A self-experiment starts with a change, but it succeeds or fails on the measure. “I want to sleep better” is a real goal. It is not yet an experiment. “I will stop caffeine after 2pm and measure sleep duration plus morning energy” is closer, because there is something concrete to compare before and after.
The best measure is not always the most scientific-looking one. It is the one you can record consistently, every day, without turning the experiment into a second job.
Pick one primary outcome
Start with the question you actually want answered. Then choose one primary outcome that sits as close to that question as possible.
If the question is “does this help me sleep?”, sleep duration might be useful, but morning energy may be closer to what you care about. If the question is “does this improve focus?”, a 1 to 10 focus rating may be less precise than software-tracked minutes, but it may capture the lived outcome better.
One primary outcome keeps the experiment honest. If you track ten things, one of them will probably move just by chance. Then the story becomes tempting: you forget the nine flat lines and talk about the one exciting number.
Use secondary signals sparingly
Secondary measures can help explain the result. They should not replace the main question.
For example:
- Sleep experiment: primary outcome is morning energy; secondary signals are sleep duration and number of wake-ups.
- Focus experiment: primary outcome is deep-work minutes; secondary signals are distraction count and end-of-day fatigue.
- Exercise experiment: primary outcome is workout completion; secondary signals are soreness and resting energy.
- Diet experiment: primary outcome is afternoon energy; secondary signals are cravings and digestion comfort.
Two or three signals can add context. Ten signals usually add fog.
Prefer daily, comparable numbers
A good self-experiment measure has four traits:
- It can be logged daily.
- It means the same thing each time you log it.
- It changes in the time frame of the experiment.
- It is close enough to the goal that you would act on the result.
That last point matters. Step count is easy to measure, but it is not a good primary outcome for a sleep experiment unless movement is the thing you are trying to change. A heart-rate metric can look objective, but if you do not know what you would do with a small change in that number, it may not be the right measure.
Ratings are allowed
People often distrust 1 to 10 ratings because they feel subjective. They are subjective. That does not make them useless.
For personal experiments, a consistent subjective measure can be more useful than a precise metric that misses the point. A daily stress rating taken the same way each night can reveal whether a routine is helping you. A lab-grade number that does not connect to the question may not.
Make ratings better by labeling the scale:
- 1 means “barely functional”
- 5 means “normal for me”
- 10 means “excellent”
The labels do not need to be perfect. They just need to make today’s answer comparable with next week’s answer.
Beware of measures that are too far away
Some outcomes are delayed, noisy, or heavily affected by other factors. Weight, mood, resting heart rate, productivity, and sleep can all move for reasons that have nothing to do with your experiment.
That does not mean you cannot measure them. It means you should be careful about claiming victory from a tiny change. If the outcome is noisy, use a baseline, keep the experiment longer, and note obvious confounders like travel, illness, unusual stress, alcohol, or major schedule changes.
A simple test
Before you start, ask:
If this number improves, would I believe the change helped? If it does not improve, would I be willing to say the change probably was not worth it?
If the answer is no, choose a better measure.
HypoMe is designed around that decision. A good experiment has a baseline, one clear change, and a measurable outcome you can log in seconds. The app can help with templates and AI-designed experiments, but the core idea is simple: decide what would count as evidence before the experiment begins.