The wrong measurement problem

A user has been on the same skincare routine for two years. She has tried azelaic acid, vitamin C, three different retinoid concentrations, niacinamide at 5% and 10%, and a barrier-repair cream her dermatologist recommended. Her redness still flares cyclically. Her texture still gets worse around her period. Her hyperpigmentation has not moved. She is told, every time she asks, to be patient. To give the product six more weeks. To try a different actives sequence.

What none of those products measure: the regularity of her menstrual cycle, her sleep over the past two weeks, the cumulative UV she has absorbed since spring, the composition of her skin microbiome, or the recovery latency of her barrier after she tape-strips it with a sheet mask. The products she is using are not necessarily wrong. The variables that determine whether they will work for her, and when, are not being measured.

Most consumer skincare is solving for the wrong measurement problem. The category measures what is on the surface of the skin and treats the layers underneath as if they are environmental noise. The basic-science literature has been clear for two decades that those layers are the signal, and the disconnect runs across microbiology, dermatology, endocrinology, sleep medicine, and environmental health. The literature is not hard to find; it sits in the major journals. What is unusual is the absence of any consumer product that operationalizes it.

This piece argues that the measurement question is upstream of the product question. If the variables that determine outcomes are not being measured, no product, however well-formulated, can be evaluated against them. The category's standard answer to skin variability is to add another active ingredient. The literature's answer is to measure the system the skin is downstream of.

Why this matters

The cost of measuring the wrong thing is paid in years. Users build routines on the belief that a product is responsible for an outcome, then stay on those routines because the procedure to test the belief does not exist outside research settings. The opposite cost is also paid. Users abandon products that were working for them, because the slow improvement was masked by variation in variables nobody was tracking.

Dermatologists writing inside their own literature have made this case repeatedly. The cosmeceutical industry expands faster than empirical evidence of efficacy can be acquired, and many products make therapeutic claims while avoiding the regulatory framework of pharmaceuticals (Glass, 2020). For retinoids, which are among the best-studied actives in dermatology, a recent review of their use in cosmeceutical formulations concluded that there is a lack of evidence from properly designed clinical trials to support the claimed efficacy of the most commonly used products (Milosheska & Roškar, 2022). The gap is not between what works in research and what works in practice. The gap is between what is known about the molecule in isolation and what is sold in the bottle as if the rest of the user's physiology were irrelevant.

What follows walks through six layers of the system that consumer skincare typically does not measure, with what the basic-science literature actually establishes about each. The argument is cumulative: even where any single layer could be brushed off as marginal, the literature is now describing them as a connected system.

The microbiome layer

The skin hosts a diverse commensal microbiota (bacteria, fungi, viruses, and archaea) that occupies the harsh outer surface and contributes to immune signaling, barrier homeostasis, and pathogen exclusion. The composition varies by body site (the oily face is different from the dry forearm, both are different from the moist axilla), by host genetics, and by environmental exposure. A landmark Nature Reviews Microbiology synthesis described the field's foundational findings and the methodologies, amplicon sequencing and shotgun metagenomics, that established them (Byrd, Belkaid, & Segre, 2018).

The clinical consequence is well-documented. Disruptions of the skin microbiome, what the literature calls dysbiosis, are now linked to acne, atopic dermatitis, rosacea, and several other conditions where the consumer-facing diagnosis is "your skin is sensitive" or "your skin is acne-prone." The underlying biology often involves measurable shifts in specific bacterial populations, the suppression of others, and downstream consequences for the antimicrobial peptides the host secretes in response.

What consumer skincare measures about the microbiome, in 2026: nothing. There are early commercial probiotic skincare products, but these are formulated ingredients rather than measurement instruments. A user does not learn, from any consumer product currently available, what her skin microbiome composition is, how it has shifted over the past six months, or how it relates to her flare patterns. The variable is not in the data the product is acting on.

The barrier layer

The skin's barrier function is measurable. Transepidermal water loss (TEWL) quantifies how much water the stratum corneum is failing to retain. Hydration sensors measure the water content of the corneocyte layer. Sebum-meters measure surface lipid. pH probes, redness imaging, and porphyrin fluorescence quantify other axes. A recent study of 200 women used 104 such noninvasive parameters to stratify sensitive skin into three functional subtypes (barrier-sensitive, neurosensitive, and inflammatory-sensitive) that the conventional dry/oily/combination heuristic conflates into one undifferentiated category (Kuang et al., 2025). Each subtype responds differently to the same product. The study is cross-sectional, focused on oily sensitive skin specifically, and drawn from a single regional population of young Chinese women; whether the same subtypes generalize across age, ethnicity, and climate is a question the broader literature is still answering. The relevant claim here is narrower: a static "skin type" label cannot represent what the instruments can already detect.

The instruments to measure barrier function precisely have existed for decades. They are standard equipment in dermatology research and cosmetic R&D. The phone in a user's pocket does not have them, but it does have a camera, an ambient light sensor, and an internet connection, which is more than enough to estimate several barrier-relevant features and combine them with self-report.

What consumer skincare measures about the barrier, in 2026: a static skin type, declared once during onboarding, treated as constant thereafter. The actual barrier varies meaningfully across weeks and months, with sleep, with cycle phase, with humidity, with recent product changes. A static label cannot represent a dynamic variable.

The hormonal layer

Skin physiology varies with hormonal state, and the variation has been measured directly. A 2023 study of 197 Chinese women aged 18-35 measured noninvasive skin parameters and 16S rRNA microbiome composition stratified by menstrual cycle regularity. Women with irregular cycles showed significantly decreased hydration, significantly increased TEWL, and shifts in the ratio of bacterial phyla on the skin surface compared to women with regular cycles (Ma et al., 2023). The skin physiology, in other words, tracked the endocrine state.

This finding does not require special instrumentation. The cycle phase is something a user already tracks, often in a separate app, often with high adherence. The link between cycle phase and skin variability is felt subjectively by enormous numbers of users. What is missing is the link in the data: a place where the cycle variable lives alongside the skin variables, so that the patterns can be sorted.

What consumer skincare measures about hormonal state, in 2026: nothing, in nearly every product. A small number of apps track cycle separately. A smaller number track skin separately. The integration of the two, what does this user's barrier do in her luteal phase, specifically, is not a standard feature of consumer skincare. The cycle data lives in one app; the skin data lives in another; nothing reads them together.

The sleep layer

The skin's barrier repair work happens in part during sleep. A 2014 study at University Hospitals Case Medical Center took 60 women and stratified them by Pittsburgh Sleep Quality Index score plus self-reported sleep duration: good sleepers (PSQI ≤ 5, 7-9 hours) versus poor sleepers (PSQI > 5, ≤ 5 hours). Both groups were tape-stripped to disrupt their barrier. Both groups were exposed to simulated solar ultraviolet light. Then their recoveries were measured.

At 72 hours after tape-stripping, the good sleepers had 30% greater barrier recovery than the poor sleepers. At 24 hours after UV exposure, the good sleepers had significantly better recovery from erythema. The poor sleepers showed elevated baseline TEWL and higher intrinsic skin-aging scores on the validated SCINEXA assessment (Oyetakin-White et al., 2014). The cohort was small (60 Caucasian women) and not demographically representative, and the specific 30% effect size should not be expected to hold at the level of a given user; what the study established is direction and physiological mechanism, not a transferable point estimate.

What matters here is the rate itself. A measurable physiological capacity, the speed at which the barrier rebuilds after damage, runs roughly a third lower in chronic poor sleepers. A skincare routine that depends on the barrier rebuilding overnight, which most retinoid protocols do, is operating at a different efficacy ceiling in a user who has been sleeping five hours.

What consumer skincare measures about sleep, in 2026: nothing, in any product the author has examined. The signal is already being recorded, on the wrist or the phone, by a sleep tracker the user checks every morning. It never reaches the skincare product.

The environmental layer

Skin aging is multifactorial. The dominant contributor remains solar ultraviolet radiation, but the broader picture, formalized in dermatology as the exposome, includes UVA, UVB, visible light, infrared-A, air pollution, ozone, particulate matter, and tobacco smoke. A 2021 review in Photodermatology, Photoimmunology & Photomedicine synthesized the evidence: broad-spectrum protection against the full solar range plus oxidative-stress mitigation against pollution-induced damage is now considered the standard of care for prevention of extrinsic aging (Krutmann et al., 2021).

The exposome accumulates. A single day of UV exposure is a data point. A summer of weekly outdoor commutes in an urban high-pollution corridor is a different data point. Skin responses are paced against the cumulative load, not the daily reading.

What consumer skincare measures about environmental load, in 2026: at best, today's UV index, pulled from a weather API. The cumulative load over the user's recent weeks is not represented in any product surface the author has found. The data is public, often gridded to within a few kilometers, and almost nobody is reading it into the skin model.

The interaction layer

The strongest argument for measuring the system, rather than any single variable, is that the layers do not behave independently. The literature is small but specific.

The Ma et al. 2023 study cited above did not only document that cycle regularity tracks skin barrier function. It also documented that cycle regularity correlates with measurable shifts in skin microbiome composition, with significant changes in the proportions of Firmicutes, Acinetobacter, Staphylococcus, and Cutibacterium across the three regularity groups, and with Spearman correlations linking specific bacterial taxa to specific physiological parameters (Ma et al., 2023). The hormonal axis and the microbial axis are not separate stories. They are one story, with reciprocal influence.

A 2023 review in Journal of Microbiology extends the argument to the barrier itself. Skin microbiome composition is mediated by, and mediates, the production of antimicrobial peptides at the keratinocyte layer, with cross-talk between bacterial communities, the epidermis, and the immune signaling that decides which bacteria thrive (Lee, Keum, & Sul, 2023). The microbes are not sitting on top of the barrier. They are building it, and being built by it, in both directions at once.

The implication for measurement: a model that tracks any single layer without tracking its neighbors will under-explain what it observes. A user's barrier dip during the luteal phase is not the cycle alone, not the microbiome alone, not the sleep alone. It is the joint distribution. Measuring one variable cleanly is a starting point; measuring them together is the methodological work.

What integration changes

Measurement and integration are separate problems. Many of the variables above can be measured today, individually, by commercial instruments or by APIs that anyone could read. The barrier can be probed with a TEWL meter. The cycle phase can be pulled from any of a dozen apps. The sleep score is on the user's phone. The local UV index is publicly broadcast.

What is missing in consumer skincare is the layer that combines them. A barrier reading on its own does not tell a user what to do with it. A barrier reading combined with cycle phase, sleep score, recent UV burden, and a prior reading from two weeks ago tells her something different: your barrier is dipping in the same place it dipped last cycle, after a week of compressed sleep and high UV. The pattern is what a single reading cannot show.

The methodological consequence is that integration requires longitudinal individual data, a model that can carry variables across time and across layers, and a procedure for separating noise from signal at small individual sample sizes. The previous piece in this series described one component of that procedure: constraint-based causal discovery with bootstrap stability filtering, run on a user's own data, refusing to call observational patterns cause (Mela Field Notes № 01). That procedure is necessary but not sufficient. It assumes the variables that matter are in the data. If consumer skincare measures the wrong things, no inference procedure, however careful, can recover what it cannot see.

Integration is also expensive. A product that integrates six layers carries six times the data-quality risk of a product that measures one. Each new layer is a potential source of confounding, calibration drift, and user friction. The case for integration is that the alternative is what consumer skincare already offers, and what consumer skincare already offers does not work well enough.

The category-scope objection

A reasonable reader will object that the piece is asking too much of a consumer category. Cosmetics are not drugs. Skincare sold over the counter operates inside a regulatory frame that defines it as a wellness product, not a therapeutic one. The variables this piece argues are missing, the microbiome composition, the cycle phase, the cumulative environmental load, are arguably outside the appropriate scope of a cosmetic product. A user whose redness flares with her cycle should see a dermatologist. The cosmetic category is for users with cosmetic concerns, and what it currently measures is what its scope permits it to measure.

The objection is partly correct. Consumer skincare's regulatory ceiling does limit what it can be expected to do, and a piece arguing for medical-grade measurement inside a cosmetic frame is arguing past the category's own boundaries.

What the objection misses is that the boundary is leakier in practice than in regulation. Users with cyclical acne, mild rosacea, post-inflammatory hyperpigmentation, and hormonal pattern changes are the consumer category's actual customer base, not the category's edge case. The user in the opening scene exists, in enormous numbers, and she is not seeing a dermatologist for her flares. She is on her sixth product. The category serves users with medical-adjacent concerns whether or not it claims to. The measurement gap is what defines the difference between what the category currently serves her and what would help.

The implication is that consumer skincare's measurement scope is path-dependent, not principled. It is what the cosmetics industry's pre-modern measurement infrastructure happens to provide, not what users would benefit from. The argument of this piece is not that the cosmetic category should pretend to be a medical one. It is that the variables already measurable today, by the user's other apps, by publicly available APIs, by noninvasive instruments that already exist, can be operationalized inside a consumer-grade frame without crossing into medical claims. The frame is wellness. The measurement is just better.

The Mela claim, calibrated

Mela is built around the integration of the six layers described above. The engine reads visible surface features from photographs, accepts user-reported barrier and irritation signals, accepts cycle-phase data from connected sources, accepts sleep data from the same, and integrates publicly available UV and environmental load. The constraint-based causal discovery procedure described in the previous piece is the layer that asks which of these variables, in this user's data, appears to be doing what.

The claim is not that Mela has solved the measurement problem. The claim is narrower. To Mela's knowledge, no consumer product currently surfaces an integrated system-level measurement model openly to users, with the procedure documented and the limits stated.¹ If a counter-example exists, this publication will issue a correction.

What is in print: Mela Field Notes № 01, documenting the causal-inference procedure. What is in active development: cold-start handling for users in their first two weeks of data, calibration sharpening as individual samples accumulate, regime-change detection that separates a true shift in a user's skin from ordinary day-to-day fluctuation. What is not yet validated: integration performance on real user data at scale. The user base is still small. The integration model's quarterly audits will be reported, including negative results, in this publication as the data accumulates.

Closing

Measuring the right thing is upstream of measuring well. A skincare category that does not measure the system its outcomes are downstream of cannot evaluate its own claims, and a user inside that category cannot evaluate hers. The literature has been telling this story for two decades. Field Notes exists in part to make the case that the story belongs in the consumer-facing product, not only in the journals.

---