A product catalogue starts with data: names, characteristics and descriptions do different jobs
A description alone cannot support catalogue filtering and comparison. Organise attributes, units, variants and sources before designing product pages.
Article language: English

In this article
One product's colour is recorded as "dark graphite", another's as "anthracite" and a third's as "premium grey". Each description sounds reasonable. But the colour filter begins to look like a collection of literary interpretations, and the team cannot tell whether these are three shades or three names for the same one. Changing the appearance of the list will not resolve a problem created during data entry.
A catalogue content model defines what information is stored, what it means and how the pieces relate. Start with a few representative products and questions about what customers need to choose, compare and understand. Screen design can then use those decisions instead of inventing them again every time.
Five products on the table instead of a hundred empty fields
Choose a small, varied set from the real catalogue, not only the best-described items. Include a simple product, a family with variants, a bundle, an item with a missing parameter and one whose name does not fit the current layout. This material helps identify differences; it is not a sample intended to prove that a prepared form is perfect.
Separate the information into parts. A name helps identify a product. A short explanation says what it is for. An attribute, a described characteristic, can support comparison or filtering. Editorial copy brings facts together in an understandable statement. Photographs, instructions and drawings are related materials. Not everything belongs in one large text field.
Start with the catalogue's job. If customers filter by width, width needs a recognisable value and unit. If they compare bundle contents, you need an ordered list of components. "Many practical extras" cannot support that comparison without manual interpretation. A useful field follows from a task, not from a competing panel having many fields.
Record questions the company cannot answer separately. A missing material must not be copied from a similar item. A missing dimension is not zero. The catalogue should allow an explicit missing-information state to be resolved, so the design does not turn an empty field into a false fact. Settle this before importing the first spreadsheet.
Data describes the product, but field names must also make sense to the person maintaining the catalogue. Manufacturer abbreviations can remain as separate identifiers. Editors still need a label, instructions and an example of a valid value. Otherwise each person may use their own definition, with the customer discovering the difference later.
Shopify describes metafield definitions as schemas defining the structure, type and rules of additional fields. This is an example of a technical mechanism, not a requirement to choose that platform. The ability to create a field does not decide what it means in your catalogue or which later task it serves.
A useful first discussion can produce a simple table: information name, meaning, assignment level, format and owner. Add a correct example and a case that raises doubt. "Dimensions" is too broad if one person enters the package and another the inside. Resolve that difference before filters, views and automatic descriptions grow around the data.
Not everything belongs to the same product level
A product family may share a story and material while individual variants have different dimensions, identifiers and stock. If variable values are stored only at family level, a product page must later guess them. If everything is duplicated in variants, changing shared information requires many manual corrections. The model should reflect real relationships.
Decide what the customer actually orders. A standalone product, family variant and bundle need not share an identical structure. A bundle needs links to components and their quantities. Do not replace these with a sentence in the description if you later want to show contents, compare bundles or check what changed between offer versions.
A shared characteristic can be inherited, meaning taken from the parent record. Decide when a special value may override it. If one variant uses a different material, that exception should be visible in the data and on the page, not only in a note remembered by the person handling orders.
Use a concrete example. A family of boxes has the same closure but three widths and two lids. The closure may be shared, usable height may depend on the lid, and the identifier on the complete variant. Variant selection design (in Polish) needs those relationships so the screen does not show correct information in the wrong combination.
Do not add inheritance without a reason. For a dozen independent objects, an elaborate hierarchy raises maintenance cost. The goal is fewer ambiguities and manual copies. The structure should be simple enough for a new colleague to find where a value displayed on a card comes from and where to correct it.
Separate stable characteristics from operational information. Material and construction change differently from availability and dispatch timing. Identify the source and update rhythm. A manually written "we ship tomorrow" pasted into hundreds of products can outlive the day it referred to. New typography cannot repair that data-organisation problem.
Associate images with the specific product, variant and role. One view shows a family, another only the red model with a high lid. This assignment supports a product gallery (in Polish) and the cart thumbnail. A filename may help the team; the catalogue should also explicitly associate each image with its role and the correct model.
An instruction document also needs a defined scope. A document for an older construction does not become current merely because its cover changes. Record which models and versions it covers, and track where it is used when replacing it. A customer downloading it from a product page should not have to decide whether a similar drawing represents their product.
Numbers, controlled terms and sentences do different jobs
For numeric fields, establish the unit and meaning. A value of 120 without a label could mean centimetres, millimetres or units. Conversions should follow a shared rule rather than separate annotations on each product. If size is a range, retain its boundaries rather than put "from 20 to 30" into a field pretending to hold one number.

Do not use one empty value for every situation. "Not applicable", "unknown" and "awaiting confirmation" mean different things. Some states may be hidden on the public page, but the data owner should recognise them. Otherwise they cannot distinguish a complete catalogue from one missing half its required information.
A controlled vocabulary helps where values group products. Colour, finish and mounting type need agreed concepts and rules for adding new ones. That does not eliminate brand language. A distinctive shade name can coexist with an ordinary filter category: the two terms do different jobs. For a large vocabulary, design search and confirmation of a specific record in a field (in Polish), distinguishing selection from merely typing a similar name.
Synonyms also require a decision. "Graphite" and "anthracite" may denote different finishes in a particular offer. Do not merge them simply because they sound similar. Confirm their meaning with the product owner, then prepare a shared label or mapping if appropriate. Catalogue organisation should remove ambiguity while preserving real product differences.
Editorial copy should use confirmed characteristics and explain their application. It need not repeat every technical value if the specification shows it nearby. Manually entering width in both prose and a field creates a risk of divergence. A product page that answers buyer questions (in Polish) needs both data and well-written explanation.
Additional fields should not replace existing concepts without reason. Schema.org's additionalProperty describes an additional characteristic as a property-value pair and points to specific properties where available. That is guidance on communicating data to systems. Your own model still needs to fit the real offer and the team's work.
Separate technical names from customer-facing labels. A field can have a stable identifier while the page label is edited and translated. Changing "overall size" to "external dimensions" should not require replacing an entire integration. Documentation should retain the relationship so the catalogue maintainer knows which terms describe the same resource.
Not every category needs every field. For fabric, sales method and width may matter; for a container, capacity and construction. Shared information can coexist with category-specific fields. The editorial form then need not require a dozen "not applicable" entries for every item that is not furniture.
Validation rules should reflect actual meaning. A dimension may need to be positive, but that alone will not detect millimetres entered into a centimetre field. Check meaning against representative examples. A formally valid record can still describe an item absent from both the warehouse and the manufacturer's offer.
The catalogue must survive the next change
Assign an owner and source to each important item of information. The owner confirms meaning; a manufacturer sheet or agreed system may be the source. Not everything needs weekly review. Decide, however, which value to accept if a new spreadsheet differs from data already used in the store.
Prepare a separate transition plan when the model itself changes. A new field does not populate merely by appearing in a panel. Establish where values will come from, what happens to the existing record and when views will use the new source. Otherwise the store may publish two different descriptions of one characteristic for a time.
When retiring a field, check who still reads it: a filter, card template, export or comparison. Do not delete the old record before confirming that needed information has moved. Equally, do not maintain two manual correction points indefinitely. There needs to be a clear transfer of responsibility to the new field.
Test a change on a typical product and an exception. If three variants inherit the family's material while a fourth has its own, migration must not flatten the difference. Compare the customer view before and after. A count of successfully imported rows does not prove that meaning and relationships survived.
First test an import on a limited set. Check units, identifiers and variant links. Pay particular attention to empty values: does the import add them or erase previous data? A catalogue update decision must include that rule, not just confirmation that a file was technically accepted.
Keep a trace of important corrections. If bundle contents change, check the card, comparison, image and related instructions. In the CMS publishing workflow used by your team, establish who examines the actual page. An admin-form preview does not yet confirm that customers received consistent information everywhere. The guide to publishing content in a CMS (in Polish) can help organise this workflow.
Prepare short catalogue editing instructions. You need not document every platform button if a few decisions suffice: data sources, when to create a variant, how to mark an absence and who approves new vocabulary values. These rules help even after responsibility for the offer changes hands.
Develop the structure from actual need. A new filter may require organising existing descriptions; a new category may need different characteristics. Do not add a field "just in case" when nobody can identify its use. Every field must be filled, explained and maintained. Multiplying columns does not make an empty catalogue richer.
Inspect the result on a small screen too. Structured data helps create a comparison readable on a phone (in Polish), but does not automatically select the right presentation. Long labels, units and missing parameters still require design. The model provides material for a good view; it does not decide what the customer should see first.
Bring a few real product cards and difficult exceptions to make a store-design conversation concrete. Aeneia Design brings content, information architecture and screens together within an agreed scope. Bring a catalogue in which the team can explain the difference between a product characteristic and a sentence that describes it well. That is a basis for the next screens.









