Smart Queries

At Insignal, I set out to simplify how users interact with web analytics.

Smart Queries is an upcoming AI-powered feature I’ve been designing that allows users to ask natural language questions like “Which traffic source brought the most users last week?” and receive instant, visual answers — without needing to dig through charts or filters.

This case study details my role in identifying the problem, shaping the vision, and designing a solution.

Context

Insignal is a modern web analytics platform that helps product and marketing teams visualize user behavior through tools like session recordings, heatmaps, and funnels. While the platform offers deep insights, I noticed that many users-especially non-technical ones-struggled to extract specific answers quickly.

I asked myself: What if users could just ask their data a question, and get an answer-instantly and clearly?

That question became the foundation for Smart Queries.

Problem & User Research

To understand the problem clearly, I conducted 1:1 interviews with 8 users and surveyed over 40 others, focusing on product and marketing professionals. My goal was to uncover where they struggled with analytics and how they would ideally like to explore their data.

Key Problems Identified:

Questions I Asked:

What I Heard:

These responses gave me clarity on both the emotional and functional friction users experienced. It also validated the need for a conversational, intuitive layer on top of existing analytics; something that could interpret a question, surface data, and guide the next steps.

My Design Process

Constraints I Worked With:

What I Focused On:

Exploration & Prototyping:

I explored two primary directions to integrate the assistant:

I built interactive prototypes for both design directions and started gathering feedback early. My goal was to understand how users responded to each concept and what helped them feel most supported while exploring their data.

User Feedback:

• Input Bar: Users appreciated its simplicity but often missed the ability to view past queries or follow up on answers. One participant shared:

  It’s clean, but I don’t know what I asked before. I feel like I’m starting from scratch each time.

  
  

• Chat Panel: This approach resonated better. Users liked the conversational flow, the ability to scroll back through previous questions, and the feeling of continuity. One participant shared:

  This feels more like a real assistant. I can think out loud and refine my questions.

Based on this feedback, I moved forward with the chat-style layout, which better supported exploratory workflows and multi-step thinking.

I also prototyped different onboarding approaches: one with a proactive assistant greeting, and another with a passive “Ask me anything” placeholder. Testing showed users were more likely to engage when there was an opening message that gave them a starting point.

During this phase, I also collaborated closely with engineers to understand capabilities and data access limitations. This helped shape what kinds of queries we could confidently support, and where to build in fallbacks.

Final Design

While I cannot share all the internal research insights or low-fidelity prototypes due to company policy, I can show the final direction that is currently being developed.

Low-fidelity sketches and wireframes, which helped me quickly iterate and validate concepts:

High-fidelity designs to define the visual style, layout, and interactions more precisely:

Where It Stands

Smart Queries is currently in development. I’m working closely with engineers to refine and build real-time response flows, and finalize the interaction details.

The first internal demo received strong feedback, with comments like:

“This feels like having a personal analyst”

“I didn’t know we had this data-now I do.”

Final Thoughts

Designing Smart Queries has been a rewarding challenge-not just because of the technical complexity, but because it allowed me to rethink how users ask questions and receive answers.

By focusing on real user pain points, testing early, and designing for clarity, I was able to shape a feature that brings analytics closer to the people who need it-even if they aren’t data experts.