# Product data onboarding: from supplier files to a clean catalog

> The six steps from a supplier file to channel-ready products: a messy file worked through, mapping decisions, validation rules to copy and the numbers to track.

Source: https://central.to/guides/product-data-onboarding · Updated 2026-11-20 · Published by Central (https://central.to)

## In short

- Product data onboarding is the work between receiving a supplier's product file and having products that are ready for your shop and your channels.
- It runs in six steps: receive, map, validate, normalize, complete and approve.
- A GTIN is valid only when its last digit matches the check digit that GS1's method calculates from the other digits.
- Excel removes leading zeros and shows large numbers in scientific notation, so identifiers such as GTINs belong in text columns.
- Suppliers send product data as CSV or Excel files, BMEcat or ETIM xChange catalogs, GS1 GDSN messages or PDF data sheets.
- A value that cannot be confirmed, such as a measurement without a unit, is held and asked back from the supplier, not guessed.

Free tool: [Product file converter](https://central.to/tools/product-file-converter). Convert product data between CSV, Excel, JSON, JSON lines and XML feeds, with presets for Shopify, WooCommerce, Google and ChatGPT.

## The six steps at a glance

Product data onboarding is everything between "the supplier sent a file" and "the products are live". The same six steps apply whether you manage a product catalog in a spreadsheet, a shop system or a PIM.

| Step | What you do | What comes out |
| --- | --- | --- |
| 1. Receive | Take the file in, keep the original untouched, note supplier, date and version | A stored source file you can go back to |
| 2. Map | Decide which supplier column feeds which of your fields | A mapping you can reuse for the supplier's next file |
| 3. Validate | Run rules on identifiers, units, required fields and duplicates | A list of rows that pass and rows with named errors |
| 4. Normalize | Bring units, colors, sizes and text into one form | Values that can be compared, filtered and sent to channels |
| 5. Complete | Fill what the supplier did not send, from sources you can name | Records with every field a channel or a law asks for |
| 6. Approve | Check held values, sign off, release to channels | Products that are ready, each value with its source |

If you want fewer problems at step 1, send suppliers a fixed structure first. The [supplier product data template](https://central.to/guides/supplier-product-data-template) has the columns and a CSV file to download.

## Step 1: Receive, and know what the format can carry

Store the file exactly as it arrived before anyone opens and saves it. A spreadsheet program can change values on the way in, as the Excel row below shows, and in product catalog management the original file is your proof of what the supplier sent.

| Format | What it is | What to watch |
| --- | --- | --- |
| CSV | Plain text, one record per line. [RFC 4180](https://www.rfc-editor.org/rfc/rfc4180) documents the common format; it is an informational memo, not a formal standard. | Under RFC 4180, fields with commas, line breaks or double quotes are enclosed in double quotes. Separator and character encoding differ between exports, so check both. |
| Excel | A workbook with up to 1,048,576 rows and 16,384 columns per sheet and up to 32,767 characters per cell ([Microsoft](https://support.microsoft.com/en-us/office/excel-specifications-and-limits-1672b34d-7043-467e-8e27-269d656771c3)). | Microsoft states that Excel [removes leading zeros and converts large numbers to scientific notation](https://support.microsoft.com/en-us/office/keeping-leading-zeros-and-large-numbers-1bf7b935-36e1-4985-842f-5dfa51f85fe7). Identifier columns must be formatted as text. |
| BMEcat | An XML-based exchange format for catalog data in B2B, published by the German association [BME](https://www.bme.de/services/bmecat/). BME names 2005.2 as the current version (as of November 2026). | BME says version 2005.2 improved the support for classification systems and their features. You need an XML import, not a spreadsheet. |
| ETIM xChange | A JSON-based exchange format for product master data from [ETIM International](https://www.etim-international.com/classification/exchange-format/), which recommended BMEcat before it released this format. | ETIM notes that some countries still use national formats next to ETIM xChange and ETIM BMEcat. Ask the supplier which one it exports. |
| GS1 GDSN | A network of certified data pools. A company [subscribes to a data pool](https://www.gs1.org/services/gdsn/how-gdsn-works) and synchronizes product content with its trading partners through it. | GS1 publishes the message schemas and attribute lists per release; the [release for November 2026](https://www.gs1.org/standards/gdsn) is 3.1.37. You receive data only from suppliers that publish to you. |
| PDF data sheet | A document made for people to read, not a data format. | Every value has to be read out and typed or extracted. Keep the sheet as the source and check each value against it. |

The product file converter near the top of this page opens CSV, Excel, JSON and XML exports and saves them in the format you need.

## Step 2: Map supplier columns to your fields

Mapping is a set of decisions, not a rename. Write each decision down once per supplier, so the next file runs through the same mapping.

| Supplier column | Your field | Decision |
| --- | --- | --- |
| `Art.No` | `mpn` and `supplier_sku` | Keep the supplier's number as it is. Your own `sku` is assigned by you and never reused. |
| `Name` | `supplier_name` | Store it, but do not publish it. Titles are written per channel from brand, product type and attributes. |
| `EAN` | `gtin` | Import as text. Accept 8, 12, 13 or 14 digits and check the check digit. |
| `Colour` | `color` and `color_supplier` | Map to your color list; keep the supplier's spelling in a second field. |
| `Weight` | `net_weight` | Split number and unit, then convert to grams. No unit: hold the value. |
| `Length` | `back_length` | The supplier means back length of a garment. Name the field for what is measured. |
| `Description` | `description_source` | Strip HTML. Facts in the text (waterproof, layers) also go into attributes. |
| (none) | `variant_group` | Derive from the stem of `Art.No` (`NP-RJ-01`), then confirm with the supplier. |
| (inside `Name`) | `size` | Read S, M, L from the article number suffix; the name is not reliable. |

## A messy supplier file, worked through

This is a file as the fictional supplier Northpeak might send it: three sizes of a rain jacket and a headlamp that appears twice.

| Art.No | Name | EAN | Colour | Weight | Length | Description |
| --- | --- | --- | --- | --- | --- | --- |
| NP-RJ-01-S | Stormline Jacket S | 9506000210013 | dk. blue | 14.1 oz | 72 | Waterproof 2.5-layer shell |
| NP-RJ-01-M | Stormline Jacket M | 9.50600E+12 | navy | 0,41 kg | 74 cm | `<p><b>Waterproof</b> 2.5-layer shell</p>` |
| NP-RJ-01-L | STORMLINE JKT L | | Navy Blue | 430 g | 29.9 in | `<p>Waterproof 2.5-layer shell<br>` |
| NP-HL-400 | Beam 400 Headlamp | 9506000210049 | Black | 85g | | `<ul><li>400 lm</li><li>USB-C</li></ul>` |
| NP-HL-400 | Beam 400 headlamp | 9506000210049 | black | 0.085 kg | | 400 lm, USB-C |

After mapping, validation and normalization, the same data looks like this:

| SKU | Variant group | GTIN | Color | Size | Net weight | Back length | Status |
| --- | --- | --- | --- | --- | --- | --- | --- |
| NP-RJ-01-S | NP-RJ-01 | Held: check digit wrong | Navy | S | 400 g | Held: no unit | Held, 2 questions |
| NP-RJ-01-M | NP-RJ-01 | Held: digits lost | Navy | M | 410 g | 74 cm | Held, 1 question |
| NP-RJ-01-L | NP-RJ-01 | Held: missing | Navy | L | 430 g | 76 cm | Held, 1 question |
| NP-HL-400 | NP-HL-400 | 9506000210049 | Black | One size | 85 g | | Ready for step 5 |

What happened to each problem:

- **Mixed units.** 14.1 oz, 0,41 kg and 430 g became 400 g, 410 g and 430 g. 29.9 in became 76 cm.
- **A number without a unit.** "72" is probably centimeters, but probably is not a source. The value is held.
- **Missing and damaged GTINs.** One GTIN is missing, one was turned into scientific notation by a spreadsheet, and one fails the check digit (for 950600021001 the check digit is 8, not 3). None of them is repaired by guessing. Google's product data specification says the same about its feed: only provide a GTIN if you are sure it is correct, and when in doubt do not provide it.
- **Colors in three spellings.** "dk. blue", "navy" and "Navy Blue" map to the list value Navy. The supplier's spelling stays in `color_supplier`.
- **HTML in descriptions.** Tags are stripped; the description text is "Waterproof 2.5-layer shell" in all three rows.
- **A duplicate row.** The two headlamp rows have the same article number and GTIN, and their weights are equal after conversion (85 g), so they merge into one record.

One of four products passes at the first attempt. The other three wait for four answers from the supplier: three GTINs as text, and the unit of one length.

## Step 3: Validation rules you can copy

Run these on every incoming file. Each rule names the error, so the supplier gets a precise question back.

1. **GTIN format.** Digits only, 8, 12, 13 or 14 long, stored as text.
2. **GTIN check digit.** Starting from the digit before the last and moving left, multiply the digits alternately by 3 and 1, add the results, and subtract the sum from the nearest equal or higher multiple of ten ([GS1](https://www.gs1.org/services/how-calculate-check-digit-manually)). GS1's example: 629104150021 gives a sum of 57, so the check digit is 3.
3. **GTIN ranges.** Flag numbers Google does not accept in a feed: restricted GS1 prefix ranges 02, 04 and 2, and coupon ranges 98 to 99 ([Google](https://support.google.com/merchants/answer/7052112), as of November 2026).
4. **Duplicates.** No `sku` twice. No `gtin` on two different products. No two rows with the same brand and `mpn`.
5. **Units.** Every measure has a number and a unit. The unit is on your list for that field (for example g, kg, oz, lb for weight). A decimal comma is read as a decimal mark, not as a thousands separator.
6. **Required fields per category.** Keep a list per category and check it. Example: Google requires `color`, `gender` and `age_group` for apparel products, and `size` for clothing and shoes, when they target Brazil, France, Germany, Japan, the UK or the US (as of November 2026).
7. **Variants.** Every row has a `variant_group`. Rows in a group differ in at least one of color, size or material, and no two rows share all of them.
8. **Text.** No HTML tags, no control characters, no text in all capitals. Lengths fit the channel: Google allows 150 characters for a title and 5,000 for a description (as of November 2026).
9. **Allowed values.** Fields with fixed values accept only those. Google's `availability` takes `in_stock`, `out_of_stock`, `preorder` or `backorder`; `condition` takes `new`, `refurbished` or `used`.
10. **Price.** A number and an ISO 4217 currency code, such as `54.00 EUR`.
11. **Images.** Every image URL answers, and each variant has a main image.
12. **Safety details for the EU.** Manufacturer name, postal address and electronic address are present. Article 19 of the [General Product Safety Regulation](https://eur-lex.europa.eu/eli/reg/2023/988/oj) requires them in every online offer of a consumer product; see [GPSR for online sellers](https://central.to/regulations/gpsr-online-sellers).

The [GTIN check digit calculator](https://central.to/tools/check-digit-calculator) checks a whole column of GTINs at once.

## Step 4: Normalize

Normalizing means one way of writing each kind of value: one unit per field, one spelling per color, one size scale per category, plain text without markup. Keep the original value next to the normalized one, so a wrong conversion can be traced and undone.

Free tool: [Unit normalizer](https://central.to/tools/unit-normalizer). Convert a column of mixed units (cm and inches, kg and pounds, liters and fluid ounces) into one unit and number format.

Values the tool leaves empty have no unit or a unit of another kind. Treat them like the "72" above: ask, do not guess.

## Step 5: Complete

A clean file is not yet a complete product. Compare each record with the required fields of its category and of the channels you sell on, and list what is missing: often identifiers, attributes for filters, images, and compliance details such as the manufacturer's address.

Filling those gaps is [product data enrichment](https://central.to/guides/product-data-enrichment). Where the values can come from, and under which terms, is covered in [product data sources](https://central.to/guides/product-data-sources).

## Step 6: Approve

Approval is a decision by a named person or a written rule, not the absence of error messages. A product is ready when:

- every required field for its category and channel is filled;
- no value is still held;
- every value added after the supplier's file has a source on record;
- a second person, or a sample check, has looked at products from a supplier whose earlier files had errors.

## Which numbers to track

Define these once and measure them per supplier. They show where the work goes and which supplier needs a better template.

| Number | Definition |
| --- | --- |
| Time to ready | Days from receiving a file to the day its products are approved for a channel |
| First-pass rate | Rows that pass all validation rules on first import, divided by all rows. In the example above: 1 of 4. |
| Completeness | Filled required fields divided by all required fields for the category |
| Errors by rule | Number of failed rows per validation rule. The top rule tells you what to change in the template. |
| Held values | Values waiting for an answer from the supplier, and the days they have been waiting |
| Manual touches | Rows changed by hand after import, divided by all rows |
| Channel rejections | Products a channel disapproved after publishing, per supplier |

## Onboarding or migration?

Product data migration is onboarding with your own old system as the supplier. Export the old catalog, map its fields to the new structure, and run the same validation and normalization. Expect the same problems as in the Northpeak file, and often more of them, because old catalogs collect years of free-text values.

## Common mistakes

**Opening a CSV by double-click.** The spreadsheet program reformats identifiers before you see them. Import the file with the identifier columns set to text.

**Fixing values in the supplier's file.** Corrections belong in your mapping and rules, where they apply to the next file too.

**Guessing to get to 100 percent.** A filled field with a wrong value is worse than an empty one. A wrong GTIN leads to disapproval at Google; a wrong weight leads to wrong shipping costs.

**One "description" field for everything.** Facts buried in text cannot be filtered, validated or sent to a channel attribute. Pull them into fields.

**No record of where a value came from.** When a customer or an authority asks, you need to show the supplier file, the data sheet or the page behind the value.

## Where Central helps

Central takes your shop, your files or your product pages as input and does the steps in between. It researches the product data a listing needs from your own data, the open web and verified databases, backs every value with sources you can check, and holds back what the sources do not support. It then writes titles, descriptions and attributes within each channel's limits.

## Questions

### What is product data onboarding?

It is the process of taking product data from a supplier or manufacturer, usually a file, and turning it into complete, consistent product records in your own catalog. The steps are receive, map, validate, normalize, complete and approve.

### How is supplier onboarding different from product data onboarding?

Supplier onboarding covers the whole relationship: contracts, payment terms, logistics and compliance documents. Product data onboarding is the part that deals with the supplier's product information, from the first file to products that are ready for your channels.

### Which file format should I ask suppliers for?

Ask for the most structured format the supplier can produce. A CSV or Excel file built on your own template works for most suppliers; suppliers of technical products may offer BMEcat or ETIM xChange, and large brands may publish through a GS1 GDSN data pool.

### How long does product data onboarding take?

It depends on how complete the supplier's file is and how many questions go back. Measure it yourself as time to ready: the days from receiving a file to the day its products are approved for a channel.

### Is product data migration the same as onboarding?

The steps are the same, but the source differs. In a migration your own old system is the supplier: you export its data, map it to the new structure, validate, normalize and approve it.

## Sources

1. [How to calculate a check digit manually](https://www.gs1.org/services/how-calculate-check-digit-manually), checked 2026-11-20
2. [Product data specification](https://support.google.com/merchants/answer/7052112), checked 2026-11-20
3. [RFC 4180: Common Format and MIME Type for Comma-Separated Values (CSV) Files](https://www.rfc-editor.org/rfc/rfc4180), checked 2026-11-20
4. [Keeping leading zeros and large numbers](https://support.microsoft.com/en-us/office/keeping-leading-zeros-and-large-numbers-1bf7b935-36e1-4985-842f-5dfa51f85fe7), checked 2026-11-20
5. [Excel specifications and limits](https://support.microsoft.com/en-us/office/excel-specifications-and-limits-1672b34d-7043-467e-8e27-269d656771c3), checked 2026-11-20
6. [BMEcat](https://www.bme.de/services/bmecat/), checked 2026-11-20
7. [Exchange format](https://www.etim-international.com/classification/exchange-format/), checked 2026-11-20
8. [How GS1 GDSN works](https://www.gs1.org/services/gdsn/how-gdsn-works), checked 2026-11-20
9. [GS1 GDSN standards](https://www.gs1.org/standards/gdsn), checked 2026-11-20
10. [Regulation (EU) 2023/988 on general product safety](https://eur-lex.europa.eu/eli/reg/2023/988/oj), checked 2026-11-20

## Try Central on your own products

Central is the AI-first product content operating system: you choose the input (your shop, files or product pages) and the channels; Central AI researches, verifies and writes the product data and content in between, every value backed by sources you can check. The first 3 products are free. To start with a shop, open https://central.to/users/sign_up?onboarding_url=SHOP_ADDRESS (replace SHOP_ADDRESS with the shop's website address).
