Frequently asked questions about thinking, epistemology, and cognitive tools. 9,738 answers
The most common failure is treating the backlog as a to-do list — feeling pressure to run every experiment on it and experiencing guilt about the ones you never get to. A backlog is not a commitment device; it is a capture and prioritization tool. Its value comes from having more ideas than you can execute, which forces you to choose the highest-value experiments rather than whatever occurs to you in the moment.
Maintain a list of behavioral experiments you want to run.
Open your experiment backlog from L-1113 and identify your top three pending experiments. For each one, write down: the primary outcome variable it measures, the life domain it targets, and the time of day it primarily operates. Now assess independence. Do any two experiments share an outcome variable?
The most common failure is running parallel experiments that share a confounded outcome variable and then attributing the observed change to whichever experiment you were most excited about. You test a new morning routine and a new diet simultaneously, your energy improves, and you credit the morning routine because you read an inspiring article about it — when the dietary change may have been the actual driver.
Run experiments one at a time for clearer results or in parallel for faster iteration.
Design a routine pilot using this four-step protocol. First, define the routine as a behavioral chain (L-1041): list every action in sequence, with each action's completion serving as the trigger for the next. Second, write three to five success criteria that are specific enough to evaluate objectively — not "feel better" but "rate morning energy above 6 out of 10 on at least ten of fourteen days.
Evaluating the pilot before the window expires. You have a bad day on day four — oversleep, skip two links in the chain, feel frustrated — and you conclude the routine is not working. This is the single most common pilot failure. Bad days are not bugs in the pilot; they are test conditions.
Test a new routine for two weeks before deciding whether to adopt it permanently.
Create a seasonal experiment calendar. Take one behavior you currently practice (or want to practice) and design four seasonal variants — one per quarter. For each variant, specify: the behavior, the time of day, the environmental conditions you expect (daylight, temperature, schedule density), the success criteria, and what you will substitute if the primary behavior becomes seasonally impractical.
Treating seasonal variation as personal failure rather than environmental signal. You installed a behavior in June when daylight, warmth, and schedule flexibility aligned perfectly. November arrives, the behavior collapses, and you interpret the collapse as evidence of declining willpower or commitment.
Some behaviors work better in certain seasons — test seasonally.
Recruit one partner — a friend, colleague, or family member — for a shared behavioral experiment. Choose a behavior change you both care about (sleep timing, daily movement, reading, screen reduction, or anything else). Define the experiment together: what you will both do, for how long, and what you will each track.
Turning the collaboration into a competition. The moment partners begin comparing results as a measure of who is doing better rather than as data for mutual learning, the experiment degrades into a performance contest. Competition activates ego defense rather than curiosity, discourages honest reporting of struggles and failures, and biases both partners toward inflating their results.
Run behavioral experiments with a partner or group for shared learning.
When a small experiment works expand it carefully to a larger scale.
Also asked: scaling habits · behavior change scaling · tiny habits expansion · diffusion of innovations behavior · scaling behavioral experiments
Identify one small behavioral experiment you have run in the past six months that produced a clear positive result. Write down the exact conditions under which it succeeded: duration, scope, context, triggers, and any constraints that made it manageable. Now design three progressive expansions — each one increasing the scope by roughly 50 percent along one dimension (duration, frequency, number of domains, or number of people involved).
Treating a successful small experiment as proof that the behavior works at any scale, then jumping straight to the full-sized version without intermediate steps. This is the most common scaling failure because success generates enthusiasm, and enthusiasm overrides the experimental discipline that produced the success in the first place.
Gather every experiment record you have created during this phase — whether in a journal, spreadsheet, notes app, or scattered across documents. If you have fewer than five entries, include informal experiments you remember running even if you did not record them formally. Set aside sixty to ninety minutes of uninterrupted time.
The most common failure is never reviewing at all — running experiment after experiment without pausing to look across them. Each experiment gets its individual assessment, but the meta-patterns that would make future experiments dramatically better remain invisible because you never create the conditions for them to emerge.