Guide Catalog operations
Product taxonomy mapping: how to do it and how to measure it.
Taxonomy mapping connects the categories sellers use to the categories your marketplace or retailer uses. It usually works in layers: a model proposes the best node from the title, attributes and image, and rules or a reviewer handle edge cases. This Youzu guide covers common standards, mapping patterns and how to measure accuracy.
Published by Youzu · Editorial policy
- Published
- Reading time
- 9 min read
Key points
- Map each source category to a leaf of your own tree, not to a parent node.
- Use standards such as the Google product taxonomy or GS1 GPC as references and channel targets.
- Send catch-all and ambiguous products to a per-product decision instead of a category rule.
- Check mappings on a labeled sample, and report critical errors separately from ordinary ones.
- Keep the mapping current when sellers, channels or your own tree change.
In this guide
What a taxonomy is
A product taxonomy is the category tree behind a catalog. Broad nodes such as Home sit at the top, narrower nodes such as Living room below them, and leaves such as Sofas or Armchairs at the bottom. Products live in leaves. Each leaf carries the attributes that matter for its products: a sofa has seats, width and material, a jacket has size, color and closure.
The tree does 3 jobs at once. It decides which filters a shopper sees, which attributes a seller is asked for and how sales and stock are reported. A product placed in the wrong leaf, or left in a parent node, loses all 3: it shows up under the wrong filters, it is asked for the wrong attributes and it is counted in the wrong line of the report.
Taxonomy mapping is the work of connecting another category structure to yours. On a marketplace, that usually means mapping each seller's own categories to your tree. For a retailer, it often means mapping the internal tree to the categories each sales channel requires.
Standards in use (Google product taxonomy, GS1 GPC, UNSPSC)
3 public classifications come up most often. None of them is meant to replace the tree your shoppers browse; they are references to borrow structure from and targets that some channels and partners require.
- Google product taxonomy. The category list Google Merchant Center uses, published as a downloadable file with an ID per category.[2] Google sorts products from broad categories down to specific ones and asks merchants to choose the most specific category that fits.[1]
- GS1 Global Product Classification (GPC). A classification for trade items maintained by GS1, the body behind GTINs. Its lowest level groups products with the same characteristics and attaches attributes to them.[3]
- UNSPSC. The United Nations Standard Products and Services Code, a hierarchy used mainly in procurement and spend reporting rather than shopper browsing.[4]
This guide cites these standards rather than copying them. Download each one from its publisher, note the version you used and record the date, because all 3 are revised over time.
Mapping patterns: one-to-one, one-to-many, many-to-one
Most mapping work falls into 3 patterns. Knowing which pattern a source category follows tells you whether a rule is enough or whether each product has to be looked at.
| Pattern | Source category | Target in your tree | How to map it |
|---|---|---|---|
| One-to-one | A seller's 'Sofas' | Home › Living room › Sofas | A category rule covers every product |
| One-to-many | A seller's 'Home / Misc' | Sofas, Armchairs, Coffee tables or Side tables | Decide per product from title, attributes and photo |
| Many-to-one | A seller's 'Couches' and 'Settees' | Home › Living room › Sofas | Several rules point to the same leaf |
One-to-one and many-to-one categories can be mapped once and reused. One-to-many categories cannot: the source category does not carry enough information, so the decision moves to the product. Catch-all categories such as 'Home / Misc' are the extreme case, and they are where most mapping errors start.
Models, rules and reviewers
Mapping usually works in layers. Category rules handle the clean cases. A model proposes a ranked list of leaves for the rest, reading the title, the attributes and the image. Reviewers decide the cases neither can settle and maintain the rules over time.
Ranked suggestions matter more than a single answer. For a seller title such as 'jkt olive', a sensible ranking is Jackets, then Coats, then Vests, each with a short reason a reviewer can check: the title word 'jkt', and a photo of a short garment with buttons. A reviewer who sees the reason can accept the first suggestion quickly or pick the second knowing why.
The worked example below shows 8 products sorted into an example tree. 7 have one clear leaf. The eighth, a plum lounge seat, fits Armchairs or Sofas equally well, so it goes to review instead of being forced into one.
Worked example
8 products mapped to an example tree
Example category tree
- Home (not a leaf)
- Living room (not a leaf)
- Sofas
- Armchairs
- Coffee tables
- Side tables
4 leaves. Products are placed in leaves only.
| Product | Leaf | Why |
|---|---|---|
| Linen sofa | Home › Living room › Sofas | The material is linen and the seat is long with a back and arms, from the title and the photo. |
| Velvet sofa, navy | Home › Living room › Sofas | The material is velvet and the seat is long with a back and arms, from the title and the photo. |
| Grey 2-seat sofa | Home › Living room › Sofas | The title says 2-seat and the photo shows 2 seat cushions. |
| Terracotta armchair | Home › Living room › Armchairs | The seat is for one person and has arms, from the title and the photo. |
| Accent chair, plum | Home › Living room › Armchairs | The seat is for one person with a high back, from the title and the photo. |
| Oak coffee table | Home › Living room › Coffee tables | The table is low and wide, which suits the middle of a seating group, from the title and the photo. |
| Oak side table | Home › Living room › Side tables | The table is small and sits beside a seat, from the title and the photo. |
| Plum lounge seat | Home › Living room › Armchairs or Home › Living room › SofasSent to review | The title and the photo fit 2 armchairs or one small sofa. |
Mapping is the same decision made for a supplier's category instead of for one product. Try it below on 5 supplier strings: 2 meet in one leaf, one needs 2 leaves and is decided per product, and the catch-all goes to review.
Worksheet
Map 5 supplier categories to your tree
Many-to-oneHome › Living room › Sofas
Many-to-oneHome › Living room › Sofas
One-to-many: decide per product
One-to-oneHome › Lighting › Floor lamps
Unmapped: goes to review
Supplier categories and the leaves they map to Supplier category Target category Mapping Couches - Home › Living room › Sofas
Many-to-one Settees - Home › Living room › Sofas
Many-to-one Living room tables - Home › Living room › Coffee tables
- Home › Living room › Side tables
One-to-many: decide per product Standing lamps - Home › Lighting › Floor lamps
One-to-one Home / Misc None Unmapped: goes to review
Read as text
Edge cases
- Products that fit 2 leaves. A lounge seat that is both a large armchair and a small sofa. Pick a rule per category, write it down and send exceptions to review.
- Bundles and sets. A sofa sold with a matching armchair. Decide whether sets get their own leaf or follow their main item, and apply it everywhere.
- Accessories. Covers, cushions and legs sold for a sofa belong under accessories, not under sofas, even when the title starts with 'sofa'.
- Parent-only sources. A seller who only ever sends 'Home'. Treat it as no category at all and decide per product.
- Seasonal or campaign categories. 'Gifts' or 'Sale' are merchandising lists, not taxonomy nodes. Map the product to its real leaf and keep campaigns as tags.
How to measure: sample design, error classes, critical errors
A mapping that has never been measured is a guess. Measure it on a labeled sample before trusting it, and again whenever the tree or the sources change.
- Design the sample. Draw it at random from real traffic, then add the difficult categories on purpose: catch-alls, near neighbors and categories with policy rules. Keep the 2 parts apart in the report.
- Label it independently. Let people who did not build the mapping label the correct leaf, and let a second person check a share of the labels.
- Classify the errors. Separate a wrong sibling leaf (Armchairs instead of Sofas), a parent node instead of a leaf, and a wrong branch (Fashion instead of Home).
- Call out critical errors. A product placed in a category with different safety, age or policy rules is critical, whatever its share of the total.
Keeping the mapping current
Mappings decay. Sellers rename their categories, channels publish new versions of their taxonomies and your own tree gains leaves as the assortment grows. Give the mapping an owner, usually the taxonomy owner, and a review routine.
- Log every source category that arrives without a mapping, and review the log on a fixed schedule.
- Re-run the labeled sample after any change to the tree or to a standard you map to.
- Keep each rule's history, so a reversal can be traced to the change that caused it.
- Retire rules that no longer receive products instead of letting them accumulate.
- Tell sellers when a mapping they rely on changes, so their next upload does not land in review unexpectedly.
Youzu helps marketplaces, classifieds and retailers turn raw seller and supplier product data into structured, reviewable catalog records, starting from your own category tree. Whatever approach you use, the measurement above is how you know it is working.
Boundaries
What Youzu does not do
- Youzu publishes no accuracy figures. Measure accuracy on your own sample against your own labels, with critical errors reported separately.
- Youzu does not replace your ERP, search or storefront; those systems stay authoritative for their own data. It can replace your PIM, or work alongside it.
- Youzu does not promise an onboarding time in advance. We size the timeline with you after seeing a sample.
Checklist
Taxonomy mapping template
One row per source category: the pattern it follows, the target leaf or the per-product rule, the owner and the date it was last checked. The first rows show the furniture example from this guide.
A CSV file of 5 rows, named youzu-taxonomy-mapping-template.csv. No form to fill in first.
| Source | Source category | Pattern | Target leaf | Decision method | Owner | Last checked |
|---|---|---|---|---|---|---|
| Seller A | Sofas | One-to-one | Home > Living room > Sofas | Category rule | Taxonomy owner | |
| Seller B | Home / Misc | One-to-many | Decided per product | Title, attributes and photo; review when 2 leaves fit | Catalog operations | |
| Seller C | Couches | Many-to-one | Home > Living room > Sofas | Category rule | Taxonomy owner | |
| Seller C | Settees | Many-to-one | Home > Living room > Sofas | Category rule | Taxonomy owner | |
| Seller D | Home | Parent only | Decided per product | Treat as no category | Catalog operations |
FAQ
Questions buyers ask
Short answers to the questions teams ask about this topic.
Most retailers and marketplaces keep their own tree for browsing and map it to standards where channels require them. Standards such as the Google product taxonomy or GS1 GPC are useful references and channel targets, but your own tree should follow how your shoppers browse and filter, not how a standard was designed.
Start per category, because most source categories map cleanly and one rule then covers every product in them. Switch to per-product decisions for catch-all or mixed categories, where the title, attributes and photo have to decide the leaf. Keep a written list of which source categories use which approach.
Label a random sample, add the difficult categories on purpose and compare the mapped leaf with the label. Report exact-leaf agreement, the error classes and critical errors as separate lines. Measure accuracy on your own sample against your own labels, with critical errors reported separately, and re-run the sample after any change to the tree.
Only as a temporary state that someone owns. A product in a parent node gets no leaf attributes and no leaf filters, so shoppers who filter will not find it. Treat a parent-node mapping as an open item in a queue, and resolve it to a leaf as soon as there is enough evidence.
Yes. Youzu maps products to your category tree and extracts the attributes your category defines; numbers and yes/no values stay typed, and values the evidence does not support are left empty. Send a sample file and your category tree to see the mapping on your own products before deciding anything.
Last reviewed
Sources
Google product category [google_product_category], Google Merchant Center Help. support.google.com/merchants/answer/6324436. Accessed . Back
Google product taxonomy, plain text with IDs, Google. google.com/basepages/producttype/taxonomy-with-ids.en-US.txt. Accessed . Back
Global Product Classification (GPC), GS1. gs1.org/standards/gpc. Accessed . Back
United Nations Standard Products and Services Code, UNSPSC. unspsc.org. Accessed . Back
About this guide
Published by Youzu · Editorial policy
- Published
Every guide defines its terms, works through one example and cites the public sources it relies on, with the date each source was checked.
Guides explain methods. They do not rank Youzu above other approaches, and an example is never presented as a customer result.
More in all guides and the catalog glossary.
Related
Keep reading
Product
Automatic product categorization and taxonomy mapping
Map every product to your category tree: a category you supply skips inference, a parent limits the choice and unresolved categories are not guessed.
Glossary
What is a product taxonomy?
Product taxonomy defined in plain words, with a worked example, common mistakes and links to the guides and Youzu products that apply it.
Glossary
What is taxonomy mapping?
Taxonomy mapping defined in plain words, with a worked example, common mistakes and links to the guides and Youzu products that apply it.