Quantifire, Data Product Design Case Study

Product · B2B Intelligence/2026 — ongoing/In build

Overview

Quantifire’s investor intelligence began as static PDF reports for 100+ clients a quarter. It is now structure, dashboards, visualisation and an AI layer people can explore, interrogate and act on.

The data is dense: 550K+ answers from 6,000+ institutions carrying revenue, exposure, movement and sentiment, each with confidence and timing.

Anything outside that quarter’s report took days to answer. So the work was judgement: deciding what users need first, then building the patterns that carry them from summary to the line items behind it.

My role

Product design lead, sole designer with 5 engineers. Owned structure, dashboards, data visualisation, the design system and Iris Core. In pilot with lead clients.

Decisions

01Turned quarterly PDFs for 100+ clients into a product analysts interrogate
0280+ components specified once, each with interaction states, in light and dark
03Made structure, navigation and interaction behave identically across 50+ views
04Kept overview and line item comparable, so deep exploration never loses its context
05Spent colour only on financial meaning, so the signal that matters reads first
06Iris Core - chat, projects, knowledge base and Market Radar as one workspace

The Product

Conviction performance, with summary detail below

Layered by question

The opening view answers the first question, with depth still available. Detail is layered rather than duplicated, so one surface carries users from overview to line item.Representative data.

Influence relationships compared across regions

Signal over volume

Across 20,000+ engaged investors, momentum and risk are ranked by what needs attention, not weighted equally. Quieter detail sits one interaction away, so the overview stays scannable.Representative data.

Original report pages, Feedback Respondents and Investor Confidence

What it replaced

Seven sections, 20+ pages, quarterly for 100+ clients. Donut charts, stacked tables and uniform cells gave every viewer the same structure, whatever they came to find.Representative data.

Complexity wasn’t the problem. Undifferentiated complexity was.

Attention overview, visitors and engagement

Hierarchy puts changes, signals and actions before supporting detail.

Iris suggestions, analysed and prioritised

Density keeps complex views readable without losing important context.

Exploration keeps deeper detail one interaction away from the main signal.

Conviction respondents, grouped by sentiment

Consistency makes complex information easier to understand and navigate.

Hierarchy

Changes, signals and actions come before supporting detail.

Density

Dense views stay readable through hierarchy, detail intact.

Exploration

Keep detail one interaction away so the main signal stays clear.

Data Visualisation

The goal was not prettier charts, but clearer signals from complex information.

Confidence distribution and performance over time

Money movement, first

Every chart answers the three questions people arrive with: what changed, by how much, against what. Direction and variance read before the legend, so the answer comes from the shape.Representative data.

Tipping point comparison, holder and non-holder intent

Comparable by default

One currency treatment, one number format, one time-comparison pattern, held across portfolio, account and line-item views, so moving through the product never means relearning it.Representative data.

AI Powered by Iris

AI that understands the user, not just the prompt.

Iris connects projects, trusted knowledge and proactive insights into one workspace, embedded in the product rather than bolted on.

Iris, in-product AI panel over the financial view

Chat is a natural conversation grounded in project data and context.

Projects give every project an organised space for conversations, files and AI.

Iris document summary with cited records

Suggestions surface findings, priorities and next steps before you ask.

Knowledge Base is the trusted information Iris uses to understand project context.

Chat

Natural conversation grounded in project data and context.

Projects

Conversations, files and AI organised around each project.

Knowledge Base

The trusted information Iris uses to understand project context.

Suggestions

Findings, priorities and next steps surfaced before you ask.

The Design System

Type & density scale

A tight ramp of data sizes keeps tables, captions and summaries locked in one clear visual hierarchy.

Data components

30+ cards, tables and 20+ chart patterns share one visual language, with empty, loading and low-confidence states built in.

Colour as signal

Colour is spent only on financial meaning and change against its starting point, so gain, loss and exposure read at any density.

Light and dark

The same components and tokens hold hierarchy, contrast and readability across light and dark modes, with no re-specification.

Conclusion

Fixed reporting became a product users can interrogate, and a system dense enough to grow into.

Hierarchy became a product decision, not a styling one. Early versions mirrored the document; later rounds cut a third of the charts so users land on the view answering their first question.

Key decision: kept Quantifire desktop only. It cost reach, but bought the room this data density needs.