Data-rich doesn’t mean decision-ready

Connected health products have gotten great at collecting data. Glucose readings, sleep stages, heart rate variability, recovery scores. The sensors keep getting better, the dashboards keep getting fuller, and the amount of information available to consumers and providers keeps climbing. 

Most of it never changes behavior in the long-term. 

The gap between having data and using it is one of the more consistent patterns we’ve seen in health and behavior work. It shows up in wearables, in chronic condition management, in almost anywhere digital tools have been layered onto human behavior. We seek more information to give us more clarity, but sometimes it just ends up creating more noise.  

A case in point 

We worked on a project involving connected insulin management, where patients and healthcare providers had collected a meaningful amount of data: glucose readings, dosing history, timing patterns. On paper, the information needed to support better diabetes management already existed. 

In practice, it lived across disconnected systems; the insulin pump had its own app, the continuous glucose monitor reported into a second app, and the provider only saw whatever made it into the EMR at the next visit. None of the three talked to each other. Patients couldn’t easily see patterns in their own behavior. Providers couldn’t easily see what was happening between visits. Correlation was hard enough to spot, let alone causation. So, the data sat there, mostly unused, and clinical decisions kept relying on self-reported information and trial and error. 

Patterns that looked erratic often had a clean explanation buried in the gap between systems. A string of afternoon lows that looked random turned out to be insulin stacking: a second correction dose going in before the first had finished working, something you could only see by laying the CGM trace and the pump’s dosing log on the same timeline.  

The goal we worked with our client on wasn’t “how do we collect more data?” It was “how do we make the data matter?” 

Where the real work was 

We spent time with both patients and providers to understand where the disconnect lived. Not what data existed, but what was missing, what signals mattered versus which ones created noise, and how information needed to be shown to support a decision in the moment. 

A few things became clear:

Patients didn’t need every data point. They needed a few that explained why something had happened, in a form they could act on the next time. Blood sugar, insulin doses, and meals all shown together on one screen instead of scattered across three apps.  

Providers had a different problem. Raw logs weren’t much use in a short appointment. What helped was patterns already surfaced, ready to talk through, instead of a data dump they had to interpret cold. 

Instead of scrolling weeks of raw logs, the provider opened the chart to a one-line summary: “3 afternoon lows this week, all following a second correction dose.” The pattern was already spotted and flagged, so the appointment started with a conversation about what to change, not fifteen minutes spent hunting for what happened. 

And different types of dosing behavior needed different interpretations entirely. A pattern that meant one thing for one patient meant something else for another. Treating all the data the same way flattened distinctions that mattered clinically. 

Take overnight lows. In one patient, they were caused by too much long-acting insulin. In another, they were caused by evening exercise. The glucose patterns looked identical, but the root cause was completely different. A generic response tuned for one would have made things worse for the other. 

What came out of it was a sharper dashboard, built around the specific moments where a patient or provider needed to make a call.

The lesson 

It’s tempting to treat data volume as a proxy for product value. More sensors, more metrics, more history. But for the people using the product, whether that’s a patient managing a chronic condition or someone trying to understand their own recovery after a workout, more data without a clear path to a decision just adds weight. 

The teams that get this right aren’t the ones with the most inputs. They’re the ones who’ve done the harder work of figuring out which signals deserve attention, when someone needs to see them, and what decision that information is supposed to support. 

That’s a research question as much as a design one. The data can tell you what happened. But understanding what it meant, and when someone needed to know it, took watching how people actually used it, where they got stuck, and what they ignored.