Zalando's product taxonomy defines which attributes describe an article and which values each attribute may take. Within that taxonomy, Zalando marks a group of attributes as high-value and recommends that partners complete them. This post explains what those attributes are, how incomplete data affects a partner brand's results on the platform, and why completing them at the PIM level is more sustainable than doing it channel by channel.
What Zalando means by high-value attributes
Every article on Zalando is described through a category-specific set of attributes. A dress, for example, has fields for silhouette, neckline, sleeve length, length, fit, shape, pattern, fastening, occasion, material composition and care instructions. Each field has a controlled list of allowed values.
Some attributes are mandatory at onboarding. If they are missing, the article is not approved. Zalando publishes an attribute requirements table per silhouette that lists them.
High-value attributes (HVAs) are mostly optional at onboarding, which is why they are often left empty. Zalando treats them differently from other optional fields. The Partner University describes HVAs as attributes that elevate product detail pages, integrator documentation from Lengow repeats Zalando's recommendation to complete them to drive better performance, and zDirect includes an Attributes insights dashboard that reports HVA completeness across a partner's assortment. Zalando has not published performance figures behind this recommendation.
Based on Zalando's article mapping guide, the HVA set for apparel includes collar, neckline, pockets, details, sleeve length, fastening, pattern, occasion, care instructions and measurements. Fit, shape and length are documented separately. The applicable list varies by category and is available inside zDirect.
One characteristic of HVAs is worth stating on its own. According to ChannelEngine's Zalando marketplace guide, Zalando does not allow content changes once a product has been exported, and after approval, attribute updates via the API are no longer possible. HVAs are the exception: they can be updated at any time, and the integrator forwards the changes to Zalando. In practice, this means HVAs are the part of a live Zalando catalog that can still be improved after launch.
How incomplete attributes affect results
Incomplete HVAs affect three outcomes: visibility in filtered results, return rates, and inclusion in curated selections.
Visibility in filtered results
Shoppers on Zalando narrow large assortments through filters, and each filter corresponds to an attribute. When a shopper filters by sleeve length, only articles with a value in that field are included in the result. An article with the field empty is not ranked lower; it is absent from that view. Integrators working with Zalando feeds describe the same effect: articles with incomplete variant data remain hidden behind filters even when they are approved and in stock.
For a brand with several hundred articles and several empty HVAs per article, a meaningful share of the assortment is not reachable through the filters that shoppers with a specific intent use most.
Return rates
Most fashion returns come from a difference between what the shopper expected and what arrived. The attributes that set expectations before purchase are fit, shape, length, sleeve length, material and measurements. When these are empty or generic, the shopper decides with less information, and the probability of a return for fit or description reasons rises.
Zalando publishes separate guidance on the Fit and Shape attributes that explains how to describe oversized, fitted or relaxed pieces accurately. We have covered the cost of fashion returns in an earlier post. On a marketplace, return rate is also a partner metric that the platform observes.
Inclusion in curated selections
Attributes such as occasion, pattern and details feed recommendations, curated pages and campaign selections. An article with occasion set to a specific value can be included in a seasonal edit built around that occasion. An article with the field empty cannot. This is the outcome where attribute completeness supports growth rather than only preventing loss.
Not sure how many of your articles are missing high-value attributes? Book a call and we will show you where your catalog stands.
Why the PIM is the right place to complete them
A common approach is to export the articles with empty HVAs from zDirect or the integrator, complete them, and re-import. This works for a single pass. It is harder to maintain as a process, for three reasons.
Most brands sell on more than one channel. Zalando has its taxonomy. Google Shopping has its product data specification. Meta has its catalog fields. Shopify has its Standard Product Taxonomy and metafields. About Zalando, Lengow, ChannelEngine and Plytix all document the same situation: each channel needs roughly the same attributes in a slightly different form. Attributes completed inside the Zalando channel are available only to Zalando. Attributes completed in the PIM can be mapped to every channel.
Catalogs change every season. New collections arrive with the same gaps as the previous ones. Enrichment done per channel has to be repeated per channel.
HVAs can be updated after launch. Because Zalando accepts HVA updates on live articles, a PIM that holds enriched values can push them to Zalando on a regular schedule rather than only at onboarding. Earlier collections improve while the current one is being loaded.
A PIM such as Akeneo, Pimcore, Plytix or Salsify, or an equivalent internal system, is where the single enriched product record should live. Channel connectors read from it. This is the same reasoning we described for Shopify metafields as the source of truth, applied to a multi-channel setup.
How AI enrichment works in practice
Enrichment at scale means deriving attribute values from data the brand already has: product images, supplier sheets and existing descriptions. Zalando offers a version of this within zDirect, where partners can upload raw data and Zalando's AI predicts attribute values for live articles. That indicates the approach is accepted on the platform side.
Doing the same on the brand's side, inside the PIM, follows four steps. They are the same steps we use for Shopify catalogs.
1. Define the rubric from the target taxonomy. Take Zalando's attribute list and the allowed values for each silhouette the brand sells. This becomes the fixed vocabulary the model is permitted to output. The model returns one of the allowed values or leaves the field empty; it does not produce free text.
2. Select articles and collect inputs. Connect to the PIM, select the articles with empty or low-confidence HVAs, and retrieve their images and any supplier text. Processing in batches of a few hundred keeps the review step manageable.
3. Classify with structured output. Send the image and text to a vision model together with the rubric, and request one value per attribute, a confidence score, and a short reason. Visually observable attributes such as neckline, sleeve length, pattern, fastening, collar, pockets and length can be predicted with high confidence. Attributes that are not visible, such as fibre percentages and care instructions, should not be predicted from images. Those remain sourced from supplier data, and the model is instructed to leave them empty.
4. Write back with review and a change log. Values above a confidence threshold are written to the PIM. Values below it go to a review queue. Each write is logged with the previous value, so a batch can be reversed if needed. The PIM's Zalando connector then sends the updated HVAs to the live articles.
The result is a catalog in which the attributes Zalando recommends completing are populated, consistent with the allowed values, and traceable, and in which the same values are available to Google, Meta and the brand's own storefront.
Want every high-value attribute in your catalog filled within a week? Book a call and we will run this process on your assortment.
Where SIXFIT fits
Today, SIXFIT applies this process to Shopify catalogs. The free Catalog Checkup scans products for missing and inconsistent attributes, and our open-source Claude skills on GitHub run the enrichment flow described above against a Shopify store.
A PIM connector that runs the same analysis against Zalando's taxonomy on Akeneo, Pimcore and Plytix data is the next step. If you sell on Zalando and want to be among the first brands we run it with, book a call. We will start with an HVA completeness assessment of your current assortment.
Frequently asked questions
- What are Zalando high-value attributes?
- High-value attributes (HVAs) are a defined group of product attributes within Zalando's taxonomy that Zalando recommends partners complete because, according to Zalando's own guidance, they drive better performance. They cover the fields shoppers use to filter and evaluate products, including neckline, sleeve length, fit, shape, length, pattern, occasion, fastening, collar, pockets, material composition and care instructions. The exact list depends on the product category and is available in zDirect, where an Attributes insights dashboard shows how complete your HVAs are.
- Can I update attributes after my articles go live on Zalando?
- For most content fields, no. According to integrator documentation from ChannelEngine, Zalando does not allow content changes once a product has been exported, and once Zalando approves an article, attribute updates via the API are no longer possible. High-value attributes are the exception. They can be updated at any time, through zDirect or through an integrator, and Zalando applies the changes to the live listing. This makes HVAs the part of a Zalando catalog that can be improved after launch without re-submitting articles.
- Do high-value attributes affect Zalando search ranking?
- They affect whether an article appears in a result set rather than its position within it. Zalando's search and category pages rely on attribute filters. When a shopper filters by sleeve length or occasion, only articles with a value in that field can be shown. An article with the field empty is excluded from that filtered view. Completing HVAs is therefore mainly a question of eligibility to appear, not of ranking.
- How does missing product data increase returns?
- A large share of fashion returns is caused by a gap between what the shopper expected and what was delivered. Fit, shape, length, sleeve length and material are the fields that set that expectation before purchase. When they are empty or inaccurate, the shopper has less information to decide with, which raises the likelihood of a return for fit or description reasons. Zalando publishes dedicated guidance on the Fit and Shape attributes for this reason.
- What is product data enrichment?
- Product data enrichment is the process of completing missing or low-quality attributes on existing catalog data. In fashion, the inputs are typically product images, supplier data and text descriptions, and the outputs are structured values that match a target taxonomy such as Zalando's, Google's or Shopify's. Enrichment can be done manually, with rules, or with AI. AI-based enrichment reads the image and text, predicts a value from the allowed list for each attribute, and returns a confidence score so that uncertain results can be reviewed by a person.
- Can AI fill Zalando attributes from product images?
- Yes. Zalando itself offers a form of this: partners can upload raw data in zDirect and Zalando's AI predicts attribute values to enrich live articles. Running enrichment on your own side, at the PIM level, gives you a review step before anything is published and lets the enriched values be reused across all channels rather than only Zalando. Vision models perform well on visually observable attributes such as neckline, sleeve length, pattern and closure type. Attributes that cannot be seen, such as fibre composition, should continue to come from supplier data.
- Do I need a PIM to sell on Zalando?
- No. Many brands supply Zalando from Shopify, an ERP or a spreadsheet through an integrator. A PIM becomes useful when a brand sells on several channels, because each channel has its own attribute schema and the same product data has to be maintained in several places. A PIM holds one enriched record per product and maps it to each channel. Brands that do not use a PIM can apply the same enrichment approach directly to Shopify metafields, which is what SIXFIT's Catalog Checkup does today.