Levels case study
Making glucose data readable for beginners
I redesigned how Levels shows glucose, so a beginner can read their day without knowing what a good number is. After a restructure, I was the company’s only designer, working directly with the CEO.
At a glance
About the company Raised a $38M Series A at a $300M valuation in 2022. About 50 people when I joined.
What shipped
- Home ring
- Metabolic panel
- Design system
Levels by the numbers
- App Store rating, from 6,700 ratings
- 4.7★
In this project
The product worked. Getting to it was hard.
I found Levels while researching an integration at Zero. Real-time glucose has immediate cause and effect. Eat a heavy meal and you see it. Sleep badly and it shows. The design team was two people, with room to work on real problems, so I joined.
In the US, a glucose monitor needs a prescription. Before seeing a first reading, members signed up, got approved by a physician, waited for a sensor, installed the manufacturer’s app, paired it, and connected Levels. Then newcomers saw a chart they could not read.
Show the pattern before the number
Every competitor showed the same thing: the current reading and a chart of the day. That works for power users. We wanted to reach beginners.
I redesigned the home screen around a ring of readings: cool while glucose is steady, warm as it climbs. The center names the state in plain words, like Spiking or Recovering, so you can read your day without knowing what a good number is. The full chart sits right below.
141 mg/dL
Your state, in plain words.
Each tick is a reading, cool while steady and warm as glucose climbs. The center says what is happening now, with the chart right below.
- Steady70–140
- Climbing140–160
- Spike160+
One number, one chart.
Every glucose app showed this. Power users could read it. Beginners saw 141 and had no idea whether that was good.
The same idea for bloodwork and food
Every logged meal gets a response score, with its glucose curve and an ingredient breakdown. Spikes become cards members can share. I also designed AI food logging that estimates a meal from a photo.
A glucose monitor gives you one marker continuously. Bloodwork gives you dozens at once. The metabolic panel uses the same language as the ring: each marker in or out of range, with Dr. Rob Lustig explaining the results in a video walkthrough. One design system tied it all together.
Running design alone
About a year in, Levels cut a large part of the company, including both other designers. From then on I was the only designer, working directly with the CEO, across mobile, web, marketing, and internal tools.
Each engineer now owned a project end to end, from finding the problem to launch. Engineers designed in Figma, and designers were expected to commit code. I could not be in every project, so I chose the ones where UX and UI mattered most and worked alongside their front-end engineers. The structure made it hard to stay focused on the right problems, so deciding where my time went became part of the job.
Results
The ring shipped as the new home screen, with the detailed chart one step away. The metabolic panel shipped in the same in-or-out-of-range language, and both used the design system I built.
What I’d do differently
Our CEO asked me to work as a design engineer, shipping inside the same process as the engineers. I was not fluent in code yet, and we only had GPT-3, so I stayed in Figma and paired with engineers on the build.
Today I would say yes. I now take my own designs to production with AI coding agents, and I would take that role on directly.
