YPulse

About YPulse

YPulse is a B2B market research information provider and leading authority on Gen Z and Millennial consumer trends and behaviors. Their subscription-based syndicated research content, comprising over 4000 articles and 15 billion data points, provides companies with insight on how to understand and reach consumers aged 13–39 in North America and Western Europe.

Challenge

YPulse conducted a customer survey that identified three challenges and areas for product development:

1. Customer satisfaction and content engagement: Customers struggled with their vast content library, spending 10–12 hours weekly consuming material, while still missing valuable insights.
2. Delivering quality experiences: YPulse needed to serve diverse audiences with industry-specific insights while balancing customization, speed, and accuracy.
3. Growth amidst expectations: In an increasingly competitive market, YPulse looked to provide a differentiated product experience to capture market share.

"After a couple of years spent in the doldrums of reduced spend on market insights, we felt that we needed a strong new product launch to serve as a lightning rod for our market and customers. "

Dan Coates, President, YPulse

The Solution: Evolving from RAG to Agents

Initially, YPulse deployed a deterministic RAG and Text2SQL pipeline, processing queries through fixed, linear workflows.

With the RAG and Text2SQL pipeline, YPulse:

• Integrated and indexed their massive content repository of articles, surveys, reports, and webinar transcripts, as well as more than 15 billion datapoints, spanning structured and unstructured data collected over the past 12 years
• Rapidly prototyped and tested multiple system architectures – using a combination of open and closed source models with supporting business logic components to create a customized solution for their business use case
• Empowered internal teams to become AI experts – with tooling to support modular AI pipeline design and fast prompt engineering and iteration
• Incorporated valuable feedback with user-friendly shareable prototypes that made it easy for end users to rate answers and judge relevance
• Solved their infrastructure and deployment requirements with Haystack’s built in parsing, embedding and storage in the vector database, as well as managed cloud service to autoscale with usage, leveraging key AWS and NVIDIA services available in the platform
  ‍‍‍‍

YPulse's RAG and Text2SQL pipeline delivered strong results and customer satisfaction. But they were determined not to stand still. With the ability to continually analyze customer interactions, they discovered an opportunity to further improve answer capability and quality.

Working with Haystack’s new agent capabilities, YPulse rapidly evolved their existing AI research assistant, upgrading from a deterministic RAG and Text2SQL pipeline to an agentic architecture that intelligently selects and combines tools to deliver better qualitative and quantitative insights to better serve customers when user queries were ambiguous, multi-threaded, or misspelled.

With Haystack agents, the system gained the ability to:

• ‍Iteratively refine ambiguous queries and route them through three specialized tools (Content Search, Brand Tracker, Behavior & Trends)
• ‍Evaluate whether each tool provides sufficient context—if not, route to the next tool, building stronger responses
• Combine structured and unstructured data into a synthesized response with source citations
• Visually bundle complex pipelines into a single component, making the system easier for internal teams to understand
• Enable future extensibility to scale with new capabilities like chart generation or email delivery as business needs evolve
  
  

Rapid Prototyping Success

"We knew that we'd picked the right partner within a couple of weeks when, after simply loading a subset of our corpus into deepset, we were getting pretty good answers to questions. Over the next 2 months, we worked closely with the deepset team to make the answers even better."

Xavier Vivar, Chief Product Officer, YPulse

Production Grade AI Scaling‍

"Most of our foundational understanding of AI and RAG came as a result of our interactions with the deepset team after our project kickoff meeting. By the launch of our chat interface product, we became fluent in the space. Now, we’ve able to continue to upskill and quickly bring to market new approaches with AI agents."

Xavier Vivar, Chief Product Officer, YPulse

The Impact

YPulse successfully went to market in months and then evolved its AI-powered chat application into a multimodal agentic research assistant, delivering measurable improvements across customer satisfaction, operational efficiency, and revenue.

Key Success Metrics

• 5x ROI within 1 year
• 30% increase in perceived product value, based on YPulse customer survey
• 50% less time required by services teams to provide clients with customized summaries
• < 4 months to launch a new production-ready generative AI agent
• 25% increase in overall user interactions
• 2–3 FTEs saved in AI Consultants and LLM Ops (data integration/system build)
• 4000+ proprietary documents in various formats for optimal use with LLMs
• 15 billion research data points unlocked from their Snowflake database
  

Conclusion

deepset enabled YPulse to build and scale internal AI competency and successively launch AI products that meet evolving customer needs, positioning them as a category leader in Gen Z/Millennial consumer intelligence with AI-powered insights.

"Customer expectations keep rising, and we need to keep pace. With deepset, we can prototype and ship new AI capabilities in weeks, not quarters. We started with RAG and Text2SQL, now run agents in production, and are ready to support whatever use cases come next."

Dan Coates, President, YPulse

‍