Skip to content
Interface language: English
Design / Signal

Shop filters: customers should know why only three products remain

Explain the active filter state, result counts, empty results and return from a product, including on a phone.

Article language: English

People on terraces with coloured panels and tables overlooking water and hills.
In this article

There were a hundred products in the catalogue. After a few clicks, three remain. The customer remembers choosing a colour, but not the material selected earlier or whether the shop still shows only items available immediately. The closed panel looks harmless. It is where the reason for the sudden contraction of the offer has hidden.

Filtering is a conversation about the conditions of a choice. A visitor says: I am interested in this, in this size, for this use. The shop responds with a matching set. Good design makes the response understandable, lets the person change their mind and recover a broader view. The number of controls matters less than agreement between the person's choice and the rule the catalogue actually applies.

Establish what a choice means before drawing the panel

Not every product detail makes a useful filter. A field should divide the offer in a way that helps buyers. Colour may matter; an internal collection code may not. Start with customer questions and common purchase conditions. Otherwise the filter becomes an exhibition of every column the system could export.

Separate a product category from its attributes. “Floor lamps” can be a place in the catalogue, while height and colour narrow that group. Mix the two orders and visitors will not know whether they entered a new section or changed the set on the same page. Earlier work on information architecture (in Polish) helps.

Check what several selections within one group mean. Choosing red and green may mean accepting either colour. Combining two equipment features may instead mean the product must have both. Check that relationship separately for each attribute group. Write the rule in words and compare it with the customer's actual task.

Shopify's Search & Discovery documentation describes combining different filters with “and” and, by default, values within one filter with “or”; some filter types can use different behaviour. This illustrates why the appearance of selected controls is not enough to infer a catalogue rule.

The interface need not use logical terminology. Clear labels and consistent behaviour are enough. If choosing two sizes broadens the set, let the visitor see that result. When another required feature is added, a smaller product count should be clear and consistent with expectations rather than look like a list failure.

Pay attention to products with variants. A family may offer a small red model and a large green one. A customer choosing red and large needs a real combination, not two attributes found somewhere in the family. Agree what the page should show and which variant it should lead to. An item that formally matches a general description may still not be purchasable in the intended version.

Value names come from data. “Graphite”, “anthracite” and “dark grey” should not become an accidental map of poetic synonyms. Equally, do not combine real differences merely to shorten the list. Work on the catalogue model (in Polish) before polishing the panel. Then you know what can sensibly be grouped and who approves a change.

The result should remember how it was reached

Show active constraints beside the list even when the panel is closed. Visitors should recognise the reason for the current set without reopening every group. Short removable labels can help if they contain enough context. “Large” alone says little if size, capacity range or set format can all be large.

Distinguish removing one value, clearing a group and resetting everything. They have different effects. “Clear” beside a price filter should not accidentally erase a chosen location or category if the visitor treats those as the starting point. The action's name should match what actually returns to its initial state.

Default constraints are constraints too. If the shop starts with items available immediately, show that clearly. No visible selection suggests the full offer. A buyer may not realise that an interesting product was omitted because of fulfilment time, not because the company does not sell it.

The result count should refer to the correct set. If the list later allows a bulk action, also check the scope of “Select all” after filtering (in Polish). Agree whether the count covers products, variants or orderable entries. Three families with a dozen versions each present a different picture from three ready variants. You need not show all these numbers at once. You do need to know what “3 products” promises and whether the cards below fulfil that promise.

Define the numbers beside individual values. Does a number mean the products returned after adding that value, or the products that have it within the current set? In a complex catalogue those figures can differ. If behaviour cannot be explained with a simple example, simplify the presentation instead of decorating it with apparent precision.

Do not mix sorting with filtering. The first determines order; the second determines membership in the set. “Most popular” should not quietly narrow the offer to selected hits. If you promote an editorial collection, name it as such. Someone seeking a specific item needs to know whether they changed the view or the scope.

Preserve the relationship with the query too. If a customer entered a name and then used filters, show both conditions. Clearing a filter need not erase the search text. Internal search (in Polish) and the narrowing panel should jointly explain why the list looks as it does, rather than act like two independent devices.

If results are empty, retain the selections and allow the person to remove a chosen constraint. The message should distinguish “nothing meets these conditions” from a data-loading problem. An empty list is not a 404 error (in Polish). The address and task still exist; the criteria or handling of a temporary failure need to change.

Aeneia conceptual photography

Do not immediately reset every choice for the customer. One important feature may be fixed while another is open to compromise. Someone choosing a product to fit a dimension may be unable to reduce it but willing to accept another colour. Design should help that conscious correction without suggesting random items under “you may like this”.

For numeric ranges, name what the boundaries describe. “Up to 40” without a unit and attribute could mean price, size or capacity. A slider does not remove the need to explain values. If a number can also be typed, check that both methods show the same state and that an accidental touch can be corrected.

Decide what an incomplete range means. Someone may want everything below a particular height without a lower limit. An empty field need not mean zero. Consider reversed boundaries and values outside the range too. The message should help correct the choice while preserving the other valid number, rather than reset the whole group.

Two slider handles also need to be distinguishable. Do not rely solely on their position when they move close together. Check keyboard operation and changing one boundary without touch. For a given task, two ordinary labelled fields may be clearer than a tiny slider that looks beautiful only in its default state.

Show the rounding rule for a range as well. When a customer selects 40, they should not have to infer from results whether the boundary itself is included and how fractions of the unit are treated. Record examples that meet and fail the condition. The visual layer and catalogue mechanism then share a definition.

A missing product parameter is not the same as failing a condition. Decide whether such an item disappears from the narrowed set or needs a separate explanation. Without that decision, partially described entries may be systematically omitted, and the team may try to repair filters when the problem lies in incomplete catalogue data.

On a phone, the panel becomes a separate stage

A small screen often calls for a drawer or full filter panel. Before drawing it, decide when the list changes: after each selection or only after the whole set is applied. Both can work. The button, counter and panel closure must describe the same moment of decision.

Carbon Design System distinguishes instant updates from applying several filters together. In the latter model, the user first sets several constraints and then refreshes the set through a separate action. Choose your own design according to the number of decisions and catalogue response time instead of copying one control's behaviour.

If a selection is still a draft, closing the panel cannot have an ambiguous effect. The close icon may cancel changes or merely hide the view. Decide explicitly and test on a phone. Do not make visitors inspect the list after each return just to guess whether the new conditions have been accepted.

With batch application, the button can show the expected result, but it must match the current state. If the number is still being calculated, do not pretend the answer is ready. Protect against a delayed response to an earlier selection appearing after rapid changes. Customers need not know how data is fetched; they need a consistent signal.

The panel must preserve access to its own actions. Check that the apply button does not disappear beneath the keyboard and that a long value list does not cut off reset. Principles for a mobile menu (in Polish) help assess opening, closing and orientation, but filters also need information about the active set.

The selection should survive viewing a product

After narrowing the list, a customer opens a product, reads it and returns. If they get the beginning of the full catalogue, they must reconstruct both filters and browsing position. Retaining conditions, order and a sensible list position helps continue the choice. It should not require remembering five controls or photographing the screen before every transition.

The same applies after adding an item to the basket (in Polish). A customer may still be choosing another product for the same use. Do not assume the first addition ends catalogue work. Agree whether “continue shopping” returns to the current list, category or home page, and why.

When a location is selected, show its scope. An “available for collection at a selected location” filter can help orientation before purchase, but must not pretend to be a final reservation. Delivery and collection conditions (in Polish) need confirmation for the selected variant and the whole order. Keep that boundary in the filter name and explanation.

A narrowed-view URL can become part of a conversation. If sharing is supported, open the link in a new session and check its meaning. The recipient should see the same conditions or a clear explanation that they depend on their location or session. Adding parameters to an address is not enough if the page cannot reproduce their effect.

Do not preserve an outdated selection without explanation. A value may disappear after a catalogue update, or an availability filter may no longer make sense in a new category. Keep as much context as remains true and say what changed. A silent reset is convenient for the system but deprives the visitor of an understanding of the difference.

For acceptance, test several paths that actually occur in the shop: a quick single attribute, several dependent conditions, an empty result and return from a product. Include long labels and missing data. This is not a contest in the number of possible selections. The question is whether the customer can explain which products they are seeing and how to change the scope.

Aeneia Design combines content, interface and agreed shop implementation. Bring example products and a few real buyer questions to a discussion about filters. After the third click, the three remaining products can then be the result of a choice rather than a puzzle to solve before buying.

Aeneia / Aeneia editorial

Editorial article. Source links are included in the text.

About Aeneia

Brands we have worked with

  • L’Oréal
  • Garnier
  • Netflix
  • Carrefour
  • Warner Bros. Discovery
  • Biedronka
  • HBO Max
  • Wolt
  • Żabka
  • COSRX
  • Embassy of China in Poland
  • Mixa
Aeneia / Initium

From knowledge to a decision

Tell us what you want to put in order. We will agree on where to start.

Contact Aeneia

We partner with and use solutions from:

  • Google
  • OpenAI
  • Microsoft
  • Stripe
  • Anthropic
  • Linear