Analytic Catalyst and Watson Analytics
All work
IBM SPSS · Watson Analytics · 2015

Analytic Catalyst & Watson Analytics — older technology, timeless process

A non-confidential example of end-to-end UX work in a complex technical domain. As lead UX/UI designer at IBM SPSS, I designed a product to make statistical analysis accessible to business users — working with product, development, and SMEs from discovery through user-validated concepts.

A note on this project

"The technology is from 2015. The process is not. This is a non-confidential example of how I work — discovery, collaboration, and user validation in a technically complex domain."

Take yourself back to 2015: statistical software was powerful but impenetrable, designed for experts and largely inaccessible to the business users who needed it most.

Overview

Making statistics accessible to business users

Existing statistical products were very complicated — focused more on test names than on utility. There was a deep gulf between people who understood statistical analysis and those who didn't. This is why we created Analytic Catalyst at IBM SPSS.

As the primary UX/UI designer, my goal was to design a product where business users could make use of statistical analysis without being trained in it — then carry those design concepts forward into the predictive portion of Watson Analytics.

The problem space

A gulf between power and usability

Statistical software of the era prioritised technical accuracy over user experience. Menus were organised by test name — meaningful to a statistician, opaque to a business analyst trying to answer a straightforward question about their data.

The opportunity was to reframe the product around the question the user was trying to answer, not the statistical method used to answer it.

Background: statistical software landscape

Discovery & framing

Discovery framing slide
Framing the problem space — who the users were and what they actually needed from statistical analysis.
User needs framing
Mapping user goals to the kinds of questions they were trying to answer — not the tests they'd need to run.
Ask simple questions slide
The design principle: let users ask simple questions in plain language.
Give understandable answers slide
And give understandable answers — surfacing insight without requiring statistical literacy to interpret it.

Design process

Process step
Working with product, development, and SMEs through mid-discovery to align on scope and direction.
Process step
Concept development — translating statistical concepts into interface patterns a non-expert could navigate.
Design concepts
Design concepts
Refinement
Refinement
User testing
Validating designs through user testing with business users — not statisticians.
User testing results

Watson Analytics

Carrying the work forward into Watson Analytics

The concepts developed for Analytic Catalyst informed the predictive portion of Watson Analytics — IBM's next-generation analytics platform. The core design challenge remained the same: surface meaningful insight from complex data for users who aren't statisticians, without dumbing down the underlying analysis.

Watson Analytics dashboard

Watson Analytics Dashboard visual design
Watson Analytics Dashboard — visual design. The bullseye model became a signature pattern for surfacing predictive strength at a glance.

Predictive models

Watson predictive spiral
The predictive spiral — a visual representation of model confidence that required no statistical knowledge to read.
Watson insight cards
Insight cards — surfacing the most significant findings from a dataset as scannable, actionable summaries.

Decision tree

Visual of a Decision Tree Model
Decision tree model — visualising a complex statistical structure in a way that reveals the story in the data.
Decision Tree Strong Predictors
Strong predictors view — highlighting the variables with the most influence on the outcome.

Outcome

From SPSS to Watson — a process that scaled.

2

Products shaped by this design process — Analytic Catalyst and Watson Analytics

0→1

New product category: statistical analysis for non-statisticians

✓

Validated with users across both products before shipping

The work shipped. The concepts moved from Analytic Catalyst into Watson Analytics and reached a broad audience of business users who would never have engaged with traditional statistical software. The process — discovery with SMEs, iterative concept development, user validation — is the part that hasn't aged.

Reflection

What still holds up

The technology is dated. The problem it addressed — making complex, expert-domain analysis accessible to everyday users — is as relevant as ever. In fact, it's exactly the problem AI is now solving at scale. Working in this space in 2015 gave me a strong foundation for thinking about how to design interfaces that sit between powerful underlying systems and users who don't need to understand how those systems work.

That framing shows up directly in how I think about AI-assisted design today: the interface should make the capability accessible without requiring the user to become an expert in the technology.