The Most Common Experiment Mistake
Most people who try self-experimentation make the same error: they run experiments for the wrong amount of time. Either they give up after a week because they do not see results yet, or they run indefinitely without ever sitting down to evaluate what the data actually says.
Neither approach gives you useful information. Duration is not arbitrary — it depends on the compound you are testing, the outcome you are measuring, and the minimum number of data points you need to distinguish a real signal from random variation.
Related: Try our Experiment Builder to test this yourself. Also worth reading: How to Design a Proper Supplement Experiment and our The Complete Guide to Supplement Tracking.
The Three Speed Categories
Every supplement falls into one of three rough kinetic categories based on how quickly the active mechanism reaches a steady state in your body.
Fast-Acting: 1–2 Weeks
These compounds produce effects on the timescale of hours to days. A useful experiment needs enough days to average out noise, but you do not need months to see whether they work.
- L-theanine: Crosses the blood-brain barrier within an hour. Test morning alertness and sleep latency over 1–2 weeks.
- Caffeine (timing and dose experiments): Effects are immediate. One to two weeks is enough to evaluate sleep impact and afternoon energy patterns.
- Glycine: Many people notice sleep quality effects within days. Two weeks gives you enough data points to calculate a meaningful average.
- Melatonin: Sleep onset effects appear the same night. Run for 1–2 weeks to assess the trend.
Medium-Acting: 4–6 Weeks
These compounds require time to build up in tissue, modulate a physiological system, or produce their effect through chronic adaptation rather than acute action.
- Magnesium glycinate or threonate: While some people notice effects in the first week, meaningful HRV and sleep score changes often stabilize after 3–4 weeks of consistent use. Run for at least 4 weeks.
- Ashwagandha (KSM-66 or Sensoril): Cortisol-modulating effects in studies typically emerge over 4–8 weeks. A 6-week experiment captures the window where effects become measurable.
- Creatine: Loading phase aside, cognitive and performance effects take 3–5 weeks to stabilize. Four weeks minimum.
- Lion's Mane: Nerve growth factor upregulation is a slow process. Effects on cognition and focus are typically reported after 4–6 weeks.
Running a 4–6 week experiment? Split it into two review checkpoints: a quick look at week two to confirm your tracking is working correctly, and a formal analysis at week four or six. Do not make any dosing decisions at week two — just verify your logging is consistent.
Slow-Acting: 8–12 Weeks
These compounds work through mechanisms that unfold over months. Running them for only 4 weeks will often produce a false negative — you will conclude they do not work when you simply did not wait long enough.
- Vitamin D: Serum levels take 8–12 weeks to meaningfully reflect a new supplementation regimen. If you can get bloodwork, get a baseline draw before starting and another at week 12.
- Collagen: Skin and joint outcomes take at least 8 weeks to assess through subjective experience. Structural change takes longer.
- NMN / NR: NAD+ pathway compounds involve metabolic changes that are slow-moving. Most studies that found effects ran 8–12 weeks or longer.
- Berberine: Metabolic markers like fasting glucose and HbA1c require at minimum 8–12 weeks to shift meaningfully. Track daily fasting glucose if you have a CGM; use bloodwork as the endpoint.
Minimum Data Points
Duration is not just about biology — it is also about statistics. You need enough daily data points to calculate a stable average and distinguish a real trend from random day-to-day variation.
As a rule of thumb: 14 days is the floor for any experiment. Fewer than 14 data points makes it nearly impossible to distinguish a real effect from normal variation in your metrics.
| Phase | Minimum Days | Recommended Days |
|---|---|---|
| Baseline | 7 | 10–14 |
| Active (fast compounds) | 10 | 14 |
| Active (medium compounds) | 21 | 28–42 |
| Active (slow compounds) | 42 | 56–84 |
| Washout / ABA return | 7 | 10–14 |
Not sure which category your supplement is in?
Ask: "In published research, how long did studies run before finding effects?" That duration is your minimum. Self-experimentation on shorter timescales cannot reliably detect effects that research took 8 weeks to measure.
When to Extend vs. When to Stop
Extend the experiment if:
- Your baseline period was unstable (high variability in your metrics due to travel, illness, or stress)
- You had more than 3–4 missed logging days during the active phase
- You made a confounding lifestyle change mid-experiment (started a new training program, changed sleep schedule)
Stop and evaluate if:
- You have completed the minimum duration and logged consistently
- You have a clear side effect that outweighs potential benefit
- Your metrics have been stable for 3+ weeks with no trend — flat data is a result
Do not stop early because:
- You think it is not working after 1–2 weeks (too soon for medium or slow compounds)
- You feel great (feeling good is the result, but you still want the quantified confirmation)
- You are bored of logging (this is the cost of getting real answers)
Be the first to try Prova
We're building an app to track whether health experiment design actually works. Join the waitlist.
Logging Through the Full Duration
The experiment duration is only meaningful if you are logging throughout. A 6-week experiment with spotty logging produces less reliable data than a tight 3-week experiment with complete daily records.
Consistency matters more than perfection. Missing one day is fine. Missing a week in the middle of your active phase means you cannot trust the average for that phase. If consistency is a challenge, keep your logging minimal — a few key metrics, logged in under two minutes per day, beats an elaborate system you abandon.