Qwiser Salesperson Search and Navigation, developed especially for e-commerce sites, is a multi-faceted solution whose algorithms effectively respond to all the different kinds of queries shoppers submit to online stores. It operates in three stages:

• Query Analysis -Understands shoppers' queries; immediately responds to clear queries; detects queries in need of refinement and forwards them to the next stage.
• Dynamic Refinement - Interacts with shoppers when queries lead to too many results or when queries are unclear.
• Analytics and Machine Learning - Learns from shoppers' queries and purchases to enhance Qwiser performance on a daily basis; produces real-time reports of shopper activity in the online store.

Qwiser™ Salesperson

Online shoppers submit many different kinds of queries in the search to find what they want on e-commerce sites. Qwiser Salesperson algorithms that work in tandem with extensive knowledge bases, as well as business rules, respond accurately and intelligently to any type of query encountered.

• Queries in natural language
• Single word queries
• Multi-word, complex queries
• Synonyms, slang, idioms
• Vague and unclear queries
• Numbers, fractions, decimals
• SKUs
• Misspellings
• Queries for items not stocked (Qwiser suggests related, alternate products)

Leading up to Qwiser Search: Automatic Indexing - The Foundation of Effective Search

Qwiser's accurate search results rely on an optimally indexed product catalogue. Qwiser works with copies of online stores' catalogues and automatically indexes their products into a pre-existing tree-like hierarchy, developed especially for the store's domain. The hierarchy used by Qwiser is rich and nuanced, and built so that fine distinctions that shoppers use to describe what they want in their queries are understood by the search engine, and are able to be quickly retrieved.

The hierarchy used in the catalogue is based on Celebros' vast knowledge of many different retail domains. The ways in which categories are broken down facilitate Qwiser Salesperson's precise and logical response to queries. For example, it is likely that a shopper requesting a shirt would want to see t-shirts too. However, a shopper requesting a t-shirt would not want to see all other types of shirts that the store carries. Qwiser indexing makes these distinctions between product categories clear.

Stores are able to make manual adjustments and revisions to the product categories in the catalogue. The detailed hierarchical product classification also lays the ground for creation of precisely-defined rules in Qwiser™ Salesperson Desktop that influence how results are presented to shoppers, and at the same time, simplify structuring of all types of sales promotions.

Query Analysis and Product Retrieval

Understanding Shoppers: Qwiser Salesperson's Concept-based Method of Query Analysis

Celebros' patent-pending Query Analysis process analyses each query submitted by shoppers using advanced algorithmic and linguistic tools that extract its conceptual and semantic information.

Analysis of the query and retrieval of results takes place within just a few milliseconds. This first step ensures that all queries are treated suitably. Qwiser Salesperson immediately retrieves accurate and relevant results for queries that are clear (i.e., blue jacket, 12" pan).

Qwiser's intelligence stems, in part, from knowledge that is stored in Qwiser™ Salesperson Desktop. In the Query Analysis stage, this includes information about the different types of queries that are associated with a particular concept, enabling Qwiser to, for example, retrieve identical results for these different queries: ¾ pants, capris, toreador pants, pedal pushers, and mid calf pants.

Interacting with Shoppers: Qwiser Salesperson's Method for Refining Results

When queries are vague or generate too many results for the shopper to comfortably review (i.e., Steven Spielberg, ring), the Query Analyser extracts any useful information from the query and forwards it downstream to additional algorithms and techniques that establish clarity.

Ranking Results Relevancy using Machine Learning

Queries are forwarded to the Ranker, which retrieves and then ranks all the relevant items in the store on how well they match the query. On sites that have implemented Analytics, real time information about shoppers' current product preferences is used to dynamically rank the products as to their degree of relevance to each shopper's specific query. This latter type of information is gathered through Qwiser's machine learning feature that works in conjunction with the Ranker. Machine learning keeps the rankings dynamic and fresh, and in tune with the changing interests of shoppers, so they are always most relevant.

Dynamic Refinement with the Help of Machine Learning

Within milliseconds, the Ranker forwards its list of ranked relevant products along with their calculated probabilities to the Best Question Finder (BQF).

Shoppers participate in the refining process through their interaction with Qwiser Salesperson. The BQF presents shoppers with relevant product categories from which to choose. Each product category is presented along with a photo representing typical items, making the interactive process intuitive and obvious to shoppers.

Qwiser's process of dynamic refinement takes place through interaction with shoppers. Product categories change and re-organize in real-time in response to the selections they make. Each set of refinement options is calculated anew by the Ranker and as a result, the most efficient path to the desired product is created.

Here too, machine learning has a significant impact on the presentation of refining questions and their answers. The refining questions that have proven to be most relevant in ferreting out shoppers' desires are shown to customers first. This is true of the options they choose from as well. The products clicked most often are shown to shoppers first, for example, flower-patterned plates as opposed to solid colour plates.

The conversation-like interaction with shoppers is concise and efficient and is among the most obvious of Qwiser Salesperson's features that mimic the interaction shoppers have with professional, live salespersons in bricks and mortar stores.