The danger of a good week
The routine has had a good run. Every day, no resistance, and it’s started to feel slightly too easy — which prompts an idea. Make it longer. Add the thing you left out. Move it to the morning, where it really belongs. Buy the equipment. Start the second routine.
Every one of those is a modification, and modifications during a good period have a poor record. The good week is not a signal that the routine can take more; it’s usually a signal that the routine has finally become cheap, and cheapness is the thing you’re about to spend.
Why the good week triggers changes
Capacity is being read off the wrong sample. A good week is a week where conditions were favourable. What you learn from it is what you can do when things go well, which is not the constraint — the constraint is what you’ll do on a bad day, and a good week contains no information about that.
Ease reads as slack. The routine feels effortless, so there must be room. But the effortlessness came from familiarity, not from spare capacity, and it’s specific to this exact version. Change the version and the effortlessness goes with it, immediately and completely.
Success makes the original design look timid. You set a small routine because you were being careful. It’s working, so the caution now looks unnecessary — a conclusion drawn from evidence produced by the caution. This is the same error as looking at a quiet period and concluding the precautions weren’t needed.
And a working routine is boring, which is uncomfortable. A good week means nothing is happening, and there’s a mild itch to make something happen. Modification scratches it. So does chasing novelty, and both cost more than the itch does.
The specific changes people make, and what each one does
Lengthening it. The classic, and the one that most reliably ends established routines, because the floor drifts up with the addition and the bad-day version stops fitting a bad day.
Moving it to a better time. Usually to the morning, on the grounds that mornings are superior. This replaces a trigger that demonstrably works with one that theoretically should. You’re swapping a tested component for an untested one on the basis of a convention. The morning isn’t automatically the answer, and a cue that has been firing for two months is worth more than any theoretical improvement.
Increasing the frequency. The safest of the four, if the sessions stay the same size and the added days have their own trigger. Weekends need cueing separately; deciding to do it seven days a week is a frequency decision and not a mechanism, and without the mechanism the extra days don’t materialise.
Starting something else as well. Immediately puts two builds in play, and new behaviours interfere with each other — often via the same slot or the same anchor that the working routine occupies.
Buying something. Harmless in itself and worth noticing as a symptom: purchasing is the form enthusiasm takes when there’s nothing structural left to do. If the equipment changes where the routine happens, though, it’s a cue change wearing a disguise.
THE TEST — before changing a working routine
· How many consecutive weeks has it run, including at
least one bad one? → fewer than four, or no
bad week → change nothing
· Is the change to what you do, or to when and where?
→ when/where → highest risk.
The cue is the part
that's working.
· Would the new version fit your worst day?
→ no → it will end it
· Is the urge coming from a gap in the routine, or from
the routine being dull? → dull → the change won't
help for longThe one change that’s almost always safe
Not a change to the routine — a change to its documentation.
During a good week you have both the appetite to write things down and accurate information about a routine in working order. That combination is rare, and it produces the two artefacts most people never have:
A written statement of the current version and its floor. Precisely what “done” means today and precisely what the minimum is. This is the reference that makes drift detectable later — without it, the routine’s definition can expand for months and you’ll have nothing to compare against, and every future you will believe the current version is what you always did.
A note on what the trigger is. Written explicitly. Weeks later, when a schedule change kills it, you’ll be able to identify what’s missing instead of vaguely feeling the routine has gone off. Cue failures are hard to spot precisely because the missing thing is a non-event.
Neither changes anything about what you do, which is why they’re safe. Both make future diagnosis cheaper.
If you’re going to change something anyway
You often will, and the appetite isn’t worthless. Constrain it:
One variable. If two things change and adherence drops, the information is unusable.
Not the cue. Whatever else you touch, leave the trigger where it is. It’s the component with the highest replacement cost and the lowest failure rate.
Keep the old floor, in writing. The floor does not move when the normal version grows. This is the single protection that makes growth reversible.
Set a date to check. Four weeks out, ask whether you’re missing more days than before. If yes, revert completely — not partially. Reverting is unpopular because it looks like regression; it’s the only way to establish what caused the drop, and it protects a version you know works.
What this costs
Holding steady during a good period means deliberately not capitalising on a window when you’d have enjoyed doing more. That’s a genuine loss of output, and if you have something you’re working toward on a timeline, it’s a loss that matters.
It also means tolerating a routine that stays modest indefinitely, while feeling capable of more. Some people find that harder than the difficult phase.
What this doesn’t fix
Never changing anything has its own failure: a routine that stays identical for a year can become so uninteresting that it stops for that reason alone. This post is about timing and constraint, not permanent stasis — the safe conditions for adding do eventually arrive.
And a good week doesn’t mean the routine is established. Four consecutive good weeks including one genuinely bad week means something. One good week means the weather was fine.