The restart is the skill
Everybody's routine collapses. Illness, a deadline, a holiday, a bad fortnight, a house move. The collapse is not the interesting part and it is not diagnostic of anything.
What determines whether you have a routine over the span of a year is not how long your longest run was. It is how quickly you restart after each collapse. Someone who runs for three weeks, breaks, and restarts in two days will end the year far ahead of someone who runs for eight weeks, breaks, and restarts in March.
Restarting is a distinct skill from starting, and almost nobody practises it deliberately.
Why the obvious restart fails
The instinctive restart looks like this: it is Sunday evening, you feel bad about the last three weeks, and you resolve to come back properly. New system. Earlier alarm. Extra steps to make up lost ground. Full version from day one.
This fails, reliably, by about Wednesday. Three reasons:
It is bigger than the thing that already broke. The routine collapsed at its previous size. Restarting above that size is starting from a harder position with less momentum.
It is powered by guilt. Guilt is a strong initiator and a terrible sustainer. It gets you through two days and then it is gone, and there is nothing underneath it.
It contains a debt. "Making up for" the gap turns the routine into a punishment. Brains do not sustain punishment voluntarily.
Restart at a third
The rule: whatever your routine was, restart at roughly a third of it.
Six morning steps becomes two. Forty-five minutes becomes fifteen. Five days a week becomes three.
It will feel insulting. That feeling is the point — it means the barrier to starting is now genuinely low, which is the only variable that actually matters on day one. You are not trying to achieve anything in week one. You are trying to prove to yourself that the routine can happen at all, which is a claim your recent evidence contradicts.
Hold that reduced size for a full week, including on the days you feel great and want to do more. Especially then. The urge to expand on day three is the same urge that built the version that broke.
Do not settle the debt
There is no debt. This needs saying plainly because the feeling is so strong.
You did not "lose" three weeks of progress that must now be recovered. Routines are not deposits. The learning, the physical setup, the knowledge of what your sequence is — all of that is intact. What lapsed was the running, and the running restarts at zero cost.
Concretely, that means:
- No doubled sessions.
- No earlier start than the original.
- No extra steps as penance.
- No "I'll do the full version to make up for it".
Change one thing, and make it the thing that broke
Before restarting, spend five minutes on one question: what actually ended it?
Not "I lost motivation" — that is a description, not a cause. Look for the mechanical answer:
- A step got bigger. The five-minute thing became fifteen. Shrink it back explicitly.
- A decision crept in. Some step stopped being pre-decided. Pre-decide it again.
- The trigger disappeared. The routine was anchored to something — a commute, a flatmate's schedule, an object in a particular place — and that thing changed. Find a new anchor.
- There was no short version. So the first disrupted day became a skipped day, and skipped days compound. Write one now, before restarting.
- A broken streak did it. You missed one day and the counter went to zero and you never opened the app again. That is a tooling problem, not a you problem — see below.
Change that one thing. Change nothing else. A restart that also introduces a new system is not a restart, it is a new project.
The first three days
Day one. Do the reduced version. It should take under ten minutes. Do not plan, do not journal about it, do not tell anyone. Just run it.
Day two. Same. The temptation to add will already be present. Do not.
Day three. Same. This is usually the day the novelty is gone and the routine has to survive on structure alone. If the reduced version is small enough, it will.
If you miss one of the first three days, that is not a failed restart. Run the reduced version the next day. The only thing that ends a restart is deciding it has ended.
Then grow slowly
After a full week at the reduced size, add one step or five minutes. Then hold for another week.
At that rate you are back to the full routine in about a month. That feels slow while you are doing it and fast compared to the actual alternative, which historically has been restarting from scratch again in six weeks.
Fix the tooling if the tooling caused it
If the thing that ended your last run was a streak counter resetting to zero, changing your behaviour will not help, because the mechanism will do it again.
Use something that grants freezes — protected days earned as the streak grows and spent automatically when you miss — so a missed day costs a freeze instead of the entire record. This sounds like a small feature. It is the difference between a bad week being an interruption and a bad week being a conclusion.
Orbit is built this way: freezes accumulate as your streak grows and are spent for you, so coming back after a gap means picking up a streak that is dented rather than deleted, and the reduced version of a routine can live as its own one-tap routine so restarting takes no negotiation.
The reframe that helps most
Stop counting consecutive days. Count days-out-of-thirty.
Twenty-two out of thirty is an excellent month by any standard, and it contains eight missed days. Under a consecutive-days framing it contains up to eight separate failures and probably ends in abandonment. Same behaviour, completely different conclusion.
The metric you choose determines whether you keep going. Choose one that can absorb a bad week.