How it works

Methodology: how COHORT judges the way you test — not what you test

COHORT doesn’t pass a verdict on your supplement, your protocol idea, or your chances of success. It asks one thing: if this experiment moves your marker, will you actually be able to trust that result?

Why methodology at all?

Self-experiments have a credibility problem, and it isn’t the sample size. It’s that many n-of-1 runs can’t be checked as they are set up: no baseline, no pre-registered definition of success, a protocol that drifts week to week, and a conclusion written by the same person who hoped it would work. The result “felt like it helped” is compatible with everything — and therefore proves nothing.

The fix isn’t more people; it’s more rigor per person. Several of the things that make a clinical trial believable have a single-person equivalent: a defined endpoint, a locked protocol, a success threshold committed to before seeing data, one variable changed at a time. Randomized switching between periods and blinding have equivalents too, but most self-experiments use neither, which is why every result carries a confidence grade.

What every experiment carries

A named target

one marker and the direction it should move

A falsifiable hypothesis

with a concrete magnitude — “≥10%”, not “improves”

A pre-registered success threshold

fixed before any data comes in

A locked protocol

dose, frequency, timing, form — pinned down

A baseline

readings before the intervention, one variable changed at a time

A pre-registered success threshold

The single most valuable decision in the whole experiment. Deciding what counts as success after seeing the data is how every wellness anecdote becomes “it worked”. A number fixed in advance makes the endpoint a pass/fail question instead of a Rorschach test.

A locked protocol and a baseline

A protocol you can’t state precisely is a protocol you can’t hold constant — and a result you can’t attribute. Readings taken before the intervention give the endpoint something to be compared with, and changing one variable at a time means the change has one candidate cause.

After the run: the confidence grade

When the endpoint reading is in, the outcome computes itself: baseline against endpoint, the change, and where it landed against the threshold you pre-registered (where none was set, or it isn’t a number COHORT can read, whether it moved in the direction you wanted). Next to it sits a confidence grade — high, moderate or low — built from three things COHORT already knows:

Adherence: how many of the planned doses your check-offs on Today record.
Baseline: whether there were dedicated readings before the intervention — one reading isn’t really a baseline.
Measurement timing: how long after the intervention ended the endpoint was measured, since an effect can wash out while you wait for the next blood draw.

The rules are fixed and visible, not a black box: whatever held the grade down is named in plain language on the result.

The grade never erases a result. COHORT is your lab notebook, not your gatekeeper. A low grade tells you, honestly, how much weight the result will bear.

What the grade is not

Not an evidence verdict. Whether the literature supports an intervention is a separate question. A well-run experiment on an emerging-evidence intervention is good science; that’s what n-of-1 is for.
Not a prediction. It doesn’t estimate your odds of improvement. It says how confidently you can interpret whatever happened — including a null result, which a well-run experiment turns into genuinely useful information.

The confidence grade is shown on the experiment’s own page, next to the result it qualifies.