A founder pitched us last spring with a digital health platform that had 240,000 registered users, 4.7 stars on the App Store, and an NPS of 72. These are genuinely good consumer software numbers. They're also largely irrelevant to whether a hospital system should buy the product, and they tell us almost nothing about whether the product is working clinically. The meeting took 90 minutes and most of it was me trying to redirect the conversation toward the question we actually needed to answer: are patients who use this product doing better?
It's not that engagement metrics are worthless. They're not. Daily active users and session length tell you something about whether people find the product useful enough to return to it voluntarily. That matters for consumer-side digital health products in particular. But they're lagging indicators of the thing we care about, and in healthcare, they're easily optimized in ways that don't correlate with clinical impact at all. A gamified medication reminder app can drive daily engagement while having zero effect on medication adherence. The engagement metric looks great. The clinical outcome is unchanged.
Why this keeps happening
Digital health companies inherit their measurement frameworks from consumer tech, and consumer tech has spent two decades perfecting engagement metrics. The tools are mature, the benchmarks are established, and the investors who came from the consumer world know exactly what 4.7 stars and 72 NPS mean. It's a known language. Using it in healthcare pitches is rational from a communication standpoint.
The problem is that healthcare buyers — hospital CMOs, health plan medical directors, employer benefit managers — don't speak this language. Or rather, they speak it but they're not impressed by it. Their decision is not "do users like this app?" Their decision is: "if we deploy this in our population, will we see measurable improvement in A1c, readmission rates, treatment adherence, time-to-diagnosis, or cost per episode?" If you can't answer that question, the engagement data becomes noise.
This mismatch drives a lot of failed enterprise digital health sales. The company genuinely has a product people like and use. They just haven't measured whether it changes anything that the buyer cares about, and they don't have the clinical evidence framework to generate that data quickly.
What outcome metrics actually look like in practice
It depends heavily on the clinical domain, but there are patterns across categories:
| Domain | Engagement Metric (insufficient) | Outcome Metric (what buyers want) |
|---|---|---|
| Diabetes management | Daily check-ins, logging frequency | A1c reduction at 90 days, hypoglycemic events |
| Mental health platform | Session completion rate, NPS | PHQ-9 / GAD-7 score change, crisis intervention rate |
| Post-surgical monitoring | Device wear compliance | 30-day readmission rate, wound complication detection time |
| Chronic disease care coordination | Message response rate, care plan views | ED utilization, preventive care visit completion |
| Medication adherence | App opens, reminder acknowledgment | Medication possession ratio, prescription refill rate |
Notice that the outcome metrics require knowing what happened to patients after they used your product. That means data collection beyond your own platform — linkage to claims data, EHR data, or at minimum structured clinical assessments conducted as part of the care pathway. This is harder to set up than tracking app opens. It requires buy-in from the clinical partner. It takes longer to generate. And it absolutely has to happen before you're credible in a Series B conversation about payor contracts.
The "validated instrument" problem
One shortcut that doesn't work: inventing outcome metrics that sound clinical but aren't validated. We've seen companies report "symptom burden reduction scores" and "wellbeing improvement indices" that were constructed internally and have no connection to any published clinical validation. These make the pitch deck look clinical without providing any of the actual rigor.
Healthcare outcomes research has standard instruments for most major clinical domains — PHQ-9, GAD-7, PROMIS, EQ-5D, SF-36, the Diabetes Distress Scale. If your platform is in a space where these instruments apply, use them. If you're in a space where no standard instrument fits, that's a real problem you need to think through carefully — because you'll need to publish your own outcomes measure, get it validated in the literature, and convince payors it's meaningful. That process takes years.
What this means for how you build
The most practical implication: outcome measurement infrastructure has to be part of the product, not bolted on later. If you're building a chronic disease management platform, the clinical assessment tools — the PHQ-9, the validated disease-specific measures — need to be part of the core product experience, not a separate research module that nobody uses. The outcome data has to flow from normal product use, not from a research overlay that creates friction.
Companies that build this way from the beginning have a data asset at Series A that is genuinely valuable. Not just for investor narrative, but for the sales process with hospital systems and payors. We have portfolio companies that routinely win contracts by walking into a procurement process with 18 months of real-world outcome data from comparable patient populations. That's the version of this that works. The DAU slides are largely irrelevant by the time you're in that room.
