Case Study

Findly

Findly is an AI-powered data analyst for enterprise customers. Its main product, Darling Analytics, is primarily used by commodity traders (i.e., LPH, LNG, Crude) and shipping analysts from industry leaders doing 100B+ in revenue.

Daring Analytics performs sophisticated data analysis and forecasting on up-to-the-minute data aggregations. The platform generates detailed and summarized reports that can be used as one-time analyses or scheduled to automatically re-run under specific market conditions. Its AI research engine integrates internal databases, documents, and real-time web information from Sonar.

Core Components

Multi-source data integration & search: Seamlessly connects internal structured data with external unstructured sources (PDFs, web content, social media) to allow exhaustive search
Natural language analytics: Allows traders to ask complex questions in plain English and receive SQL-powered insights
Automated reporting: Generates scheduled reports and triggers alerts based on market conditions

Findly's technology ensures accuracy, scalability, speed, and security with SOC Type II and ISO 27001 compliance certifications.

Pain point

Commodity traders are faced with workflows that require them to digest vast amounts of data with outdated tools and processes. Many traders turn to overworked analysts to provide them with daily summaries of key data points.

The sheer size and fragmentations of commodities datasets makes the task of merely pinpointing where a certain measurement came from incredibly difficult. Team members rarely remember what database tables contain what information, turning to the data team for even basic questions.

Powered by Sonar Pro

Sonar Pro is integrated within Findly’s core infrastructure to enrich research answers with grounded, real-time, verifiable information from the web.

Findly chose Sonar to ensure:

Up-to-Date Information: Sonar allows Findly to access real-time web search data. Findly uses the search_recency_filter parameter to ensure that all responses are from within the last month. This is critical for time-sensitive queries and commodities trading and shipping clients.

Verifiability and Trust: The Perplexity provider is designed to return objects containing source URLs. This allows Findly to display sources alongside the every generated answer, laying the foundation for user trust and empowering clients to conduct independent fact-checking.

Control over Sources: Findly uses Perplexity’s search domain filter to improve answers by excluding low-quality or undesirable websites from the search results.

Enhanced Answer Quality: Sonar grounds answers in web search results. Findly has discovered this significantly improves the accuracy and reliability of results. Traditional LLM models that rely only on information gained during internal training pose steeper risks of hallucination.

Security Compliance: Sonar runs stateless calls over HTTPS; no customer data is stored or used for model training, satisfying our SOC 2 and ISO 27001 requirements.