Configure the engine

Facets and filters

Order, display type, ranges, dependent facets.

A facet is a filter with a count next to it — Material: V4A (212). It is how a buyer gets from 400 results to four without typing another word. This article covers which facets appear, in what order, and in what form.

Before you begin. Facets come from attributes marked Filterable (facet) in PIM › Data model › Attributes, and their values come from select options or measurements. A facet over a free-text field is not worth configuring — fix the data model first.

Choose the facets for a category

Facets are configured per category, not once globally. Thread size is the first thing a buyer narrows on for fittings and pure noise on packaging material. A single global facet list is the most common mistake in B2B search, and it makes every listing page slightly wrong.

  1. Open the category in PIM › Categories and go to its search settings.
  2. Add the attributes that are useful for this branch of the tree, in the order a buyer decides in.
  3. Set the display type per facet (see the table below).
  4. Save. The change takes effect on the next request — facet configuration does not require a reindex, unlike adding a new facet field.

Three rules that hold up in practice:

  • Five to eight facets visible. Beyond that buyers stop reading and start scrolling. Put the rest behind More filters.
  • Order by how buyers decide, not alphabetically. In technischer Großhandel the first cut is almost always dimension or thread; manufacturer comes later. In Ersatzteilhandel it is the machine.
  • Inherit down the tree. Set the facet list on Fittings and let its children use it, overriding only where a subtree genuinely differs. Otherwise you maintain the same list forty times.

Display types

TypeUse it forNotes
ChecklistSelect attributes with up to ~15 options: material, manufacturerMulti-select, OR within the facet
Range sliderMeasurements: pressure, length, weightNeeds a numeric field, not text
Range bucketsMeasurements with clustered values: up to 10 bar, 10–100 bar, over 100 barBetter than a slider when values are uneven
SwatchColour, surface finishOnly where the visual difference is the decision
Boolean togglein_stock, rohs_compliantOne line, no counts needed
Searchable listFacets with hundreds of values: manufacturer in a large GroßhandelThe list itself gets a small search box

Numeric ranges

A numeric facet only works if the value is stored as a number with a known unit. This is the practical reason measurement attributes exist: the index holds the value converted to the family's standard unit, so 100–250 bar compares correctly against products entered in bar, in MPa and in psi.

Two decisions per numeric facet:

  • Slider or buckets. Use buckets when values cluster — most pressure ratings in a catalog sit on a handful of standard figures, and a continuous slider invites buyers to select 137–214 bar and get nothing. Use a slider when values are genuinely continuous, like length.
  • Where the boundaries sit. Set them on the standard sizes your industry uses, not on equal intervals. M4–M8, M10–M16, M18 and above is a useful filter; three equal thirds of the numeric range is not.
Twenty-four checkboxes is not a filter. A pressure rating with 24 distinct values rendered as a checklist is worse than no facet at all: it is long, each option has a tiny count, and the buyer has to know the exact figure. Convert it to ranges.

Dependent facets

Some facets only make sense once another choice has been made. Thread size is meaningless until the buyer has narrowed to fittings. Bore diameter is meaningless until they are looking at bearings.

The mechanism is straightforward: a facet is declared with a condition, and the storefront only requests and renders it when the condition holds.

ConditionFacet that appears
Category is under FittingsThread size, connection type
Material = v4a or 1.4571Surface finish
Category is under BearingsBore diameter, outer diameter, width

Dependent facets are how you keep a filter panel short without hiding anything important. They are also how you avoid the facet that shows (0) against every option, which is what happens when an attribute is only filled on part of the catalog.

Facet counts and why they sometimes look wrong

The count next to an option is the number of products that would remain if you selected it, given the filters already applied. Two behaviours confuse people, and both are correct:

  • Counts within one facet do not shrink when you select in that facet. Selecting Material: V4A does not reduce the other material counts, because material choices combine with OR. Counts in other facets do shrink.
  • Counts never include products the buyer is not allowed to see. A closed channel, a customer-specific assortment, a disabled product — all removed before counting. This is the right behaviour and the usual explanation for "the count does not match the product list in the Cockpit". See Common search problems.

What to check

  1. Open your three biggest categories on the storefront and count the visible facets. More than eight, cut.
  2. Select one option in each facet in turn. Every one must reduce the result count and none must produce zero.
  3. Look for options with a count of one or two across a whole category. That is usually inconsistent data — V4A and v4a as separate options — not a real filter. Fix it in the catalog.
  4. Check a numeric facet against a product you know. If the range excludes it, the unit conversion or the attribute type is wrong.

Next