Your skin barrier is a layer of dead cells about twenty micrometres thick — thinner than a sheet of paper — held together by a lipid mortar. It is the only thing standing between your bloodstream and the room you're sitting in. Almost everything sold as "barrier repair" is trying to do one of three simple things to it, and two of them cost less than a coffee.
The barrier is the stratum corneum: roughly fifteen to twenty layers of flattened, dead cells embedded in a fatty matrix. When people say their barrier is "damaged," they mean that matrix has been depleted or physically disrupted, water is escaping faster than it should, and irritants are getting in more easily than they should. It repairs itself. Your job is mostly to stop interfering.
Bricks and mortar
The standard model, and it's a good one: corneocytes are the bricks, the intercellular lipid matrix is the mortar.
The bricks are dead skin cells that have lost their nuclei and filled with keratin. They aren't inert. Each one is packed with what's called natural moisturizing factor — a mix of amino acids, lactate, urea, and pyrrolidone carboxylic acid, produced when a protein called filaggrin breaks down. These are humectants, sitting inside the cells, pulling in water and holding it there. People with certain filaggrin gene mutations produce less of this, which is one reason atopic dermatitis and ichthyosis vulgaris run in families.
The mortar is where most of the action is. It's a lamellar structure — organised sheets, not a random smear — made of ceramides, cholesterol, and free fatty acids in roughly equal molar proportions. That ratio turns out to matter more than the absolute quantity. Work in the 1990s showed that applying these lipids in the wrong proportions can delay barrier recovery rather than speed it, which is a useful thing to know when you're reading an ingredient list that mentions ceramides and nothing else.
Water leaves your body through this structure constantly. The rate is called transepidermal water loss, and it's the standard laboratory measure of how well a barrier is working. Higher TEWL, worse barrier. It's the number most "clinically proven" barrier claims are based on.
What it's for
Water retention is the part everyone talks about, but it's not the whole job.
The stratum corneum also excludes most of what lands on it — microbes, allergens, and the overwhelming majority of cosmetic ingredients. It hosts a resident microbial population that competes with pathogens. And it participates in immune signalling: when it's disrupted, keratinocytes release inflammatory cytokines, which is why a compromised barrier doesn't just feel dry, it feels reactive. Products you tolerated last month suddenly sting.
That's the mechanism behind most sudden-onset "sensitive skin." It's usually not that you developed an allergy. It's that the barrier thinned, and molecules that previously stayed on the surface are now reaching nerve endings.
What damages it
Four things, in rough order of how often they're the culprit:
Surfactants. Cleansers remove oil. That's what they're for. Harsh ones remove structural lipids along with the dirt. Sodium lauryl sulfate is the reference irritant in dermatology research precisely because it does this reliably enough to be a laboratory tool.
Over-exfoliation. Acids and physical scrubs accelerate desquamation. Used sparingly, useful. Used daily alongside a retinoid and a vitamin C serum, you are removing the barrier faster than it rebuilds. It's the most common self-inflicted version of this, and it tends to be invisible to the person doing it — the routine feels productive right up until it doesn't.
Environment. Low humidity, cold air, indoor heating, hot water, hard water, wind. Winter skin is a barrier problem, not a moisturiser-strength problem.
Time and biology. Barrier turnover takes roughly a month in a young adult and slows considerably with age. Lipid production drops. This is why the routine that worked at 25 stops working at 55 — and why the answer is different products, not more of them.
What repairs it
Three mechanisms. Nearly every barrier product is one or more of these.
Occlusion — physically slowing water loss. Petrolatum is the benchmark and reduces transepidermal water loss by more than 95% when applied thickly enough. It costs almost nothing. Nothing outperforms it. Plant butters, waxes, silicones and heavy oils all work by the same principle to varying degrees.
Humectancy — pulling water into the stratum corneum. Glycerin is the most evidenced humectant in cosmetic science by a wide margin. Urea, panthenol, and hyaluronic acid work similarly. Humectants alone in dry air can make things worse; they need something above them to hold the water in.
Lipid replacement — supplying ceramides, cholesterol and fatty acids to rebuild the mortar. Plausible mechanism, decent evidence, and the most expensive of the three. Whether topically applied ceramides incorporate into the lamellar structure or simply sit on the surface acting as emollients is still debated.
The fourth intervention is the one nobody sells: stop the insult. Cut back to a gentle cleanser and one bland moisturiser, drop every active for two weeks, and most compromised barriers recover on their own. Acute disruption in healthy skin recovers most of the way within about a day. Chronic damage takes weeks.
What we don't know
How much of what people call "sensitive skin" is barrier dysfunction at all is unclear. Some of it is neurosensory — nerve fibre reactivity independent of barrier integrity. Some of it is almost certainly expectation. The measurement tools are also imperfect: TEWL readings drift with room humidity, temperature, and how long the subject has been sitting still, which makes cross-study comparison harder than the marketing claims suggest.
We also don't have good evidence on the optimal ratio of topical lipids for a healthy barrier, as opposed to a diseased one. Most of the ceramide research was done in atopic dermatitis populations, and extrapolating from that to a person with mildly dry cheeks in February is a longer leap than it's usually presented as.
