Response Accuracy
Improved from 80% to 98%, resolving 90% of user complaints about incomplete answers.
Client Overview
Client: Global Financial Advisory Service
Location: United Kingdom
Industry: Financial Services
Project Background
The client relied on a Retrieval-Augmented Generation system to inform clients of the latest trends and investment strategies. However, the system failed to meet its promises and delivered superficial answers that dismissed critical information. Such inconsistency damaged the audience's trust in the platform, leading to a 25% drop in repeated usage and an estimated yearly revenue loss of $2 million.
Technical Challenges
Incomplete Outputs : The system retrieved data from only one document, leaving relevant information in other sources, resulting in 20% of incomplete responses.
User Dissatisfaction : 30% of users reported dissatisfaction due to missing critical details, leading to declining user engagement.
Information Retrieval : Difficulty aggregating information from multiple sources led to response gaps affecting the platform's credibility.
Technical Implementation
Query Transformation
We choose multi-step query splitting to break down complicated queries into 3–5 sub-questions, facilitating access to financial insights.
Aggregation Mechanism
Our system enabled us to collect responses from multiple documents into a cohesive answer, ensuring the user could access correct information.
Enhanced Algorithms
We improved retrieval algorithms to effectively locate and extract relevant information from various sources, handling queries across 100,000+ documents with 98% accuracy.
Business Benefits
Efficiency
Reduced query handling time by 25%, ensuring an average response time within 3 seconds.
User Retention
Boosted customer satisfaction by 40%, leading to a 20% increase in repeat usage.
Revenue Recovery
Retained users contributed to a revenue recovery of $1.5 million annually.
Key Innovation
We used multi-step query transformations, which allowed the system to break down complex queries into smaller, precise sub-questions smoothly. Paired with a strong aggregation mechanism, the platform presented a cohesive response, massively improving reliability. Resolving incomplete answers with advanced query transformations and improved aggregation techniques enabled us to help the client regain the trust of their audience while enhancing the user experience and positioning themselves as a trustworthy medium to access financial insights. This approach drove significant business growth and customer loyalty. Are you facing similar challenges with your RAG system? Let us collaborate to design innovative solutions that empower your business to achieve unparalleled results. Contact us today!
Solution
We incorporated query transformations into their RAG pipeline. Using multi-step query splitting, we broke down complex user queries into smaller, more manageable sub-questions. Each sub-question targeted a specific aspect of the information spread across several documents, aiming to deliver accurate answers.
Impact Stories: From Vision to Product
Case Study
Pioneering Progress
The client, a market leader in immigration and visa services with a vast network spanning 60 offices in 27 countries, sought to revamp their outdated digital infrastructure.
View Case StudyCase Study
