Rebuilding a Medical Analytics Platform at GSK

Medical Insights Analysts were opening AIM, clicking a filter, and going to make coffee while the dashboard loaded. Thirty-plus minutes, on a good day. Most of them had stopped opening it at all. AIM was GSK’s global platform for surfacing trends across trial and field data; in practice, the Power BI implementation was so slow and rigid that the tool was being bypassed. I was the lead and only UX voice on the project. The redesign had been demoted to a second-tier priority. The platform itself was the UX problem, and the fight to prove it would take eight months before a rebuild could begin.

Diagnosing the real constraint

Initial testing produced a SUS score of 34. Iterations within Power BI improved that only to 51, making it clear that this was not a polish problem. The technology itself was limiting performance, flexibility, and the kinds of visualizations analysts needed for real exploratory work.

The eight-month case for a rebuild

Replacing an enterprise platform isn’t a decision UX gets to make alone, and at GSK it wasn’t a decision the organization wanted to make at all. The eight months that followed were a structured exercise in making the case unavoidable.

We started by working directly with the data team in a scrum format to optimize the underlying big-data pipeline, on the theory that Power BI might be salvageable if the data layer was faster. It wasn’t. We brought in Microsoft experts to audit the build and recommend optimizations. Their changes brought load times from roughly 32 minutes down to 28: a real improvement, but not one any analyst would feel. Engineering ran a platform evaluation, with Plotly emerging as the strongest candidate. I ran the parallel UX evaluation on each step: was anything we tried changing the user’s actual experience, and was it pointing toward a product that could succeed for its target user? The answer remained no, on both.

The case for Plotly with a custom front-end was that it combined performance gains with the visual and interaction control AIM actually needed. I built the before-and-after deck, mapped the opportunity space, and circulated it to anyone who’d read it. The deck had one job: make the alternative undeniable.

Rebuilding trust while rebuilding the product

Analyst confidence had eroded over the years AIM had been failing them, and the rebuild on its own wouldn’t fix that. I created a weekly “You Said, We Listened” feedback loop and used live demos in focus groups to show how user input was shaping the product in real time. What actually closed the trust gap was instant load times. The first time a filter responded immediately, the room shifted.

Results and impact

The redesign improved SUS from 34 to 81, dropped load times from 30+ minutes to seconds, increased insight generation speed by roughly 3x, and reached 85% adoption. The work received GSK’s Gold Award in May 2021, the company’s highest internal recognition.

Reflection

This project was one of the clearest product wins of my career and one of the worst organizational experiences of my life.

The product improved dramatically, but the effort required to get there was unsustainable. I was operating in a low-maturity environment with weak trust, poor collaboration, and constant pressure to prove what should already have been obvious. Eight months of structured evaluation to demonstrate that a 30-minute dashboard wasn’t going to be saved by edge-case optimization. The pattern I learned to recognize: organizations will sometimes do everything possible not to change as a way of proving they must.

That experience became a turning point. Design quality alone doesn’t produce strong product outcomes; the organization has to create the conditions that let design succeed. GSK is where I started moving toward Design Operations.