When the preview says zero and the data is there

Director of Design

Led the UX side (two designers) alongside Datameer's UI lead, who ran the visual side — a functional-head peer to product and engineering, across a team split between Germany and New York

Five years designing a collaborative analytics platform for data analysts at top US, Canadian and UK banks and insurers, and moving design from executing specs to shaping what got built.

The core challenge

Big data analytics platforms have a built-in UX problem: when a dataset is too large to process in real time, the preview runs on a sample. And a sample can come back empty, not because the data isn't there, but because the sample didn't catch it.

Diagram: a query against the full dataset returns real matches, but a preview built from a small sample can miss all of them and report zero — not a bug, a property of sampling itself. View full size ↗
Fig. 01

Why a sampled preview can report nothing while the data is sitting there.

Users would build a query, see empty results in the preview, and reasonably conclude their data was missing. The data was fine. The preview was misleading them, not through a bug, but through a property of sampled data.

That made it an information design problem, not a visual one: how do you represent uncertainty honestly without making people lose confidence in a tool they depend on every day?

Approach

  1. Customer Research

    With executive support for user research, I joined the sales team on a customer research trip and ran sessions in person with US banking and insurance customers. Where access was harder, iterative screening on usertesting.com found the hard-to-reach personas.

  2. Platform Redesign

    A complete UX and UI redesign of the analytics interface, from information architecture down to the details: column charts, big-data pivot tables, aggregated JSON views, result cards and a comprehensive filter system.

  3. Design System

    Co-built the platform's design system. I owned component behavior and interaction specs across the library, including details like data-table column headers, inspector panels and custom motion easing.

  4. Organizational Elevation

    Datameer was betting on a new product, and that bet needed research and design in the room where it was decided. Research got executive support, and design moved into the roadmap conversations.

Results

The biggest change was organizational. Datameer kept its existing product and went after a new one, and that new product needed research and design where its direction was set. Research got executive backing, design entered the roadmap conversations, and the design system got built.

UX Maturity
Execution function: designers received specs Strategic partner, with executive support for user research
Design System
Co-built, with component behavior and interaction specs owned by UX
User Research
In-person sessions with US banking and insurance customers, plus remote screening for hard-to-reach personas
Sample Uncertainty UX
Visual and text feedback that tells users when a preview is a sample and how it differs from the full query

Want to hear more?

  • Ask me how a preview that reported zero results while the data sat right there turned out to be an information design problem, and how research with bank and insurance analysts got executive backing in the first place.
Get in touch