study first
What next
Next: “Data quality and data owners”, “Systems landscape and data exchange”, “Marking and trading leftovers”.
Documents on this topic
- AI in fashion analytics and content: what can be computed, limits, risks
- Fashion business IT architecture
- Systems and data: 1C and ERP, PIM and MDM, BI, integrations (e-document exchange, marking, marketplace API), cost of ownership and data quality
- Customers’ personal data: Federal Law 152-FZ for a brand and a retailer. Grounds and consents, storage, localization, loyalty, mailings, liability
- A data cube for planning: measures, dimensions, grain and the season calendar