Search Engines & Filtering
A large e-commerce catalogue means as many ways to lose a sale. The customer types ’white trainers size 8’ and receives sports socks and black trainers from another brand. They leave the site. Your merchandising teams spend their days creating manual redirect rules to compensate for a search engine that understands nothing.
Meanwhile, the conversion rate for internal search stagnates, although it is the journey where buying intent is clearest. The filters? They display impossible combinations: size 8 in blue when stock is zero, or worse, hide available products due to poorly indexed attributes. Your PIM is clean, your ERP up to date, but between these systems and the customer, search destroys value.
Each day of delay means revenue going to competitors whose sites ’guess’ what the user is looking for.
This solution belongs to the E-commerce & Booking family.
Challenges holding back your performance
Your e-commerce CMS’s native search functionality was designed as a secondary feature, not as a conversion tool. An open-source engine maintained internally? Your teams spend more time keeping the infrastructure up than working on relevance. Mainstream SaaS solutions work on simple catalogues: they collapse when you have dozens of attributes per product, variants in several languages and business rules that change according to season, stock or margin.The makeshift approach of managing synonyms in a CSV file cannot cope with reality: your customers type ’American fridge’, ’US refrigerator’, ’double-door fridge-freezer’ for the same product. And nobody has the time to maintain these lists manually across all your SKUs.
Our Technical Approach
Our approach stems from a fundamental observation: e-commerce search is not a technical problem, it is a business understanding problem.
We build an indexation layer that truly reflects your catalogue: not a flat copy of your database, but an enriched structure where each product carries its context: sector-specific synonyms, category hierarchies, popularity signals, stock constraints. The engine learns from your existing search data: queries with no results, clicks following reformulation, abandoned filters.
Typo-tolerance and natural language processing (NLP) are not gimmicks: they absorb the misspelt queries that currently return zero results. Facets are calculated in real time on available stock, not on the theoretical catalogue. Merchandising integrates directly: your teams can boost a product on promotion or bury end-of-line stock without touching the code.
Search As-You-Type
Display of relevant results and thumbnails from the first letters typed.
Typo-Tolerance
Algorithms for distance correction that automatically rectify common input errors.
Dynamic Facets
Instant recalculation of attributes (colours, remaining sizes) following filter selection.
Active Merchandising
Business rules to push certain products to the top of search results.
Asynchronous Indexing
Invisible index update as soon as a price or stock level is modified.
Synonyms & NLP
Natural language comprehension to link ’basket’ and ’sneaker’ automatically.
Technical Architecture
Search, check availability and price, order: these three steps also structure Online Auction Platform and Direct Booking Engines.
- Search As-You-TypeDisplay of relevant results and thumbnails from the first letters typed.
- Typo-ToleranceAlgorithms for distance correction that automatically rectify common input errors.
- Dynamic FacetsInstant recalculation of attributes (colours, remaining sizes) following filter selection.
- Active MerchandisingBusiness rules to push certain products to the top of search results.
- Asynchronous IndexingInvisible index update as soon as a price or stock level is modified.
- Synonyms & NLPNatural language comprehension to link ’basket’ and ’sneaker’ automatically.
Smooth and frictionless integration
We connect to your existing PIM or catalogue feed via API or scheduled import. Indexing runs in parallel with your current site: no interruption. You can switch search functionality page by page: first the product listing, then the homepage, then the category filters. Your front-end remains yours, we provide the results.
Merchandising teams maintain their usual practices, with a dedicated interface that doesn’t require an IT ticket for each modification.
Audit of catalogue and search
Analysis of your product structure, existing search logs and zero-result queries. Identification of the critical queries that concentrate most of the traffic.
Pilot modelling and indexing
Construction of enriched index schema on a catalogue segment. Configuration of synonyms, typo-tolerance and initial relevance rules.
Front-End Integration & Merchandising
API connection to your existing site, deployment of the merchandising interface for your teams. A/B testing on a portion of the traffic.
Toggle and continuous optimisation
Production deployment across all search traffic. A dashboard for performance monitoring and regular adjustments based on actual data.
Measurable results for your organisation
- Internal search treated as a conversion tool, not as a secondary feature
- Zero-result queries absorbed by typo-tolerance and automatic synonyms
- Response times that stay stable on a large catalogue, whatever the number of references
- Autonomous merchandising: boosts and business rules applied without technical intervention
- Manual redirect rules replaced by relevance rules
- Incremental indexing: a new product becomes searchable once published in the PIM
Clarifying your decision-making
We already have an open-source search engine internally, why change?
An open-source engine is an excellent base, but it is a low-level component. What is missing is the e-commerce business layer: sector-specific synonyms, visual merchandising, dynamic facets on live stock. Clarendis can build on your existing cluster where the infrastructure is sound.
What impact on site performance during indexation?
Indexation has no impact on site performance: it runs asynchronously on separate infrastructure. Your site sees only an optimised query interface. Even a full reindex of the catalogue generates no client-side latency.