Mind

How to Stay Consistent With Training

Cameron Usher · Founder, PX

1 September 2026 · 12 min read

12 years in competitive sport; trained alongside the Australian Olympic squad; coached 100+ people one-on-one.

You have started before. Probably more than once. There was a version of you six months ago who had the plan, the kit, the alarm set, and a genuine intention to see it through, and that version was not lying to anyone. Then a week went sideways, and then a fortnight, and the plan became something you used to be doing.

The question of how to stay consistent with training is usually answered as though it were a question about wanting it enough. Get up earlier. Want it more. Find your why. That answer has been tested at large scale, and it is wrong, in a way that is worth knowing precisely, because the true answer changes what you actually do on Monday.

This is a guide to the mental side of training: what actually breaks consistency, what the research says holds, and how a week gets built so that it survives contact with a real life. It is not a motivation piece. Motivation is the one thing here with an evidence base against it.

The problem is not that you don't want it

Start with the number that should change how you think about your own history.

Researchers pooled 25 samples covering 29,600 people to measure the gap between intending to exercise and actually exercising. Of everyone studied, 38.7% intended to be active and were. And 33.0% intended to be active and were not. Expressed as a proportion of everyone who formed the intention, the intention–behaviour gap came out at 47.6%: roughly half of all the people who genuinely meant to train, did not.

Read that again with yourself in it. A third of the entire sample was sitting exactly where you have sat. Not people who never intended to train. People who intended, and whose intention did not survive the week. Intention is a necessary but insufficient antecedent of physical activity. Wanting it is the entry fee. It is not the mechanism.

Which means the standard diagnosis is a category error. If half the intenders fail, the failure is not distributed by character. It is distributed by something structural, and if it is structural then it can be designed around, which is a far more useful thing to know than that you should want it more.

Where it actually breaks: the first four weeks

Consistency does not decay evenly. It has a cliff, and the cliff is early.

Researchers pooled three funded, supervised exercise trials covering 947 participants. These were people in a structured programme, with staff, equipment and appointments, paying nothing. Two hundred and ninety-five of them, 31%, dropped out, and roughly two-thirds of that dropout happened before the first month was complete. The most common reason given was lack of time, at 40% of dropouts. Among the people who did stay, adherence ran above 82%.

That last detail is the interesting one. It is not that people trickled away steadily across the year. It is that a third left during ramp-up, and almost everyone who cleared ramp-up kept going.

The same shape shows up in the wild, at a scale no trial can reach. In an analysis of 389,481 training-app users, including 100,709 beginners, only 10.1% of beginners were still training at week 52, with a median time to dropout of 19 weeks. Experienced users did better, at 25.8%, which tells you something on its own. And the strongest predictor of who was still there at twelve months was not how hard they trained or how ambitious the programme was. It was how consistently they trained in the first 28 days.

So the honest picture is this. The first month is not the easy part before the real training starts. It is the part where the outcome is decided, and it is the part almost every programme designs worst, because ramp-up is when plans are most aggressive and life is least adjusted.

Why motivation cannot be the fix

There is a study that should have ended the motivation industry, and it is worth a paragraph of your attention.

Researchers ran a pre-registered megastudy across 61,293 members of a large American gym chain, testing 54 different four-week interventions designed to increase attendance. Planning prompts, reminders, framing, bonuses, social incentives: the full behavioural-science armoury, run properly, at enormous scale. Forty-five percent of the interventions significantly increased gym visits while they were running, by between 0.14 and 0.40 extra visits a week.

Then they stopped, and the researchers kept watching.

Only 8% of the interventions had a statistically significant effect after the intervention ended, against 2.5% you would expect by chance alone.

That is the whole thing in one result. Motivational input works, in the sense that it moves behaviour while it is being applied, and it does not persist, because it was never building anything that outlives it. Every January is a four-week intervention on yourself, and the data on four-week interventions is now extremely good.

This is not an argument that motivation is worthless. It is an argument that motivation is a fuel and not a structure, and that building consistency out of it is like building a house out of petrol. The thing that survives the intervention has to be something other than how you felt during it.

The five things the mental side is actually made of

"The mental side of training" is a phrase that usually gestures at nothing in particular. In practice, when consistency fails, it fails through one of five specific mechanisms. Each of them is concrete, each of them is common, and each has its own answer.

The argument with yourself about resting. You take a day off and spend it feeling like you have got away with something, or you train through a day you should not have because sitting still felt worse than being tired. Both cost you. It turns out that guilt-driven motivation is genuinely weaker than the alternative, and there is a number attached: across 184 independent datasets, exercise correlated with doing it because it matters to you at about twice the strength it correlated with doing it out of guilt and obligation. The full case, and what to do instead, is in guilt about skipping a workout.

The morning renegotiation. The plan says intervals. You slept adequately. Your legs are ambiguous. And so, at 6am, in the worst possible state for judgement, you reopen a decision that was already made, and you make it differently than you would have the night before. Do this daily and you are not following a plan, you are holding a daily referendum on one. That is the subject of how to stop skipping workouts, and the answer is less about discipline than about when the decision gets made.

The session that costs you the night. You train at 8pm because 8pm is what the calendar gave you, and then you lie there at 11 with your heart still going. This one has a surprising evidence base, including the finding that evening training mostly does not harm sleep at all, and a clear account of the specific conditions under which it does. See why you cannot sleep after an evening workout.

The week at work that never appears on your training log. Your body does not keep separate ledgers. A week of difficult work suppresses the same recovery signals a week of hard training does, and in one study it measurably slowed how fast strength came back after a session, while the participants reported feeling no different. That is how work stress affects your HRV, and it is the single most common reason a plan that looked reasonable on Sunday is wrong by Wednesday.

The weight of having stopped before. The fifth restart is heavier than the first, and not because you are weaker. Because you now have evidence. What to do with that evidence, and why the comparison you are making is statistically unfair to you, is in finding the motivation to start working out again.

Two more sit slightly apart, because they are conditions rather than daily frictions: rebuilding confidence returning to exercise after injury, and the question of whether mood affects workout performance, which turns out to be one of the more sensitive signals available and one of the least used.

None of these is a character flaw. All five are predictable, most are measurable, and every one of them is a thing a system can be built to absorb. Underneath all of them sits the slower question of how any of this becomes routine rather than effortful, which is how to build an exercise habit.

What has actually been tested, and what held

If motivation does not survive its own withdrawal, what does? Three things have real evidence behind them, and the third is the one almost nobody does.

Planning what you will do is not enough. Planning for what goes wrong is.

There is a clean three-arm trial on this. Participants were split into standard care, a group that made action plans (when, where and how they would exercise), and a group that made action plans plus coping plans (what specifically they would do when the obstacle they expected actually arrived). Two months later, weekly exercise came out at 95 minutes for the control group, 113 for the action-planning group, and 179 for the group that had also planned for the obstacles.

The detail that matters most is buried in the middle number. Action planning on its own was not significantly better than doing nothing.

Worth saying plainly, because this is the version of the advice everyone has heard: writing "Tuesday 7am, gym, legs" in your calendar is, on this evidence, close to inert on its own. Writing "and if Tuesday's meeting overruns, it becomes twenty-five minutes on Wednesday morning" is where the effect lives.

Two caveats to keep us honest. This trial ran in cardiac rehabilitation patients with a mean age of 59, so the size of the effect will not transfer directly to a thirty-five-year-old founder. And the broader literature on if-then planning for exercise is more modest than its reputation: across randomised trials specifically, the pooled effect is small and only reliably significant when the plan is reinforced rather than made once and filed. Use the shape of the finding, not the magnitude.

Self-monitoring is the highest-value single component

A meta-regression covering 101 studies and 44,747 participants examined which behaviour-change techniques actually accounted for the differences between interventions. Prompting self-monitoring of behaviour explained more between-study variation than any other single technique, and interventions that combined self-monitoring with other self-regulatory techniques achieved roughly 0.42 against 0.26 for those that did not.

Which is to say: knowing what you actually did, as opposed to what you meant to do, is not admin. It is the intervention.

Feedback beats persuasion, and this one is uncomfortable

A meta-analysis of 27 physical-activity interventions specifically designed to raise self-efficacy found the overall effect small. More usefully, it looked at which techniques were associated with higher self-efficacy and which with lower. Feedback on past performance came out highest. Persuasion and barrier-identification were associated with lower self-efficacy.

The moderator analysis is observational rather than randomised, so treat it as a strong signal rather than proof. But the direction is consistent with everything above, and it is a hard finding for the entire genre of encouragement. Being told you can do it does not appear to help. Being shown what you have already done appears to.

Sizing: the honest version of "start smaller"

Everyone says start smaller. It is worth being precise about what is actually known here, because the confident version of this advice runs ahead of the evidence.

What is well supported: dropout clusters during ramp-up, when the prescription is furthest ahead of the body and the schedule. What has not been cleanly tested, as far as I can find, is a direct comparison of starting doses against long-run adherence. Nobody has run the trial. Anyone telling you the research proves you should start at 50% is telling you something the research has not said.

So here is the defensible version.

The number that should govern your week is not the number you can hit on a good week. It is the number you can hit on a bad one. Consistency is not an average across the month, it is a floor under it, and the floor is set by your worst week rather than your best. A programme built for the good weeks fails every bad week and then asks you to feel bad about it, which is a design flaw dressed up as a character test.

In PX this is a capacity dial, running from one to ten, and it is driven by adherence rather than ambition. Complete 80% or more of what was planned over a fortnight and it steps up. Fall below 50% over a week and it steps down. It recalculates every Sunday. That is the whole mechanism, and it is deliberately unflattering: it sets the level from what you did, not from what you intended, which is exactly the input the evidence above says matters.

Stepping down is the part people resist and the part that does the work. A dial that only goes up is a wish. A dial that goes down honestly is the thing that makes the next four weeks survivable, and the first four weeks are where this is decided.

What a plan that knows all this looks like

Put the findings side by side and a specification falls out of them.

The decision has to be made before the morning, because the morning is when you are worst at it and most likely to renegotiate. It has to be sized from what you actually did, because adherence is the input that predicts the next fortnight and intention is the input that does not. It has to have already planned for the collapse, because coping planning is what separated 179 minutes from 113. It has to give you feedback rather than encouragement, because that is the technique associated with rising self-efficacy rather than falling. And it has to absorb the bad week rather than punish it, because bad weeks are not exceptions, they are the sample.

That is the whole product argument, and it is why PX decides your day before the day happens rather than scoring it afterwards. Your wearable, your calendar and your own reports go in. One plan comes out, across training, mind and food, sized to the capacity you currently have, with the reason attached to every call it makes.

The reason is the part that matters most, and it is the part nothing else does. Not a number. A sentence. Three hard days and a short night have outrun your recovery, so today is easy and Thursday takes the intervals. You can disagree with that. You cannot disagree with a recovery score of 62, you can only feel something about it, and feeling something about a number is precisely the loop that has not worked for you so far.

One more thing, and it follows from the megastudy. A system that works by pushing you is a four-week intervention with better branding. The point of deciding in advance is not to summon willpower on demand. It is to need less of it, permanently, by moving the load from the person to the plan.

None of this makes training easy. It makes it decided, which is a different and more useful thing. The pursuit is daily, and quiet, and mostly consists of doing a reasonable amount of work on a day you did not feel like it, having already agreed what reasonable meant while you were thinking clearly.

Start your plan. PX reads what you already track and builds one plan across training, mind and food, before your day starts. Try PX.