Turning clinical data into metabolic health anyone can read

My role in the project

I was the sole designer on GoodFlip's CGM, owning it end to end: the initial research and success metrics, the onboarding and core metrics dashboard, and the post-launch data that told me what to fix next. Working alongside the clinical team to get the medical detail right and pitching direction to stakeholders, I shipped the MVP, then used what the launch data showed to rework the flow, adding an AI-powered read of each user's day and their glucose spikes.

I got close from three distances

Before I designed a screen, I got close to this product three ways: the domain, the device, and the whole journey.

Inside each view I made one more call, metric by metric. Show it, simplify it, or hold it back.

The question was never how much to show. It was what each person could act on, and what would only scare them.

Then I looked at what happened

Six weeks after launch, the February 2026 data was in. I only claim what it can hold, so here’s what held.

The dashboard held. The onboarding was where reality pushed back.

The device didn’t behave

We started getting a steady stream of complaints about accuracy. People were comparing their CGM reading against a blood glucose monitor reading and treating any gap as a defect. A CGM is built to show trends over time; a glucometer is built for a precise reading at a single moment. That mismatch in expectation was quietly driving our ratings down.

I tried to fix it with an explainer addressing the gap directly in the app. It didn’t land: only 2% of the people who saw it came away satisfied. We ended up solving it upstream instead, by changing the sales pitch so people understood what a CGM does, and doesn’t, tell them before they ever opened the box.

If I had to do it differently, I’d take these into account:

• Explain at the moment of doubt, not inside Help
• Set expectations before purchase and during setup
• Talk to five people who complained before building anything

Two responses, each mapped to a specific failure

Building a better relatability for users with Kaira AI

Let's continue our story from above. Imagine that you are now able to track your blood glucose levels and see glucose fluctuations with every activity (eating, exercising, waking up, etc.) you do. You now have visibility, but your curious self wants to know more.

I figured there had to be a way to answer these questions a user might naturally have. This is where we introduced Kaira on the CGM dashboard, with the following aims:

1. Provide insights to the user that helped them understand why things are happening the way they are, not just what.
2. Create a habit of food logging, allowing users to track meals as well, because what can be tracked can be improved. We used the Hooked Model to achieve this.

The Hooked Model

Heading into the future

This project was really special. It put GoodFlip on the map alongside other health-tech platforms. From a design perspective, I gained experience drafting data-rich experiences that people actually use and enjoy.

A designer is always stuck in a perpetual loop toward perfection, and the team is already taking significant steps toward a more robust experience. Some of my ideas and the work ahead include the following.