Running experiments

An experiment in COHORT is an n-of-1 trial: an intervention, a target marker and a locked protocol, measured against your own baseline.

Before you start

Experiments are curated by the COHORT science team. Find them under Experiments → Find, each with its protocol and the evidence behind it. Start experiment begins the run straight away; a curated experiment can't be scheduled for a later date. Before it starts, COHORT checks that the birth year in your profile shows you're 18 or over, and that you aren't already running the same experiment or another curated sleep experiment: sleep experiments ask the same daily questions, so you run one at a time. If another experiment is already running on the same marker, COHORT warns you first: two interventions on one marker muddy both results. You decide whether to go ahead. Unless your profile records your sex as male, every curated experiment first asks whether you're pregnant or breastfeeding: COHORT's experiments aren't designed for pregnancy or breastfeeding, so a yes means the experiment isn't available to you now. Some experiments also ask, on their page, whether any of a short list of situations applies to you, and which of the medicines or supplements they name you use, under Medicines and Supplements. Each of those lists has a None of these answer, and Start experiment waits until every question is answered. If an answer means the experiment isn't offered to you, the page says so and the experiment can't be started. For the other medicines and supplements you tick, the page shows what its sources say before you start, with caution notes apart from the ones to monitor. Things done on a given day, such as alcohol, caffeine, evening light or a hot bath, aren't asked: the page shows what the sources say about them for the days of the run, before you start and while it runs. Your answers stay in your browser: COHORT doesn't receive or keep them. Check any new intervention with your doctor or pharmacist, especially next to prescription medications.

Some experiments can run in more than one way. Their page shows each design with what it involves, how long it takes and what its result can show; the first is the default and is chosen until you pick another. A design that switches back and forth between your usual routine and the change also asks how many cycles to run: more cycles take longer. The design and the number of cycles you start with stay fixed for the run.

A curated experiment's protocol is fixed: its dose, how often it's taken and how long it runs stay as they were when it started (reminder times can still change), and it can't be moved to the backlog or restarted. You can stop it at any time from its page. While it is dosing you can also pause and resume it (a paused dose is off Today, and the end date doesn't move) and log a day as skipped, taken at less than the dose, or taken as planned; those entries stay recorded on the run.

On Experiments → Find, pick a goal first, such as better sleep quality, to see the curated experiments for it. They're ordered by how closely each fits the goal, the grade of its evidence and how soon it reads out; nothing about you or other members' results changes the order. Your choice of goal stays on the page and isn't saved. Some experiments aren't listed for everyone: melatonin is offered to members in the US only, an experiment on total sleep time can appear only after your baseline is in, and an experiment measured by a wearable is listed only when your connected device records what it measures. A goal with nothing on offer for you isn't shown.

The science team can retire an experiment. A retired experiment leaves Experiments → Find and can't be started. A run already in progress usually carries on to its end. When a retirement stops a run, its page says so and shows the date; everything you logged is kept, and a stopped run has no result.

While it runs

• Doses appear on Today; check-offs build the run's adherence. Some curated experiments begin with baseline nights, with nothing to take, so there is something to compare against: the dose appears on Today when the intervention period starts, and the adherence on its page counts from that day.
• Log a lab reading (the baseline at the start, the endpoint at the end) with Log on the experiment page. A panel you upload in Health → Labs after the run ends files the endpoint for you. Wearable markers need no logging.
• A curated experiment adds the daily questions its protocol needs to your journal, marked *Required* (Today & the check-in explains how they work). You can also attach your own questions to the run (experiment page → "Track during this run") for side effects and subjective outcomes.
• The experiment page shows progress, your readings, and "Journal during this run" — the context that could confound the result.

Finishing

When the duration elapses, the run moves to *measuring* — include the marker in your next blood draw (or log the reading directly) and the outcome computes automatically: baseline vs endpoint and % change, plus a confidence grade based on adherence, baseline quality and measurement timing. How it's graded → Wearable-marker experiments (HRV, sleep) use a trailing 7-day average automatically. A curated experiment's result shows the change without calling it a success or a failure. An experiment with a success threshold or a linked goal also shows whether the threshold was met or, without a threshold COHORT can read, whether the marker moved the way the goal aims.

Scale journal questions tracked during the run get a subjective start→end delta alongside the hard numbers.

After

At the end of a curated run, under its result, you're asked whether to keep the change in your routine. The question is the same whatever the run showed, and you answer it once. If you keep it, a supplement goes on your list in Health → Supplements, and anything else becomes a daily habit in Health → Daily habits; either way it shows on Today as one daily tick, and you can change or remove it there. A change COHORT has withdrawn since your run isn't offered.

Share results on the experiment page publishes a read-only results page under a name you choose — nothing is public unless you do this.