IstoVisio · B2B · Desktop app · Science

Designing 0-1 Annotation Tool for Scientific Research

syGlass

ISTOVISIO · January 2025

This platform helps scientists explore massive, complex datasets and reveal patterns they couldn’t see before.

I was the sole designer, and I owned the problem framing, scope, design direction, and final sign-off. I worked directly with two engineers and collaborated closely on feasibility, performance constraints, and trade-offs.

PROBLEM

BUSSINESS GOAL

Scientists used syGlass for data annotation, but couldn’t complete publication-ready workflows in syGlass. They had to switch tools at the final step, breaking their flow and increasing the risk of churn as competitors offered complete workflows.

Increase retention and renewal confidence by enabling scientists to complete their full annotation workflow inside syGlass, positioning the platform as an all-in-one scientific workspace rather than one step in a multi-tool pipeline.

DESICION-CHANGING RESEARCH

Because of this insight:

I reframed the problem from “organizing annotations” to “supporting end-to-end, evolving scientific workflows.”

The founding team believed this was a relatively simple request from power users, based on high-level customer feedback asking for better label organization.

To validate this direction, I spoke with five power users from academic labs and walked them through an early prototype built on the founding team’s understanding of the problem.

I learned that the original scope would have shipped something that technically worked, but failed in real scientific workflows.

Success shifted from “did we ship all features?” to “does this still work when the experiment changes?”

Scope expanded in complexity, but narrowed in intent

DECIDING WHAT TO BUILD (AND WHAT NOT TO)

During research, I didn’t just collect feedback. I tested where the workflow was breaking. As I walked users through real projects, I asked one question:

“What needs to exist here so you don’t have to leave syGlass?”

Once I understood what users needed, I had to balance those needs with engineering and time constraints.
At that point, I faced a critical trade-off.

I could design a highly flexible system that covered every edge case, but that would increase technical risk, slow delivery, and impact overall performance.

Instead, I prioritized a constrained solution that fixed the broken workflow immediately without compromising stability or usability, and created a foundation that could scale as user needs evolve.

What I chose to build:

  • Rename masks directly in the product

  • Assign one clear color per mask

  • Create reusable labels and apply them to masks

  • Filter and search masks by name or label

  • Edit colors for multiple masks at once

What I chose not to build (on purpose):

  • Complex group hierarchies

  • Multiple colors per mask or color inheritance rules

  • Automated label imports

These decisions helped us ship a feature that was stable, usable, and useful.

FINAL DESIGN

The final design focused on fixing the exact points where users previously had to leave syGlass.

Instead of maximizing flexibility, the interface guides users through a clear annotation workflow that works.
Defaults are predictable, changes don’t break existing work, and annotations can be reused without reorganizing data.

00 STAGE - MAIN MENU

01 STAGE - NAVIGATE TO MASKS MENU

02 STAGE - MASK CUSTOMIZATION

03 STAGE - CREATE LABEL

04 STAGE - ADD LABEL TO MASK

IMPACT

Before this feature shipped, syGlass supported only parts of the annotation workflow.
After release, users could stay in the product through publication, reducing a key churn risk and making syGlass more competitive as an end-to-end solution.

REFLECTION AND NEXT STEPS

This project wasn’t about adding features. It was about defining the minimum needed to earn user trust and stopping there.

With an MVP in place, the next step is expansion.
Future releases will add advanced capabilities that users requested, such as customization and automation, building on this foundation without compromising performance or usability.