Lexical Keyword Search
Lexical keyword search forms the foundation of Smart Search. It matches what customers type against indexed product data and applies configured ranking, filtering and merchandising rules to return precise and predictable results.
It supports matching across product codes, names, descriptions and attributes, alongside synonyms, spelling corrections and configurable search logic. Merchants retain control over ranking, sorting and filtering, ensuring results align with commercial priorities.
If semantic search is not enabled, lexical keyword search remains the primary search method.
AI Semantic Search
AI semantic search extends product discovery by interpreting the intent behind a query. It can return relevant results even when the query does not share the same exact words as the product content.
For example, a search for “trainers” can return products described as “running shoes”.
Semantic search operates alongside keyword search rather than replacing it, ensuring that discovery is expanded without losing precision.
Hybrid Search Model
When semantic search is enabled, lexical and semantic search stages combine into a single ranked result set.
Keyword search ensures precise matching and commercial control. Semantic search expands the candidate set by introducing conceptually relevant results. The combined output balances accuracy and discovery within one coherent result set.
The interaction between these two stages can be configured using parameters such as confidence thresholds, candidate pool size and score balancing, allowing behaviour to be tuned to the catalogue and trading priorities
tradeit™ Lexical Search delivers configurable, high-performance keyword search designed for ecommerce environments where precision and control are essential.
Smart Search indexes a broad range of product data, including product codes, names, descriptions, attributes, categories, relationships, bundles and variations. It supports multi-language and multi-domain catalogues, along with B2B-specific requirements such as product code aliases.
Indexing is handled incrementally for operational updates such as stock levels, pricing and product changes. This allows search results to stay aligned with the live catalogue without requiring full rebuilds for routine updates. Full indexing is still required for structural changes such as synonym updates or schema changes.
Lexical search includes built-in behaviours that improve result quality without manual intervention. These include handling different word forms, tolerating common spelling errors and removing non-essential words from queries. These behaviours apply consistently across all installations.
Search behaviour is controlled through configurable search passes. Each pass can target specific product fields and define how matching is performed, including weighting, fuzziness and match position.
This enables merchants to prioritise exact product code matches, emphasise product names over descriptions and fine-tune relevance based on catalogue structure.
Smart Search supports configurable faceted navigation, allowing customers to refine results through structured filters. Filtering remains a core part of the search experience and operates independently from semantic matching.
Sorting is fully configurable and can combine multiple signals, including product attributes, business metrics and merchandising rules. Supported metrics include orders, baskets, views, ratings and sales value.
Products can also be prioritised based on availability, recency or defined product groups, allowing ranking to reflect commercial strategy.
All existing merchant controls remain in place, including synonyms, spelling corrections, filtering and metric-based ranking. These controls continue to operate as part of the lexical search stage, providing predictable and manageable outcomes.
A diagnostic view provides visibility into how results are ranked, including which fields matched, which search passes contributed, and how scoring and sorting affected the final order. This enables search behaviour to be reviewed and adjusted without development effort.
Lexical search provides precise matching across product data, full control over ranking and sorting, flexibility for complex catalogues and efficient handling of ongoing data updates.
tradeit™ Semantic Search introduces meaning-based retrieval to extend the reach of keyword search.
Semantic search matches products based on intent and context rather than exact keywords. This helps bridge the gap between how customers search and how products are described.
By introducing additional relevant candidates where keyword matching alone may fall short, semantic search reduces dead-end searches and increases the likelihood of returning useful results.
Semantic search works in combination with lexical search. While semantic search expands the result set, lexical search reinforces exact matches and maintains precision within the final ranking.
Semantic search behaviour is configurable, allowing the balance between semantic and keyword matching to be tuned. Parameters such as confidence thresholds and score weighting ensure that the system remains controlled rather than acting as an opaque model.
The effectiveness of semantic search depends on the quality of product data. Rich descriptions and clear product names improve results, while catalogues dominated by internal codes rely more heavily on keyword matching.
Semantic search operates alongside keyword search and does not replace it. Keyword search remains essential for handling exact matches, particularly for product codes, brand names and specialist terminology.
It does not convert queries into structured filters. If a query includes elements such as colour, price or category, these are not automatically applied as restrictions. Customers continue to refine results using the standard facet and filtering interface.
Semantic search broadens discovery, reduces failed searches, and improves the likelihood of returning relevant results, while maintaining control through its integration with keyword search.
tradeit™ Recommendations complements search by helping customers discover additional relevant products throughout their journey. While search responds to explicit queries, recommendations use behavioural and business data to surface products aligned with customer context and activity.
Recommendations are generated using data such as browsing behaviour, past purchases, basket contents and engagement metrics. These are used to present products that are relevant to the customer’s current context, supporting exploration beyond initial intent.
Recommendations can be driven by a range of metrics, including product popularity, engagement and customer interaction patterns. Multiple metrics can be combined and weighted to control how products are prioritised.
Product relationships such as related items, alternatives, complements and bundles are also supported, including similarity-based recommendations where appropriate.
Recommendations can be deployed across product pages, category pages and search results. They are configurable in both layout and behaviour, allowing merchants to control how and where they appear.
Merchants can define which recommendation types are used, how they are prioritised and how they are presented. This ensures recommendations align with trading objectives and site structure.
Some of the same business metrics used in recommendations can also be applied within search ranking configurations where required.
Recommendations improve product discovery beyond search, increase engagement through relevant suggestions, support cross-sell and upsell opportunities, and provide configurable control over how products are surfaced.
Recommendations and Smart Search fulfil different roles within the ecommerce experience. Search is designed to help customers find products based on specific queries, delivering precise and relevant results. Recommendations extend this by presenting additional products based on behaviour and context. Together, they support both targeted product finding and ongoing product discovery across the site.
Smart Search supports a range of capabilities that improve performance, reliability and control across search and browsing:
When semantic search is enabled, relevance is driven by the hybrid model of semantic and keyword search.
Keyword search continues to support detailed attribute-based filtering and sorting, while semantic search extends discovery without replacing existing controls.
Smart Search is built on OpenSearch, an open platform widely used across the industry. This provides a proven, scalable search foundation without vendor lock-in. Search is hosted within Red Technology’s infrastructure, ensuring consistent performance and data control, with all search activity remaining within your platform environment.
Please complete the form and your download will be sent to you.
01865 880 800
info@redtechnology.com
