Giving Wellness a Face

A methodology clients couldn't see

SonderMind's clinicians developed a new methodology for measuring wellness. Instead of tracking symptoms alone, through questionnaires like the PHQ-9 and GAD-7, it looks across six areas: Emotional, Social, Physical, Cognitive, Environmental, and Spiritual. The scores were already being calculated behind the scenes, but that picture lived in the clinical model, not in anything clients could look at.

The Wellness Picture gives each area a state on a five-step scale, Needs Care, Emerging, Growing, Thriving, Flourishing, with plain-language copy about what that state feels like. It's meant to read as a reflection, not a grade.

Testing the framework with strangers first

I ran an unmoderated concept study with 20 participants who were in therapy or using wellness and self-care apps. All 20 said the six areas felt like a complete picture of wellness, and all 20 understood the scale.

Small copy decisions did a lot of the work. Pairing each area with a one-line explanation, like “values, sense of purpose, inner peace” for Spiritual, kept people from misreading the labels. Some participants liked letter grades because they felt concrete, but grades conflict with the clinical principles behind the product, so we kept descriptive states.

Then asking who it could hurt

Provider interviews raised a question the study hadn't been read for: could the clients most eager to see their picture be the ones least able to absorb a low reading? I re-analyzed all 20 responses against that question and built a persona framework from them.

The answer changed the brief. 18 of 20 wanted to see where they landed, so appetite barely told people apart; resilience did, and nothing in the product can see it. The most self-critical participants were among the most eager, so making the feature opt-in doesn't protect them. Every design choice had to be safe for the most vulnerable reader, not the average one, and notification cadence defaults to conservative.

A launch built to validate it

I wrote the rollout plan around harm rather than engagement. It proposed:

• A phased rollout behind a feature flag, with each phase gated on safety and comprehension signals
• A permanent holdback group, so outcomes can actually be compared against something
• Automatic rollback triggers, like a label that flips two states with no new data
• Treating score improvement as a goal, not a success metric, so there's no pressure to make scores move

Where it stands

The framework, the scale, and the supportive tone held up in the concept study, and the rollout plan was ready for review. My role ended in a company-wide layoff before the feature launched, so this preview covers the work through the study and the plan.