Search Relevance Tree
Rank search features from highest to lowest priority. Drag to reorder. The search engine applies these tiers in order when sorting products.
PRIMARY factor for ranking the STANDARDS list (not products). Standards whose standard_name contains the full query phrase rank first (e.g. 'gate valve' → API 600 'Steel Gate Valves' with 30 products ranks above JWWA B 122 'gate valve' with 3 products). Among name-matched standards, those with more connected products rank higher. Standards without the phrase in their name fall below all name-matched ones.
STRONG signal: the product's own name contains the query words (e.g. searching 'gate valve' → a product named 'Slab Gate Valve DN 200' ranks at the top because its name directly contains 'gate' and 'valve'). This is a direct textual match between the user's search and the product name — no standard or manufacturer inference needed.
STRONG signal: the product's category name contains the query words (e.g. searching 'gate valve' → all products in the 'Gate Valves' category rank high, even if their individual product name doesn't contain 'gate valve'). The category tells us the product family, so a category match is a strong relevance signal.
Product is directly linked to a standard that matched the query (e.g. EN 15512 → products with EN 15512 in their standards array). This is the strongest possible signal.
Builds a co-occurrence matrix: when a standard is directly matched, collects all manufacturers with products linked to it, then counts how many of those manufacturers use each other standard. Products linked to standards used by more of these manufacturers rank higher (e.g. ISO 10297 → 10 manufacturers → ranks above EN 331 → 9 manufacturers → ranks above UL 312 → 7).
Products found via the matched standard's group name (e.g. EN 15512 → 'Industrial racking and shelving' group → all racking/shelving products, even if not yet linked to the standard).
Products whose name contains the product-type phrase extracted from the standard title (e.g. 'Cold-Water Meters' → 'water meter' → all water meters).
Products matching the broad noun from the standard title (e.g. 'meters' → all meter-type products). Less precise than the category phrase.
How many distinct query words appear in the product name. 'AWWA C508 Swing Check Valves' ranks above a plain 'Swing Check Valve' for query 'AWWA C508'.
How many of the matched standards the product is linked to. A product linked to 3 matched standards ranks above one linked to 1.
Products from manufacturers whose name matched the query (e.g. searching 'Topas' → all Topas products, even if the product name doesn't contain 'Topas').
Products from similar manufacturers (pre-computed via shared standards fingerprint). When a manufacturer matches the query, products from its similar manufacturers also surface — expands the result set to peer companies making equivalent products.
Total number of standards linked to the product (not just matched ones). Standards-rich products rank higher among equals.
Products from standards related by name keywords (fallback when no direct products exist). e.g. SII 1498 → EN 1176 playground equipment.
Each product shows which relevance factors are active for it.
When a user searches, the engine collects products from multiple sources (direct Algolia match, standard links, group expansion, manufacturer bridge, etc.). It then sorts them using the ordered tiers above — top to bottom.
Factor #1 is the strongest filter: if a product matches factor #1 and another doesn't, the first always ranks higher. Only when two products tie on factor #1 does factor #2 break the tie, and so on.
Disabling a factor removes it from the sort entirely — products won't be compared on that dimension. Use this to eliminate noise (e.g. disable "Broad Category Noun" if it brings too many irrelevant matches).
Changes take effect immediately after Save. The config is stored in AppSettings key search_relevance_tree.
The standards list shown alongside search results is ranked differently from products, using a two-tier logic that depends on the query type:
1. Standard-number queries (e.g. "EN 15512", "ISO 9001")
When the query looks like a standard reference, standards are matched by number prefix and returned in Algolia's relevance order. No re-sorting is applied.
2. Text / keyword queries (e.g. "gate valve", "water meter")
Standards are re-sorted using two priorities, applied in order:
- Standard name match (primary): Standards whose
standard_namecontains all the query words rank first. The standard name is the most important signal — a standard titled "Steel Gate Valves" is clearly relevant to a "gate valve" search, even if it has few linked products. - Connected product count (tiebreaker): Among name-matched standards, those with more linked products rank higher. This puts API 600 ("Steel Gate Valves", 30 products) above JWWA B 131 ("Water supply gate valve", 0 products) — both match the name, but API 600 has real products behind it.
Standards that don't match the name at all (e.g. API 6D "Pipeline Valves" for "gate valve") fall below the name-matched group, then sort by product count among themselves.
This logic lives in the algoliaSearch backend function and runs automatically for every text query. It is not part of the draggable product tree above.