Why Are My Recovery Scores Worse in Perimenopause?

Nothing obvious changed. Training is the same, bedtime is the same, the ring is on the same finger. But the green mornings got rare, amber became the default, and the app has started suggesting rest on days that feel ordinary. The number that used to be encouraging is now mostly a source of low-grade concern.

The usual reading is that something is deteriorating. There is a better one available: your baseline is moving, and the score is still comparing you to a version of you that no longer exists.

What the wearable is actually seeing

Perimenopause is the transition into menopause, and it is a normal life stage rather than an illness. It commonly runs for several years. The defining feature is not that hormones fall in a straight line, it is that they stop being predictable. Estradiol can run high one month and low the next. Progesterone generally trends downward. Cycles shorten, lengthen, or skip.

Wearables do not measure any of that directly. They measure the downstream signals, and those signals move.

Resting heart rate commonly drifts upward. Heart rate variability tends to fall and, more importantly, to scatter. Skin temperature becomes erratic rather than following the clean ovulatory step it used to. Sleep fragments more, and night waking often increases. Respiratory rate ticks up in some people.

Now look at what a recovery or readiness score is built from. Almost universally: resting heart rate, heart rate variability, temperature deviation, and a measure of sleep quality.

Every input moves in the direction the score reads as worse, at the same time, for years.

This is not the luteal problem, and phase matching will not fix it

If you have read why recovery drops in the luteal phase, the mechanism there was a blended baseline. A rolling two or three week average mixes both halves of a cycle, so a normal luteal week has to read as a decline. The fix is to compare this luteal phase to your last one.

Perimenopause breaks that fix, and it is worth being precise about why.

Phase matching assumes the phases repeat. In perimenopause they often do not. A cycle that arrives at day 24, then day 38, then not at all, has no stable phase to match against. The temperature step that marked ovulation is the thing that gets unreliable first, so the very landmark the comparison depends on is the one that goes missing.

There is a second, larger problem. In a regular cycle the baseline is stable and the phase moves around it. Here the baseline itself is drifting. Comparing today to your own average from eighteen months ago is no longer comparing like with like, because the person that average describes has a different resting heart rate and a different HRV range.

The two references that both fail here

A population reference range compares you to strangers. A personal baseline compares you to yourself, which is better, and is the whole argument for personal baselines over population ranges.

Perimenopause exposes the case where even a personal baseline needs a qualifier. A baseline is only useful if it describes the current you. When your physiology is in a multi year transition, a long lookback window is the more flattering statistic and the less accurate one, while a short window loses the ability to see anything slow. Most apps pick one window and never tell you which.

That is the actual complaint behind "my scores got worse". Often the reading did not get worse. The reference got old.

What this costs, in practice

The cost is not that the number is discouraging. It is that people act on it.

The common pattern is training less, for years, at the life stage where the evidence for maintaining strength and load is strongest. Some people conclude they are chronically overtrained and deload repeatedly against a signal that was never about training. Some conclude the device is broken and stop wearing it, which discards the useful signal along with the confusing one.

There is a quieter cost. A persistent amber that means nothing specific trains you to ignore amber. When something genuinely unusual appears, it has to compete for attention with eighteen months of noise, and it usually loses.

What to do with the number instead

Change what you ask it. The useful question stops being "is this good" and becomes "is this different from my recent weeks". Absolute values are the part that is drifting. Short horizon deviations still carry information.

Prefer trend direction over daily scores. A single morning in a period of high variability is close to meaningless. Four weeks of direction is not.

Let the baseline be re-earned. If your physiology changed, a reference built on the last several months describes you better than one built on two years. Expect a rebuild to need time. Our guide on how long a baseline takes to become accurate covers what that period looks like.

Track what actually varies with how you feel. For many people in this stage, sleep continuity and night waking track subjective days better than a composite recovery score does. Consider watching the input that moves with your experience rather than the summary that does not.

Separate the confounders that arrive at the same time. Midlife brings load that has nothing to do with hormones. Alcohol, heat, illness, caregiving and work stress all move these same four signals, and a transition does not exempt you from checking them. The method in why did my recovery score drop still applies.

What would be worth attention

A gradual, noisy shift across this transition is expected, and it is not a finding.

Worth raising with a clinician: heavy or markedly irregular bleeding, symptoms that are disrupting your sleep or daily function, a sharp change rather than a gradual one, or anything that concerns you. Those are clinical questions. They are not questions for a score, and they are not questions for an app. A wearable is not the instrument that establishes where you are in this transition, and any app that implies otherwise is overstating what a wrist sensor can know.

Where NUVARD sits on this

NUVARD is built on the assumption that a baseline is a living thing rather than a fixed reference. The model holds 53 physiological variables across 15 body systems against your own history and updates what normal means for you as it changes, which is the specific failure this page is about.

It carries a life stage setting that includes cycling, hormonal contraception, perimenopause, menopause, pregnancy and not tracking, because these signals behave differently across those states and one mode cannot cover them. Selecting perimenopause tells the model to widen its expectation of variability and to stop treating an unreliable temperature step as a missing input. That is a setting you choose. NUVARD does not detect, identify or confirm perimenopause, and nothing in the product is a substitute for a clinician.

While the history is still short, the reads are hedged and say so, rather than projecting confidence the data has not earned.

Two limits worth stating plainly. A transition measured through downstream signals is inherently noisier than one measured directly, and no amount of modelling removes that. And if you are in the earliest part of this shift, there may not yet be enough separation between your old baseline and your new one for anything to be said with confidence. NUVARD will tell you that rather than fill the gap.

It reads from more than 300 devices and apps, and it sells no hardware, so no device gets favoured. What you take is tested against your own signals, and no measurable signal is a result it will report rather than a gap it will hide.

NUVARD is wellness and health intelligence. It does not diagnose, treat, cure, prevent or detect disease.

NUVARD releases in full on 21 August 2026, and everyone gets the same access on the same day. The waitlist at nuvard.ai is the launch notification list, and it gets you the download link first.