All work
Case 02 / 04 — 2025

Notch

A keyboard-first tracker for product teams, built on one measurable promise: no interaction that costs more than a hundred milliseconds, and no action that needs the mouse.

Product design Research Design system
ClientSelf-directed concept
Year2025
ScopeProduct design · Research · Design system
Timeline14 weeks · Desktop web & macOS
Notch

The problem

Every tracker a product team touches was built for the person reporting work, not the person doing it. Filing an issue meant a modal, six fields and a mouse trip to a dropdown that could have been two keystrokes.

Speed was the whole thesis, so it had to be a constraint rather than an adjective. I set a hard budget on paper: no interaction over one hundred milliseconds, and every action reachable from the keyboard.

That budget shaped everything downstream — including a list that has to stay responsive above two hundred rows, where naive rendering drops frames on a 2021 MacBook Air.

What we designed

A command palette as the primary surface, a single-line filing bar, and a navigation model where every destination has a two-key path. The mouse remains supported and never required.

A density study in week nine changed the row height, the hit area and the slot order — all things the earlier screens had already assumed. Forty-two tokens and thirty-four components hold the result together across six screens.

Chapter 01

The problem

What a tracker costs the person actually doing the work.

Overview
The problem
Chapter 02

Research

Six teams, and the difference between how they file work and how they read it.

Research method
Findings
Who it is for
Chapter 03

Interaction model

The keyboard map, the two-key rule, and the density study that rewrote the list.

Navigation
Keyboard map
Density study
Chapter 04

Identity & design system

Forty-two tokens and thirty-four components, built to survive a dense dark interface.

Identity
Design system
Chapter 05

The product

Six screens, and the states that carry the weight between them.

Screens — issue list
Screens — issue detail
States
Chapter 06

Outcome

Three rounds, nineteen people, one prototype.

Outcome

What moved

51s → 19sMedian time to file one issue, same task, round three
+58 ptsTask success on “what changed in this cycle” — 34% to 92%
3.4s → 2.1sTo find a named issue among forty, after the density change
0Participants who needed the mouse — was 11 of 19 in round one

What I would do differently

  • 01

    I designed the list before I measured it

    The density study landed in week nine. It changed the row height, the hit area and the slot order — all things the earlier screens had already assumed. Measuring first would have saved a fortnight of redrawing.

  • 02

    Refusing custom fields costs somebody their job

    Two of the six teams used a custom field for a genuine regulatory requirement. “No custom fields” is a good constraint for the product and a real problem for them, and I do not have an answer beyond saying so.

  • 03

    I never tested with a screen reader

    Contrast, focus order and shape-not-colour were specified and checked with tooling. Nobody navigated this with VoiceOver, which means the accessibility board describes intent rather than evidence.

Next case study

Nightjar