Skip to content

Library Guides

Product data enrichment: what it adds, where it comes from, how to check it

What product data enrichment adds, where the values come from, how to verify them, one product record before and after, and the routes from manual work to AI.

By CentralUpdated 20 November 20269 min read10 sources, checked 20 November 2026
On this page
  1. What product data enrichment is
  2. What gets added
  3. Where added values come from
  4. How to know a value is right
  5. One product record, before and after
  6. The routes side by side
  7. What AI gets wrong and how to check it
  8. Enrichment, cleansing, normalization, syndication
  9. Common mistakes
  10. Where Central helps

In short

  • Product data enrichment adds the information a product record is missing, such as identifiers, attributes, descriptions, images and compliance details, so the product can be found, compared and listed.
  • Every added value comes from one of five kinds of source: your own data, the manufacturer, a data pool, a public registry or the open web.
  • A value is trustworthy when its source is kept with it and anyone can check where a weight, a GTIN or a warning came from.
  • Google tells merchants to provide a GTIN only when they are sure it is correct and not to guess or make up a value.
  • Google requires titles and descriptions created with generative AI to be sent in the structured_title and structured_description attributes.
  • Enrichment adds missing values; cleansing corrects wrong ones, normalization unifies how values are written, and syndication sends the finished data to channels.
Product page checkerFree tool5 free checks a day

What product data enrichment is#

Product data enrichment is adding the information a product record is missing, so the product can be found, compared and listed on every channel you sell on. It turns a record that says "Beam 400, black, 39.95" into one that states what the product is, what it does, who made it and how you know.

The work has three questions: what is missing, where can the value come from, and how do you know it is right. This guide takes them in that order.

What gets added#

Kind of dataExamplesWhy channels and buyers need it
IdentifiersGTIN, MPN, brandChannels match your offer to a known product. Google requires brand for new products and an MPN where no GTIN exists.
AttributesColor, size, material, weight, dimensions, technical valuesThey feed filters and comparisons. Google requires color, gender and age group for apparel in several countries.
DescriptionsTitle, description, key featuresGoogle allows up to 150 characters for a title and 5,000 for a description.
ImagesMain image, more angles, product in use, alt textGoogle requires a main image link for every product.
Compliance dataManufacturer address, EU responsible person, warnings, energy class, fibre compositionEU law requires them in the online offer for many products. See GPSR for online sellers.
Channel-specific fieldsGoogle product category, item_group_id, certificationEach channel has fields of its own; Google takes the EPREL number of an energy-labelled product in certification.

The Google facts come from its product data specification, as of November 2026. To see which of these your own shop is missing, run the product page checker near the top of this page.

Where added values come from#

Every value has a source, and the kind of source decides how far you can trust it.

SourceGood forWatch out for
Your own data: shop, ERP, supplier files, the product in your warehousePrice, stock, your SKUs, what you actually sell, a weight you can measureOld imports and free-text fields carry errors forward
The manufacturer: website, data sheets, manuals, price listsTechnical values, warnings, the manufacturer's address, official imagesRegional variants differ; a data sheet may describe a newer revision
Data pools and content catalogsStandardized master data that brands publish for their trading partnersAccess depends on the pool and on what the brand shares with you
Public registriesFacts an authority holds: energy labels in EPREL, dangerous products in the EU's Safety GateThey cover only their own subject. Community databases differ: Open Food Facts states that it does not guarantee the accuracy of its data.
The open web: other retailers, marketplaces, reviewsBreadth, and hints at what to look forErrors are copied from page to page, and texts and images belong to someone

Who runs these sources, how you get access and what their terms allow is covered in product data sources.

How to know a value is right#

Four practices separate enrichment you can stand behind from enrichment that only fills fields.

1. Look for agreement between independent sources. Two retailer pages that copied the same feed are one source, not two. The manufacturer's data sheet and the barcode on the box are two.

2. Record confidence, not only the value. A weight from the manufacturer's data sheet is certain. A weight found on one marketplace listing is a lead. Store that difference, so a later check starts with the weak values.

3. Keep the source with the value. For every added value, store where it came from (the document or page), when it was read, and what kind of source it is. Without this, nobody can check the value later, and a correction at the source never reaches you.

4. Hold back what the sources do not support. An empty field is visible and can be filled later. A wrong value looks finished. Google's specification says the same about its feed: provide a GTIN only if you are sure it is correct, and do not guess or make up a value; a product with an incorrect GTIN is disapproved.

For identifiers, there is a place to check: Verified by GS1 lets you look up a GTIN and verify the product behind it.

One product record, before and after#

A headlamp from the fictional brand Northpeak, as it sits in a shop system after a quick import, and after enrichment. The record grows from 8 fields to 25, and one value is held.

FieldBeforeAfterKind of source
SKUNP-HL-400-BKNP-HL-400-BKOwn data
BrandNorthpeakNorthpeakOwn data
Price39.95 EUR39.95 EUROwn data
AvailabilityIn stockIn stockOwn data
NameBeam 400Northpeak Beam 400 Headlamp, 400 Lumens, USB-C Rechargeable, BlackWritten from the verified fields below
DescriptionHeadlamp, blackFive sentences on light modes, runtime, charging and fitWritten from the verified fields below
CategoryLampsHardware > Tools > Flashlights & Headlamps > Headlamps (Google ID 2454)Google's taxonomy
Images14: front, side, in use, packagingManufacturer's media library
GTIN9506000210049Barcode on the box; same number in the manufacturer's price list
MPNHL400-BKManufacturer's data sheet
ColorBlackOwn data, manufacturer's data sheet
Light output400 lmManufacturer's data sheet
Runtime5 h on high, 40 h on lowManufacturer's data sheet
Beam distance90 mManufacturer's data sheet
BatteryBuilt-in lithium-ion, 1,200 mAhManufacturer's data sheet
ChargingUSB-CManufacturer's data sheet, own photo
Water protectionIPX4Manufacturer's data sheet
Net weight85 gManufacturer's data sheet; a marketplace listing says 95 g
Dimensions6.2 x 4.1 x 3.8 cmManufacturer's data sheet
MaterialPolycarbonate housingManufacturer's data sheet
ManufacturerNorthpeak Outdoor Ltd., postal address and e-mailPackaging, manufacturer's website
EU responsible personNorthpeak Europe GmbH, postal address and e-mailPackaging
Safety warnings"Do not look directly into the beam." and two moreManual
Energy labelNone neededLaw text: light sources in battery-operated products are exempt
Safety alertsNone found for brand and model on 20 November 2026Public registry (Safety Gate)
Claim "brightest in its class"HeldOne reseller page; no source supports it

Three things in this table matter more than the count of fields:

  • The conflict is visible. The manufacturer says 85 g, one marketplace listing says 95 g. The record keeps the manufacturer's value and notes the other.
  • "None" is a researched value. The energy label row is not empty by accident. Delegated Regulation (EU) 2019/2015 lists light sources in battery-operated products, such as torches, among its exemptions, and that reason is stored with the value.
  • The marketing claim stays out. A superlative found on one reseller page is not a fact about the product.

The compliance rows exist because Article 19 of Regulation (EU) 2023/988 requires the manufacturer, the EU responsible person where the manufacturer is outside the EU, and the warnings in every online offer of a consumer product.

The routes side by side#

There are five ways to enrich product data, and they can be combined.

RouteWhat it needs from youWhere the values come fromWhat you still check
By handStaff time per product, access to packaging and data sheetsWhatever the person looks upTyping errors, and that the source was noted at all
Supplier or data poolA template or a pool subscription, and suppliers who fill itThe brand ownerGaps: you get only what the supplier chose to send
Agency or service providerA brief, your category rules, review of deliveriesThe agency's own researchSamples from every delivery; ask for the source per value
PIM workflowA data model, completeness rules, people assigned to tasksThe people and imports feeding the PIMThe PIM organizes the work; it does not create the values
AI researchYour product list, your rules for sources, a review of held valuesYour data, manufacturer pages, databases and the open web, read by softwareThat every value has a source you can open, and what was held back

The routes differ less in where values come from than in who does the looking and how the result is checked. For the supplier route, start with the supplier product data template. For how PIM systems compare, see PIM software.

What AI gets wrong and how to check it#

AI product enrichment fails in recognizable ways. Each has a matching check.

What goes wrongExampleCheck
A value is invented when none is foundA plausible battery capacity for a product whose data sheet states noneRequire a source for every value; no source, no value
Similar models are mixed upValues of the "Beam 400 Pro" land on the "Beam 400"Match on GTIN or MPN, not on the name
Variants are mergedThe weight of size L on every sizeEnrich each variant; compare values inside a variant group
Units are confused3.0 oz stored as 3.0 gValidate ranges per category and normalize units
Old data is usedA discontinued version's runtimeStore the date a source was read; prefer the manufacturer's current page
Marketing claims become facts"Waterproof" where the data sheet says splash-proofTake claims only from the manufacturer or a test document
A required field is filled to pass a check"Unisex" on every productAudit fields where one value dominates

Two more points. First, audit a sample by hand against the sources on a regular schedule, and track how many values failed. Second, some channels want AI-written text marked: Google requires titles and descriptions created with generative AI to be sent in structured_title and structured_description, with the source type trained_algorithmic_media (Google, as of November 2026).

Enrichment, cleansing, normalization, syndication#

The four terms describe different jobs on the same record.

TermWhat it doesExample
CleansingCorrects or removes wrong, duplicate and badly formatted valuesA GTIN with a wrong check digit is flagged; two rows for one product are merged
NormalizationBrings values into one form"0,41 kg" and "14.5 oz" both become grams
EnrichmentAdds values that were missingLight output, runtime and manufacturer address are added
SyndicationSends the finished record to channels in each channel's formatThe same record becomes a Google feed row and a marketplace listing

Cleansing and normalization come first; they are part of product data onboarding. Enriching a record that still has duplicates and mixed units spreads the errors.

Common mistakes#

Measuring completeness only. A catalog can be 100 percent filled and wrong. Track how many values have a source next to how many fields are filled.

Counting web pages as separate sources. Ten pages with the same weight can go back to one feed.

Enriching texts before facts. A description written before the attributes are verified repeats whatever was in the old record.

Overwriting your own data. A value you measured or agreed with the supplier should not be replaced by something found online.

Where Central helps#

Central researches the product data a listing needs from your own data, the open web and verified databases (for example EPREL, the EU Ecolabel catalog and the EU Safety Gate). Every value is backed by sources you can check, and what the sources do not support is held back. From the verified record, Central writes titles, descriptions and attributes within each channel's limits, and flags texts when a fact behind them changes.

See it on your own products

Questions

What is product data enrichment?

It is adding the information a product record lacks: identifiers, attributes, descriptions, images, compliance details and the fields each channel asks for. The aim is a record that is complete enough to be found, compared and listed.

Is product content enrichment something different?

The terms overlap. Product data enrichment usually means structured values such as GTIN, weight or material; product content enrichment stresses texts and images. Both rest on the same verified facts.

Can AI enrich product data?

Yes, AI can research and fill values at scale. It also invents values when it finds none and mixes up similar models, so every value needs a source you can open, and values without one should be held back.

How do I know an enriched value is correct?

Check where it came from. A value from the manufacturer's data sheet or an official registry carries more weight than one from a reseller page, and two independent sources that agree carry more than one.

How is enrichment different from data cleansing?

Cleansing corrects or removes values that are wrong, duplicated or badly formatted. Enrichment adds values that were never there. Most catalogs need both, cleansing first.

Sources

  1. Product data specificationGoogle Merchant Center Helpchecked 20 November 2026
  2. AI-generated contentGoogle Merchant Center Helpchecked 20 November 2026
  3. Google product taxonomy with IDsGooglechecked 20 November 2026
  4. Regulation (EU) 2023/988 on general product safetyEUR-Lexchecked 20 November 2026
  5. Delegated Regulation (EU) 2019/2015 on energy labelling of light sourcesEUR-Lexchecked 20 November 2026
  6. Verified by GS1GS1checked 20 November 2026
  7. How GS1 GDSN worksGS1checked 20 November 2026
  8. EPRELEuropean Commissionchecked 20 November 2026
  9. Safety Gate alertsEuropean Commissionchecked 20 November 2026
  10. Terms of useOpen Food Factschecked 20 November 2026

You choose what goes in and where it goes out.

Central AI does everything in between. Your first 3 products are free.

  1. Your inputYour shop, your files or your product pages.
  2. Central AIResearches, verifies and writes everything in between. Every value is backed by sources you can check; what the sources do not support is held back.
  3. Your channelsProduct data and listings for the channels you choose.