Apparent age — a leverage model 51 interventions / 7 damage channels

Fifty-one things you can do. Five of them are half of it.

Nearly everyone allocates effort in inverse proportion to what matters — buying serums while sleeping badly, avoiding sugar while ignoring their posture. This is a map of relative leverage: every intervention placed by how many years of apparent age it plausibly buys against how well that claim is evidenced. Run the model over all 51 and five items account for roughly 55% of everything available. The next nine take you to 94%. Everything after that is decimals.

~50%Heritability of perceived age in twin studies. Half the outcome is settled before you act.
80–90%Share of visible facial aging attributed to ultraviolet exposure rather than time.
~100msTime an observer needs to fix your apparent age. It is decided before you speak.

Intervention Years Evidence Ratio
low ongoing cost high ongoing cost or risk ↑ years of apparent age preserved → strength of evidence click any point for the full entry
Section 01 — mechanism

Aging is seven processes running at once

This is the single most useful reframe available. "Anti-aging" is not one problem, and the interventions that look redundant on a shelf are often hitting completely different channels — while three products you own may all hit the same one. Coverage beats intensity. A protocol that touches all seven weakly outperforms one that hammers two.

Section 02 — perception

What observers actually read

People optimize skin because skin is what they see in the mirror at 30cm. But apparent age is assigned by strangers at two to five metres, in well under a second, using channels that mostly are not skin. Bars below are rough signal bandwidth — how much of the age judgment each channel carries at conversational distance.

Section 03 — protocol

Four tiers, in the order they pay

Tier 0 is assumed. If any of it is missing, nothing below it matters — the returns are so lopsided that Tier 1 work on a Tier 0 gap is wasted motion. Tiers ascend in cost and descend in certainty.

Section 04 — frontier

The experimental end, graded honestly

Everything here has a real mechanism and insufficient human evidence. That combination is exactly what makes a field profitable to sell into. Grades reflect how much a careful person should currently believe, not how interesting the biology is.

Section 05 — negative leverage

Things people do on purpose that add years

Underdiscussed, and the failure mode of anyone who takes this seriously enough to overshoot. Each of these is chosen deliberately by someone trying to look younger.

Section 06 — audit

What you already hold, and what's left

Tick what you genuinely do — consistently, not aspirationally. The model discounts overlapping interventions within a channel, so the total will be lower than the naive sum. That discount is the point: your fourth collagen-stimulating input is worth far less than your first.

Captured
0.0yr

Nothing ticked yet. Start with the six highest-leverage items and see how quickly the curve flattens.

Still on the table
0.0 yr
Section 07 — method

How to read the numbers

The year values are modeled estimates, not measurements. No literature reports "tretinoin buys you 3.9 years of apparent age." What exists is: RCTs on wrinkle scores and collagen histology, observational data on perceived-age ratings, twin studies, and photoaging asymmetry cases. The numbers here are a synthesis — a defensible ordering with honest ranges, built so that relative magnitudes are useful even where absolute values are soft. Treat the ranking as reliable and the decimals as decoration.

Baseline assumption: a person adopting the intervention around age 30 and holding it for 30+ years, measured as the gap in perceived age against an otherwise identical person who did not. Starting later compresses the range, sometimes a lot — the starts late field in each entry says how much.

Evidence tiers. 1 mechanism or in-vitro only. 2 small, uncontrolled, or surrogate-endpoint human data. 3 consistent observational evidence in humans. 4 randomized trials, limited in size or endpoint. 5 multiple RCTs or overwhelming convergent evidence.

The stacking discount. Within a damage channel, the nth intervention is credited at 0.62n−1 of its standalone value. Four retinoid-adjacent collagen inputs do not sum. This is why coverage across channels beats depth within one, and why the audit's total will look disappointingly small once you've ticked the obvious things — which is the correct feeling.

What this model cannot see: your genotype, your phototype, your bone structure, your hormonal trajectory, and the specific face you're working with. Every number here is a population estimate applied to an individual, which is the oldest error in health reasoning. Nothing on this page is medical advice; the prescription-only and procedural entries in particular require a clinician who can look at you.