Challenges

Turned messy robot math into something normal people can use. Instead of scripting or Excel geometry, we built a nocode 3D workspace with guided steps that still preserves the level of precision robots need  like down to centimeters and degrees  and runs smoothly on an iPad with gloves on.

The Process

1. Weeks 1–3: Field research & task analysis (shop-floor observations, operator interviews).
2. Weeks 4–7: Rapid wireframes → interactive prototypes of the 3D workspace and parameter panels; weekly usability loops.
3. Weeks 8–10: High-fidelity UI, design tokens, HMI pattern library.
4. Weeks 11–12: Dev handoff, implementation QA, pilot instrumentation (analytics + in-app survey).

The Outcome

• −42% time-to-first-setup (avg. 68→39 mins across 12 tasks, n=18 operators)
• −63% parameter-entry errors (unit/axis mistakes per session)
• −31% support tickets in the first 6 weeks post-launch
• Training time −75% (2h coached → 30m guided onboarding)
• SUS 86 / CSAT 4.6/5 in pilot usability rounds
• Adoption: 3 pilot sites, ~45 daily operators, 400+ sessions, 99.7% crash-free

Behind the Decisions: Reflections & Trade-offs

OnRobot was about taking something that should require an engineer  robot setup and making it feel natural to operators who would rather just get the job done.

The core trade off was precision vs. approachability. Robotics teams needed millimeter-level control; operators needed to see “move this here” and trust that nothing dangerous would happen. We decided to bring the complexity into a visual 3D workspace + parameter panels, instead of stuffing it into forms and manuals. That made setups feel like manipulating a scene, not programming a machine.

Nothing about this could have been solved from a conference room. The most important design work happened on the shop floor: watching how people actually rigged tasks, where they guessed, where they double checked with a colleague. Weekly reviews with robotics, HMI devs, and safety took that raw reality and hardened it into a system that cut setup time, slashed errors, and still passed internal safety scrutiny.

‍

‍

‍