Product Development Process

The development of ANA started with a simple question. How do you make decades of ESOMAR knowledge instantly accessible to every member, regardless of how complex the topic is or where the information sits. ESOMAR’s archive is massive. Reports, guidelines, event papers, publications, regulatory documents, methodologies, and industry frameworks all stored in different formats and written across different eras. Finding the right insight meant digging, scanning, guessing keywords, and opening dozens of PDFs. We needed to replace that workflow with something fast, intuitive, and accurate.

We began by mapping the structure of ESOMAR’s content. What kinds of documents exist. How they relate to each other. How members think about topics. How researchers phrase questions. This helped us design a system that could interpret intent instead of relying on basic keyword matching. The first prototype focused on question understanding and clean document retrieval. It was intentionally narrow, built to validate one thing. Could the system understand a member question and surface the right source within seconds.

From there, we expanded into multi document reasoning. ESOMAR content is often fragmented across several papers, so ANA needed to merge insights into a single answer. We trained the system to read multiple documents at once, extract the relevant pieces, and present them as one unified response with linked citations. This was tested against real research questions from ESOMAR members to make sure the reasoning held up.

We kept the interface simple. A clean layout, quick access to sources, and the ability to switch between summarized answers and full documents. Researchers move fast, so the product had to stay out of the way. Every design decision supported clarity, trust, and speed.

Security was built in early. Members needed to upload their own documents confidently, knowing the system would process them privately and isolate them from the public ESOMAR archive. We built a secure document pipeline that respects organisational boundaries and handles sensitive files safely.

As ANA matured, we refined the model with real usage patterns. Which questions were common. Where the model struggled. How users validated answers. These insights informed continuous improvements to accuracy, tone, and retrieval quality.

‍

‍Technical Approach

ANA is built on an AI architecture designed to understand research language, retrieve the right information from ESOMAR’s archive, and generate answers that members can trust. The core technical foundation combines three layers: intent understanding, retrieval, and verified synthesis.

At the retrieval layer, ANA indexes ESOMAR’s entire library of reports, guidelines, publications, and event papers. Every document is processed, structured, and linked to metadata so the system can locate relevant content quickly and accurately. This ensures that answers always originate from ESOMAR’s verified sources, not external data.

The reasoning layer allows ANA to merge insights across multiple documents. Research topics often span several papers or guidelines, so the system is trained to extract relevant sections, compare perspectives, and synthesize them into a single, coherent answer. Each answer includes citations and direct links to the underlying sources to maintain transparency and trust.

The natural language layer is optimized for research specific terminology. ANA understands industry concepts, methodologies, and global market language, allowing users to ask questions the way they think about their work. This goes far beyond basic keyword search and supports deeper exploration across themes and frameworks.

Security is handled with strict isolation. Uploaded documents are processed within a private environment, stay separate from the public archive, and are accessible only to the user who uploaded them. This protects sensitive content and allows ANA to support individual research workflows securely.

The interface is designed around clean interaction and fast access to sources. Users can switch between answers, citations, document previews, and summaries without friction.

‍

Methodology

Our methodology for building ANA focused on one goal. Turn a complex, decades-old archive into a simple and reliable search experience that researchers trust. Instead of designing a generic AI assistant, we grounded every step in how ESOMAR members actually look for information, interpret insights, and validate sources.

We began by studying how researchers interacted with ESOMAR’s content. Which documents they used most. How they phrased questions. How they moved between topics. These patterns shaped the foundation of ANA’s question understanding and retrieval flow. The first iterations were tested with real queries taken directly from member use cases, allowing us to validate whether the system understood intent rather than just keywords.

From there, we moved into rapid cycles of refinement. Each release focused on increasing accuracy, improving source retrieval, and strengthening the quality of the synthesized answers. When the model missed context or returned incomplete citations, we adjusted both the data structure and the reasoning logic before expanding the feature set.

User feedback played a central role. We observed how members navigated the interface, how they compared answers to source documents, and where they needed more clarity. This helped us refine the interface into a clean, research friendly workflow that feels natural to use.

We validated every improvement against two criteria. Does it save researchers time, and does it preserve ESOMAR’s authority. If a change did not serve both goals, we removed or redesigned it.

This iterative, research driven methodology ensured ANA became more than an AI tool. It became a trusted, intuitive assistant that fits seamlessly into the daily workflow of ESOMAR’s global community.

‍

Conclusion

ANA transformed how ESOMAR members access and use decades of research. By combining natural language understanding, reliable retrieval, and transparent citations, the platform turns a complex archive into a clear and fast search experience. Members save time, find insights effortlessly, and make decisions with more confidence.

For ESOMAR, ANA strengthens the digital value of membership and positions the organisation as a leader in applying AI to research knowledge. It supports researchers, analysts, marketers, and strategists with a tool that fits naturally into their daily workflow.

The result is a trusted AI assistant that preserves the authority of ESOMAR’s content while giving users an entirely new way to explore it. ANA makes the depth of ESOMAR’s global intelligence accessible in seconds.

‍