We are collecting more information about women’s bodies than ever before. But what needs to happen before that information can meaningfully improve their health?

Every week, pitch decks land in my inbox from founders building in women’s health. They address different problems: fertility, menstrual health, menopause, maternal health and metabolic health. Increasingly, however, I am seeing a familiar proposition and it looks like this:

A company has developed an app that collects data from a woman’s wearable. It combines that data with information the woman enters herself: her symptoms, pain, mood, menstrual cycle, medication and lifestyle. Sometimes it also pulls in blood tests or medical records. Artificial Intelligence then brings everything together, connects the dots and tells her what those patterns mean.

The promise is care that is earlier, more continuous and more personalised. But the more of these propositions I see, the more I find myself asking the same question.

Even if we can collect all this data, do we yet know enough to interpret it in a way that is clinically useful?

Collecting data is one thing. Understanding what it means is another. Knowing what to do as a result is something else again. That is where I believe an important opportunity, and a real bottleneck, lies.

This is one part of the much broader women’s health investment opportunity. I see investable opportunities across biotechnology, diagnostics, medical devices, care delivery and digital health. This article focuses on one of them: the infrastructure required to turn data about female biology into evidence that can support a clinical decision.

How women’s health wearables and AI are supposed to work

Wearables have come a long way from counting steps. They can now track sleep, heart rate, heart-rate variability, activity and skin temperature. Women can record menstrual cycles, symptoms, bleeding patterns, pain and mood. Continuous glucose monitors provide another stream of metabolic data, while medical records and laboratory results are increasingly digital.

One measure of how far the category has come is Oura. In its September 2026 IPO filing, the company reported five million paid members and nearly 42 billion hours of longitudinal biometric data. Approximately 72% of its members are women. Oura describes itself as an “always-on health intelligence platform”. The ring captures the data, but the long-term value is expected to come from what the company can learn from it.

We are therefore beginning to build something we have never really had before: a longitudinal picture of what is happening in a woman’s body between medical appointments.

AI appears to provide the final piece. It can process more information than any individual woman or doctor could review. It can identify relationships across thousands, or millions, of observations. It can potentially learn what is normal for one woman rather than constantly comparing her with a generic average.

The logic appears straightforward. We measure what is happening in the body. We find a pattern. We interpret what the pattern means. We decide what to do. The woman receives better care.

But when you look more closely, that chain is not yet intact.

Why wearable data is not yet clinically useful

The first problem is that we are not always measuring the thing we think we are measuring. A wearable may accurately detect a change in temperature, heart rate or sleep, but the signal alone does not tell us what caused it or whether it matters clinically. A rise in temperature might be associated with ovulation, hormonal change, infection or illness. To interpret it, we need context: the woman’s age, reproductive stage, medication, symptoms, medical history and what else was happening at the time.

The relationship between the signal and a particular condition must then be clinically validated before it can support a decision. For example, a wearable may detect a sustained rise in temperature and infer that ovulation has occurred. Before that inference can be relied upon, it must be tested against an accepted clinical reference, such as hormone measurements or ultrasound. We need to know how often it identifies ovulation correctly, misses it or places it on the wrong day.

Only then can the information help a woman or her clinician decide whether to continue monitoring, run a test or intervene. The more consequential the decision, the stronger the evidence needs to be. General guidance about sleep is one thing. An alert that could influence diagnosis, medication or treatment is another.

The final part of the chain is what happens afterwards. Was the interpretation correct? Was a diagnosis confirmed? What action was taken, and did it improve the woman’s health? Without that feedback, the system may continue collecting data without learning whether its conclusions were useful.

With it, something more powerful becomes possible. The measurement informs an interpretation. The interpretation supports a decision. The decision leads to an outcome. That outcome improves how future measurements are interpreted.

That is the loop we need to close. It is also where I believe a genuine data advantage can be built: by connecting measurements to clinical context, decisions and outcomes, rather than simply accumulating the largest possible volume of wearable data.

What women’s health data infrastructure must do

The starting point must be the clinical question, not simply the data already available. What decision are we trying to improve? What biological signal would help answer that question? Are we measuring it directly, or relying on a proxy? In some areas, the existing sensors may be sufficient. In others, progress will require new measurement technologies capable of capturing information that cannot yet be measured reliably.

We also need datasets that reflect female biology. But a dataset is not sufficiently representative simply because everyone in it is a woman. Age, reproductive stage, menstrual-cycle phase, pregnancy, postpartum status, menopause, medication, ethnicity and existing health conditions can all affect what a signal means. The US National Institutes of Health now requires researchers to consider sex as a biological variable because it can influence disease processes, treatment responses and the interpretation of research findings. We need the same discipline within female-specific datasets, or we risk replacing one generic average with another.

The relevant information must also be connected. The wearable may hold one part of the picture, the cycle-tracking app another, the laboratory the blood-test results and the clinician the medical record. Standards such as FHIR can help systems exchange information, but connecting data is not the same as understanding it. Integration only becomes useful when the underlying data is well defined, placed in context and linked to reliable clinical outcomes.

The interpretation then has to be validated in the population in which it will be used and delivered at the point where someone can make a decision. The US Food and Drug Administration’s guidance on digital health technologies asks whether a technology is fit for its intended purpose. A device’s ability to capture a measurement does not automatically make that measurement appropriate for a particular clinical use.

Even a validated insight has to fit into real care. A clinician cannot review months of unfiltered wearable data during a short appointment. If a system creates more alerts but nobody is responsible for responding to them, it may add workload without improving care. The right person must receive the information, trust it, know what action to take and be able to follow what happens next.

That is what it takes to turn data into clinical infrastructure.

Why I see an investment opportunity in women’s health data infrastructure

One of my investment convictions in women’s health is in the infrastructure that turns data about female biology into evidence that can support a clinical decision.

Better measurement is part of that foundation. In some areas, existing sensors can already capture useful signals and the missing layer is interpretation. In others, progress will depend on new measurement technologies that can generate clinically meaningful data where reliable measurement is not currently possible.

But measurement alone is not the investment thesis. The strongest companies will combine world-leading scientific and engineering talent with the ability to validate what they measure, connect it to decisions and outcomes, and build a credible distribution model that supports sustained use.

I do not believe the company with the largest dataset will necessarily have the strongest moat. The more defensible position will belong to the company that can show why a signal matters, what should happen because of it and whether that action improved the outcome.

That is one area where I believe durable value will be created.

Related Posts

Women’s Health Investment Opportunity: What Investors Should Know