Did it actually work?
Now you can find out.
A supplement, a habit, a change in routine — everyone has a story about something that worked. Some of those stories are real. Many are placebo, a lucky week, or an ordinary rebound after a rough patch. Oltris runs an actual test, so you find out instead of guessing.
The honest problem
You can't tell if something works just by taking it and watching.
Expectation alone changes how you feel — that's the placebo effect, and it's real. Add in ordinary good weeks and bad weeks, and it's easy to credit a supplement for a swing that would have happened anyway.
There's a subtler trap too: people usually start something when they feel worst. Almost anything looks like it's working after that, because a bad patch is often followed by a return to normal — regression to the mean, not the pill.
"It worked for me" is good evidence when an effect is large and immediate. It's much weaker for small effects on things that drift on their own — mood, energy, sleep, pain — which is where most supplements live. The fix isn't paying closer attention or being more honest with yourself. It's a different kind of test.
How it works
What makes a test real
-
A real baseline
Compared against your own untreated self, not your memory of how you used to feel.
-
Randomised order
The app decides the sequence, fixed in advance, so you can't unconsciously arrange a flattering result.
-
Washout
The days right after a switch are discarded, so the old condition doesn't leak into the new one.
-
Enough time
A handful of days can't tell a real effect apart from an ordinary good week.
-
Rules set in advance
How long it runs, what counts as the outcome, what would count as an effect — decided before the data exists, so the answer can't be bent to fit what you hoped.
Matching the test to the question
Fast-acting or slow-building — the design has to match.
Test something that clears your system in hours with weeks of switching on and off, and you'll get a clean answer. Test something that builds up over weeks the same way, and you'll get a false "no effect" — because by the time you switch off, it's still in your system. Getting this wrong is the single most common mistake in self-experimentation. That's why the app asks what you're testing before it picks a design.
-
On/off (crossover)
For things that act fast and clear out quickly — caffeine, l-theanine, a habit like an evening walk. Alternates between doing it and not, in a randomised order, each condition compared against your own recent self.
-
Before/after
For things that build up over weeks — creatine, vitamin D, omega-3, iron. A baseline period, then a long period actually taking it, without switching back and forth.
What you can test: supplements, foods and drinks, and habits or behaviours.
What you can't: prescribed medicines (more on that below), and anything where stopping and starting would be unsafe.
It takes weeks, not days. One thing at a time. And some questions can only ever get a weaker answer than others — the app tells you which you're getting, not just what you're getting.
Two worked examples
What this actually looks like.
Both of these are demonstrations of the method, run on the app's own real experiment engine — not personal results. Real numbers, so you can see exactly what a plain answer looks like, including an uncertain one.
A worked example — a demonstration of the method, not a personal result.
On/off · crossover
Does an earlier last coffee change how you sleep?
The question: does stopping coffee earlier in the day lead to better sleep?
The design: an on/off test. Six blocks of three days each, order picked at random by the app, a two-day washout between every switch — thirty days in total. Some blocks ran the rule "last coffee before 2pm," others had no cutoff. Sleep quality was rated each morning.
What the data showed: nights following the early cutoff scored 1.55 points higher on average than nights without one — a consistent gap, not a one-off. (One day did break the rule — a coffee logged after 2pm during a cutoff block. The app flagged it and left it in; one slip in thirty days doesn't change the picture.)
Before 2pm At or after 2pm
Illustrative data from the app's own test fixtures — not a real user's results.
Promising, but not yet conclusive. The gap sits right at the edge of what a month-long test can distinguish from chance. A few more weeks of the same pattern would likely settle it — but a short run only earns "promising," and that's what it gets.
A worked example — a demonstration of the method, not a personal result.
Before/after
Does creatine build up energy over time?
The question: does taking creatine daily raise everyday energy?
The design: before/after — the only design that fits something that accumulates. Fourteen days of nothing, as a baseline, then forty-two days of taking creatine every day. The rules were fixed before day one: what counted as the outcome, and how long it would run.
What the data showed: energy sat flat around the same level for the first two weeks of taking it. Weeks three and four ticked up. By weeks five and six it had settled noticeably higher than the baseline — a gradual rise, not a jump.
Baseline Creatine
Illustrative data from the app's own test fixtures — not a real user's results.
Your average was higher during the intervention phase. That's it — deliberately. This design can't produce a p-value; there's no on/off comparison to test, just a before and an after.
This design can't rule out
- Time or season — you might simply feel different now, for reasons unrelated to creatine.
- Expectation — you always knew which phase you were in, so what you expected may have shaped what you noticed.
- Regression to the mean — a rebound toward normal was likely anyway.
- Anything else that changed in your routine over this time.
The answer
A plain verdict, not a lecture.
Every test ends in one of three places, in plain language:
-
Likely real
The pattern held up consistently enough to trust.
-
No detectable effect
Treated as a real answer, not a failure — you now know something you didn't.
-
Not yet certain
Promising, but the test hasn't run long enough to be sure yet.
Results are yours — exportable any time, to keep or to bring to a doctor.
Oltris doesn't run experiments on prescribed medicines. Track them freely — just don't stop and start one to run a test. Medicines are prescribed for reasons a short self-test can't see, and changing how you take one to satisfy an experiment can be harmful. These tests are for the things you already choose to take yourself.
Coming
We're running these ourselves.
Publishing what we find, as we find it — including the results that show no effect.
Oltris is on the way.
Leave your email and we'll tell you the moment it's ready. No spam, no selling your address — just one message when it matters.
Check your inbox — we've sent a confirmation link. If it doesn't arrive in a few minutes, check your spam folder or try again in case of a typo.
We'll only email you about Oltris. Unsubscribe any time.
We handle your email in line with our Privacy Notice.