EcoOnline supports organisations in maintaining safe working environments when handling hazardous chemicals. This feature allowed customers to replace outdated or expired Safety Data Sheets (SDS) while preserving associated data such as risk assessments, substance details, and compliance records.

Existing problems with the legacy feature:

• Automatic bulk selection without warning led to unintentional updates
• The update process was long, complex, and difficult to navigate
• Users often missed key signals, such as expired SDS indicators

Success metrics extended beyond adoption and migration to focus on what mattered most: users' ability to complete updates efficiently, avoid costly errors, and understand the impact of their decisions. This directly supported platform adoption and customer retention.

• Improved discoverability of expired SDS and feature entry points
• Users can complete single and bulk SDS updates without external assistance
• Reduced risk of accidental SDS updates
• Decreased time to completion compared to the legacy feature
• Strong task-level usability and satisfaction scores (approximately 4/5)

To balance the aggressive timeline with validation needs, I built an early prototype using the internal design system and existing patterns. I mirrored the legacy flow to reduce risk while simplifying key paths to minimize cognitive load.

Understanding the users

The UX Researcher and I conducted moderated a few interviews with customers and identified our primary users as highly knowledgeable professionals responsible for chemical compliance and safety.

Typical roles included: Health & Safety managers, chemists, and lab technicians, and they all had in common the following shared goals:

• Accurately update SDS documentation
• Perform both single and bulk updates
• Manage legal requirements with confidence
• Avoid irreversible or high-risk actions

This framed the feature as a high-risk workflow with low tolerance for error. So, before testing, I facilitated an internal design review with engineering and product to ensure technical feasibility, align stakeholders on scope and risk tolerance, and validate that our testing scenarios reflected real-world usage. This early alignment prevented costly late-stage changes.

In collaboration with the UX Researcher, I conducted seven moderated user tests with an early prototype to contextualise the experience for customers. The sessions surfaced the following key findings:

• 7/7 users found the feature’s name “Search & Replace” confusing
• 7/7 users missed the success confirmation message
• 5/7 users failed to notice expired SDS entry points
• Users expected additional chemical identifiers (CAS no., revision date, article no.)
• Legal requirements differed by country

Based on research insights, I iterated on:

• Clearer naming and terminology
• Stronger visual signals for expired SDS
• Safer bulk-update interaction patterns
• Richer data visibility at key decision points
• Safe, pre-selected default configurations

I worked closely with the UX Researcher to adjust the scenarios, partnered with the copywriter to refine terminology, and maintained ongoing alignment with engineering and product through weekly design reviews.

Pre-launch, I validated the UX improvements by:

• Creating a high-fidelity prototype of the full flow and running an unmoderated test with a broader audience within a short timeframe.
• Partnering with other teams to test alternative feature names through internal surveys (screenshot below).

Post-launch, the project demonstrated clear success:

• Time to complete updates was reduced by 3 minutes
• NPS score of 4,05 indicating strong usability
• No critical incidents reported in the first 30

Additionally, customer support tickets related to SDS updates decreased significantly, and feedback from early adopters indicated increased confidence in the migration process.

With additional time and resources, I would prioritize follow-up research 3-6 months post-launch to identify friction points in real-world usage. I'd also explore progressive disclosure patterns to further reduce cognitive load during bulk updates. Additionally, I would establish a library of reusable safety-critical UI patterns to scale these learnings across other high-risk features in the platform.