AI & product platform

From Raw Feedback to Product Decisions: Designing an AI-Driven Insight Platform

Focused on onboarding, data ingestion, and insight exploration to transform messy qualitative input into structured, decision-ready outputs.

Top-level dashboard surfaces key metrics, risk, and opportunity, connecting insights directly to business impact.

Overview

Univrs is a platform designed to help product teams turn qualitative user feedback into structured, decision-ready insights tied to business impact.

I contributed to early product definition and led design across onboarding, data ingestion, and insight exploration. My focus was on shaping a coherent system that transforms messy, real-world input into clear outputs that teams can act on.

My Role

I partnered closely with a product collaborator who led much of the underlying modeling, scoring, and data processing logic, while I focused on the product experience, interaction model, and interface coherence.

The Challenge

Univrs needed to bridge a difficult gap: turning noisy, disjointed customer data into clear product priorities tied to measurable business outcomes.

The core challenges were:

The design challenge was to create a system that guides users from raw input to confident decision-making, helping them understand what matters, why it matters, and where to act.

Design Strategy

I approached the product as a single continuous loop:

Input → Structure → Insight → Decision

To support this, I focused on four principles:

These principles shaped decisions across onboarding, ingestion, mapping, and insight exploration.

Onboarding

Onboarding is designed to establish context quickly and reduce friction, helping users reach meaningful output as early as possible.

Organization setup introduces key dimensions that shape how insights are structured and interpreted across the system.

Organization setup establishes context early, allowing the system to tailor insights to meaningful business dimensions.

Data Ingestion

The ingestion flow guides users through uploading and validating their data, providing clear feedback and reducing the likelihood of errors.

The experience balances simplicity with transparency, ensuring users understand what has been uploaded and how it will be used downstream.

Empty state encourages upload with clear affordances and guidance.
Uploaded state confirms structure and prepares users for mapping.

Data Mapping

The data mapping interface transforms raw input into structured data. It balances flexibility and clarity, allowing users to classify fields while maintaining context.

Evolving the Mapping Model

Early exploration treated mapping as a direct translation between uploaded fields and system-defined fields, using a two-column layout.

While straightforward, this approach required users to mentally reconcile two separate representations of the data, increasing cognitive load and making it harder to validate mappings with confidence.

This led to a shift toward a unified table model, combining classification, real data preview, and validation into a single continuous view.

Early exploration paired uploaded fields with database targets in a two-column layout—inline previews helped, but users still had to mentally bridge separate representations row by row.
Inline classification and progressive disclosure keep users in context while managing complexity.

Key Tradeoffs

Designing the mapping interface required balancing:

These tradeoffs led to decisions like inline classification, real data previews, and progressive disclosure of mapped fields.

Key Decisions

Because input data was often inconsistent and partially structured
→ Auto-detection was visible and editable, allowing users to validate and override system suggestions.

Because mapping required comparing classifications with real input
→ Example data remained visible alongside field definitions, reducing errors and increasing confidence.

Because switching between separate modes would increase cognitive load
→ Classification and validation were handled inline, keeping the workflow continuous and in context.

Product Structuring

Users define product-level context, enabling the system to connect insights to meaningful business dimensions.

Structured product inputs create a foundation for connecting qualitative insights to specific products and teams.

Insight Exploration

The platform enables users to move from raw feedback to prioritized insights. The interface is designed to support both rapid scanning and deeper investigation.

A tabular layout allows users to quickly compare insights across metrics such as impact, revenue at risk, and theme. A persistent detail panel provides immediate access to supporting context without breaking flow.

Tabular insights paired with a detail panel support both high-level scanning and deep investigation, enabling faster decision-making.

Key Decisions

Because users needed to identify what matters quickly
→ Insights were structured for high-density scanning rather than narrative browsing

Because trust depends on context
→ A persistent detail panel allows users to inspect evidence without losing their place

Because insights must drive action
→ Business impact metrics are surfaced alongside qualitative signals

Outcome

The resulting experience created a more structured and approachable path from raw input to insight:

While early-stage, the product reached a level of clarity and polish that supported demos, stakeholder alignment, and continued iteration toward production readiness.

Reflection

This project was built as a scrappy, fully bootstrapped effort alongside full-time work, with progress unfolding in cycles as our team balanced professional and personal commitments.

These constraints forced a focus on what mattered most: establishing a clear product model and getting to a working, demonstrable system quickly. Rather than over-optimizing individual screens, the priority was creating a coherent end-to-end flow from input to decision.

The v1 product is now live and able to demonstrate a core capability that is often missing in product teams: connecting qualitative feedback directly to measurable business impact.

This experience reinforced the importance of designing systems, not just interfaces. The value of the product comes from how each part connects, turning messy, real-world data into structured, decision-ready outputs.

← Back home