Login

Guides

Did it work? How to tell

You are the only person who will ever run a trial on you. Two statistical traps make almost everyone get the answer wrong, and avoiding them takes one decision made before you start.

Last scientifically reviewed: November 22, 2026Editorial review: MDS Labs. External scientific review pending.11 minute read

In brief

  • "I felt better" is not evidence the pill did anything. People given placebos, and openly told they were placebos, still reported a 29% improvement in fatigue.
  • You start a supplement when you feel worst. That is the one condition that statistically guarantees you will feel better afterwards, whatever you took.
  • Pick your measure before the first dose, and pick one you can count. Placebo moves self-reported pain far more than it moves how far someone can walk.
  • Change one thing at a time, and give it the window the trials actually used — 4 weeks for Boswellia, 8 for ashwagandha, 12 for bacopa, about a year for omega-3.
  • Decide your stop rule in advance. A supplement with no endpoint is a subscription, not an experiment.

Why this page exists

Every other page here tells you what happened on average to a few hundred strangers. That is the right way to know whether something works in general, and it cannot tell you whether it works for you. Averages hide the people who responded strongly and the people who did not respond at all, and researchers studying individual nutrition responses say so directly: the limited measurement in group trials cannot be used to infer one person's response 1.

So at some point you try it yourself. That is reasonable. The problem is that the human mind is close to the worst available instrument for the job, and two specific effects will give you a confident wrong answer unless you plan around them.

Trap one: feeling better proves nothing

In a randomized trial, 74 cancer survivors with fatigue were given placebo pills. They were told openly that the pills were placebos, containing nothing active.

They took them twice a day for 21 days and reported a 29% improvement in fatigue severity and a 39% improvement in fatigue-related quality of life, compared with people getting usual care. These were medium and large effects 2.

Read that again, because it is the most useful fact on this page. The improvement did not require believing the pill was real. Knowing it was a sugar pill did not switch the effect off.

This is not a reason to be cynical about your own experience — feeling less tired is a real, worthwhile thing. It is a reason to understand that your sense of improvement cannot, on its own, tell you whether the active ingredient did anything. If you want to know that, the next two sections are the method.

Trap two: the moment you decide to start

This one catches nearly everybody and almost nobody sees it.

Think about when you actually buy a sleep supplement. After a terrible run of nights. When do you start something for your joints? During a bad flare. For focus? In your worst week.

Regression to the mean is the statistical fact that unusually extreme measurements tend to be followed by measurements closer to your average. It makes natural variation look like real change, and it gets stronger when the thing you are measuring is noisy and when you selected the moment to measure based on an extreme value 3.

Which is exactly what starting a supplement at your worst moment does. Your bad week was going to be followed by a more average week whether or not you bought anything. Take the capsule on Monday and you will credit the capsule.

What to do about it

Measure for a week or two before you start. That baseline is the single most valuable thing in a personal trial and it costs nothing but patience. It tells you what your normal range looks like, including how bad a normal bad day is, so that later you are comparing against reality rather than against the memory of your worst night.

If you have already started, you have lost the baseline for this attempt. Note that honestly rather than pretending otherwise.

Pick a measure the placebo cannot move so easily

Not all outcomes are equally vulnerable. A meta-analysis of 21 osteoarthritis trials compared placebo responses across different kinds of measurement and found them significantly smaller for objectively measured outcomes than for self-reported pain: muscle strength and range of motion both showed notably lower placebo responses 4.

The ranking in that analysis is a practical guide. The placebo-to-treatment ratio was largest for self-reported pain (0.44) and smallest for walking time or distance (0.30) 4. The more countable the measure, the less room there is for expectation to do the work.

So when you choose what to track, move down this list as far as your situation allows:

  1. Something counted by a device or a clock — minutes to fall asleep, number of night wakings, weight lifted, reps completed, steps, time to walk a set distance
  2. Something counted by you, objectively — number of days per week with symptoms, number of painkillers taken, number of naps
  3. A rating scale, used consistently — the same 0 to 10 question at the same time of day
  4. A general impression — this is what most people use, and it is the one the placebo moves most

One measure. Chosen before you start. Written down.

Give it the window the trials actually used

The most common way to waste money is to stop too early; the second most common is to carry on forever. The research gives real numbers for both, and they are longer than most people expect.

If you are tryingJudge it no earlier thanMeasure
Boswellia for joints4 weeks 5Pain-free days, stiffness on waking, painkillers taken
Ashwagandha for sleep8 weeks, at 600 mg or more 6Minutes to fall asleep, night wakings
Magnesium for falling asleepA few weeks 7Minutes to fall asleep
Bacopa for mental switching12 weeks 8A timed task, done the same way each time
Creatine for strengthThrough a full training block, with a loading phase 9Weight lifted or reps at a fixed weight
Omega-3 for executive functionAbout 12 months 10Honestly, this one is hard to self-assess

That last row is worth sitting with. Some things cannot reasonably be judged by a personal trial at all, because the effect is too slow and too small to separate from a year of ordinary life. Knowing which those are saves you from pretending.

The five rules

  1. Baseline first. One to two weeks of measuring before anything changes.
  2. One change at a time. Start two supplements together and no result can be attributed to either. This is the rule people break most.
  3. One measure, chosen in advance, as far down the objectivity list as you can manage.
  4. A fixed window, taken from the table above, decided before you start and not revised halfway through.
  5. A stop rule. Write down now what result would make you continue, and what result would make you stop.

If you want to go further, the formal version of this is called an N-of-1 trial: alternating periods on and off the intervention with repeated measurement throughout, which is how researchers establish individual response properly 1. Stopping for a few weeks and seeing whether your measure drifts back is a reasonable amateur version, and it is far more informative than simply continuing.

When to stop

Three honest stopping points, and the third is the one nobody plans for.

The window closed and your measure did not move. This is a result, not a failure. You now know something specific about yourself that no study could have told you.

It worked and the job is done. This applies especially to deficiency correction, which has a defined endpoint and a recheck rather than an indefinite course.

You cannot remember why you are taking it. Most long-term supplement use arrives here eventually — something started for a reason that passed, continued out of habit, never evaluated. If you cannot say what it is doing and how you would know, that is a stop.

What this adds up to

Reading evidence tells you what is worth trying. This tells you what happened when you tried it, and the two together are the whole skill.

The method is unglamorous: measure first, change one thing, pick something countable, wait the real amount of time, decide in advance what would count as working. It takes nothing but a note on your phone and the patience to start two weeks later than you wanted to.

What you get in return is the ability to answer the question for yourself, for anything, for the rest of your life — including all the compounds that have not been studied yet, and the ones where the studies will never be run. That is worth more than any individual product on this site.

Keep reading

Before running a trial on yourself, it is worth checking whether the answer is simpler: Are you actually short of anything? For how long a compound stays in your system, which sets how quickly any effect could appear or fade, see Bioavailability and half-life. For reading the trials themselves, Understanding evidence.

The research-grade version of personalising a protocol is covered in the Prometheus trial, which applies this thinking with laboratory measurement rather than a notes app.

A personal trial is not a substitute for diagnosis. If the thing you are measuring is a symptom that has lasted, the more useful experiment is a doctor's appointment.

Back to Top
Added to cart
Your Cart
0
Close

No products in the cart.