Why you can't tell what your supplements are doing
Six products, started at different times, none ever stopped. Why the question has no answer as things stand, and how to make it answerable.
By Stack Almanac
Magnesium before bed. Vitamin D with breakfast. A B-complex at some point, fish oil most days, a greens powder when you remember. You read the research, you bought the brands people recommended, and you have been consistent for months.
Then someone asks which of them is doing anything, and you hesitate.
The honest answer is that as things stand, the question has no answer. Not because none of it is doing anything. Because nothing about the way you are taking it could tell you either way, and no amount of reading will change that. The useful thing to know is which of the four reasons below is yours, because three of them are fixable in a fortnight.
Why the question has no answer as things stand
Everything started at a different time, and nothing has ever stopped. A stack (everything you take) grows by addition: a podcast, an abstract, a friend, a sale. There is no before and after for any single item, because every start was also a week with other things in it, and nothing has had an off period since.
There is no baseline. Whatever you would say now about how you slept in March, you did not write it down, so your memory of March is the thing being tested rather than the evidence. It is the complaint behind the best line anyone has given us about supplements: very expensive urine, and no idea which bottle to stop.
Too much moves at once. In the fortnight you started magnesium you also had an easier run at work, cut the 4pm coffee and got to bed earlier twice. Any one of those could carry a change in sleep on its own. With four of them in play, nothing in your week can separate them.
And expectation is doing some of the work. In a meta-analysis of 32 insomnia drug trials, covering 82 treatment conditions and 3,969 people, the placebo groups reached 63.56% of the drug response, measured on sleep-lab equipment rather than by asking people how they slept. Pooled across trials in chronic fatigue syndrome the placebo response was 19.6%. The size depends entirely on what is being measured, which is why no single percentage is worth quoting, but the direction is consistent: something improves when you start taking something.
What a fair test on one person looks like
There is a formal design for this and it has a name. An n-of-1 trial is a prospectively planned, multiple crossover trial in a single person: on, then off, then on, then off again. Its reporting standard is CENT 2015, published in the BMJ. Reading it is useful mainly for what it says about your own version. The parts that make the design work are randomising the order, blinding, and repeating the periods. At home you have the third one.
So the home version is weaker than a trial and much better than what you are doing now. Five things make the difference:
- One change at a time. Change two and you have learned nothing about either, which is how a stack gets to its current size in the first place.
- Decide what would count as an answer before you start, write it down, and score it the same way every time. The moment you start, your memory of the baseline stops being evidence.
- Run off periods as well as on periods, and more than one of each. A single on-and-off pair is an anecdote. Several is a pattern.
- Give it the length the research used rather than a week, and expect a null result. That is the most common honest outcome of a test like this, and it is still an answer.
- Keep the rest of the week as ordinary as you can, and write down the things that moved anyway: illness, alcohol, travel, a hard fortnight. Those are what you will argue with at the end.
One part of all this needs no experiment at all. What you take, how much of each thing you are getting in total across every product, and what it costs a month are answerable this afternoon, off the labels. If three products carry the same mineral, you can know that today. We wrote that up separately: how to read a supplement label before you buy and the hidden duplicates in a stack.
What your own record can and can't show
Once you have at least 3 days on a supplement and 3 days off it, each with a check-in, your record puts the two side by side. It shows the difference. It can't prove the cause. Don't stop anything your GP or pharmacist prescribed.
That last pair of sentences is the whole limit. A record kept by one person cannot isolate one supplement, because sleep, alcohol, illness, the season and everything else you take move the same numbers. It is not a read on your symptoms and it knows nothing about your body. What it does is hold what happened, dated, so that in six weeks you are arguing with a record rather than with a memory.
What belongs with your GP or pharmacist
Anything prescribed, any symptom that is new or getting worse, pregnancy, and anything you want measured. Do not run an off period on a prescribed medicine to see what happens. A pharmacist is the easier of the two to reach. The NHS says pharmacists can give advice about medicines, including how to use them and any worries about side effects, and that most pharmacies can help without an appointment. Take your full list with you, or a photograph of the bottles.
If you want the step by step
The method is written out in full in how to tell if a supplement is doing anything, with the three kinds of outcome and what a blood test can and cannot settle. There is also a longer piece on running an n-of-1 on yourself, and one on why a pattern in your own data is not a cause.
Stack Almanac holds what you take and the check-ins you make, dated, and puts them next to each other. Start a record of what you take at Stack Almanac.
If you follow a written routine rather than a loose collection, our page on following a protocol says what the evidence covers and what it does not, and points you at your own record rather than at a verdict.
The facts
- An n-of-1 trial is defined in the CENT 2015 reporting standard as a prospectively planned, multiple crossover trial in a single individual, often described as challenge-withdrawal-challenge-withdrawal.
- Winkler and Rief, Sleep 2015: 32 studies, 82 treatment conditions, 3,969 participants. Placebo groups achieved 63.56% of the drug response on polysomnography.
- Cho, Hotopf and Wessely, Psychosomatic Medicine 2005: the pooled placebo response in chronic fatigue syndrome trials was 19.6%, which the authors noted was lower than the figure commonly assumed for other conditions.
- NHS: pharmacists can give advice about medicines, including how to use them and worries about side effects, and most pharmacies can help without an appointment.
- Only claims on the Great Britain nutrition and health claims register may be used in Great Britain, and no authorised wording links any supplement to sleep except melatonin at 1 mg per quantified portion. The NHS says melatonin is available on prescription only in the UK.
Our judgement, kept separate
- The accumulation is the problem, not the supplements. A stack that only ever grew is a stack that cannot be read, whatever is in it.
- The insomnia figure is the one to sit with. Most of the measured benefit of an actual sleeping pill also turned up in the groups taking nothing, on sleep-lab equipment rather than on a questionnaire.
- The part of this you can settle today, for nothing, is overlap and cost. People skip it because it is less interesting than the question they came with.
- Expect the honest answer to be that you cannot tell, for most of the list, and decide what to keep on the basis of cost and overlap instead. That is a worse answer than the one you wanted and a much better one than a guess.
What we could not verify
- An earlier version of this post was framed around closing a feedback loop so you could tell what was working, and said that with 30 days of data you might notice your sleep scores improved the week you started magnesium glycinate, or that an expensive stack correlated with nothing. That framing is gone. One person's data cannot isolate one supplement, and "is it working" is not a question this or any tracker can answer.
- We could not open the BMJ full text of CENT 2015 today, so the design detail here is limited to the definition and the reporting-guideline record. The randomisation, blinding and washout detail is summarised from secondary sources rather than quoted.
- We could not open the full text of Cho 2005 today, so the 19.6% figure and the trial count come from the published summary and secondary reporting rather than the paper.
- We know of no published trial of the home version of this method, with no randomisation and no blinding, against anything. Treat it as a way to get a better-argued answer, not a validated protocol.
Sources
- CONSORT extension for reporting N-of-1 trials (CENT) 2015 Statement, Vohra and others, BMJ 2015;350:h1738, with the explanation and elaboration paper, Shamseer and others, BMJ 2015;350:h1793: bmj.com. Reporting-guideline entry: equator-network.org
- Effect of placebo conditions on polysomnographic parameters in primary insomnia: a meta-analysis, Winkler and Rief, Sleep 2015;38(6):925-931: pmc.ncbi.nlm.nih.gov
- The placebo response in the treatment of chronic fatigue syndrome: a systematic review and meta-analysis, Cho, Hotopf and Wessely, Psychosomatic Medicine 2005;67(2):301-313: pubmed
- How pharmacies can help, NHS: nhs.uk
- Great Britain nutrition and health claims (NHC) register, Department of Health and Social Care, last updated 19 May 2026: gov.uk. The authorised wordings and their conditions of use are the Annex to Regulation (EU) No 432/2012 as it applies in Great Britain: legislation.gov.uk
- About melatonin, NHS: nhs.uk
- CAP Code section 15 (food, food supplements and associated health or nutrition claims) and section 12 (medicines, medical devices, health-related products and beauty products), Committee of Advertising Practice: asa.org.uk
Checked 10 October 2026. Stack Almanac is a supplement tracker and advisor. It does not diagnose, treat or advise on medical conditions, and nothing on this page is medical advice. We don't sell supplements.
Related reading
Mentioned a supplement you want to learn more about? Browse our ingredient reference for UK reference values, forms, timing and pairings.
Check this across your whole routine
Paste the supplements you take. The free check adds up the totals, shows where products overlap, shows general timing information and works out monthly spend, before you create an account.
Audit your supplement routineOccasional field notes
Useful reading, when we have it.
New evidence guides and meaningful product updates. No daily drip, no supplement sales pitch.