Same plan, same lifts, same ten weeks. Only the delivery changed, and one version lost almost half its lifters before the finish line.
Every lifter in this trial had at least a year of training and knew every exercise in the plan. Handed that plan as a PDF, 22 of 47 still fell short of 26 out of 30 sessions.
In a trial published in 2025, researchers led by Simon Gavanda and Steffen Held gave 112 lifters the same 10-week plan and changed only how it was delivered. One group trained with a certified coach in the room at every session. One followed it through an online coaching app. One got it as a PDF.
Ten weeks later, they counted who was still doing it. Per 10 lifters who started, about 9 made it with the coach, about 8 with the app, and about 5 with the PDF.
Those are plain headcounts from one trial, and the paper runs no statistical test on them. They are still hard to unsee. If you have a downloaded programme somewhere that you never finished, one of those three rows is going to look familiar.
The easy explanation is that the PDF group were novices who didn't know what they were doing. That explanation doesn't survive the details.
Same plan, three deliveries: in this trial the formats differed most in who made it to week 10, and much less in what the people who got there went on to gain.
- What the coaching app added over the PDF was finishers, not bigger gains: among lifters who reached week 10, the two groups' average squat and bench gains were almost identical.
- Illness was the exit reason for 12 of 47 PDF lifters against 1 of 17 coached ones. The researchers wonder whether "sick" was sometimes a polite way to quit, while saying it may partly be a matter of chance.
- Other trials found smaller gaps between supervised and unsupervised training, each counting something different, and this trial ran no statistical test on its own finishing shares.
- Only the coached group's fat-free mass rose (everything in the body that isn't fat, on an electrical body test). A 2022 review pooling body-makeup results from 6 trials found supervision made almost no measurable difference.
Online vs In-Person vs PDF: Who Reached Week 10
The rule for finishing was strict: at least 26 of the 30 sessions, so no more than four missed sessions in ten weeks.
With the coach, 15 of 17 made it; with the app, 39 of 48; with the PDF, 25 of 47. Everyone below the bar was counted as a dropout, and only the people above it made it into the results.
Now the part that stings. These were not beginners. Every lifter, men and women aged 18 to 58 with no health problems, had at least a year of lifting behind them and already knew every exercise in the plan.
And the PDF wasn't a vague list. It spelled out the exercises, sets, reps, tempo and rest, plus the rule for when to add weight.
The group that had the whole plan in writing is the group that lost almost half: 22 of 47.
If you're newer to lifting, the flip side matters too. This trial can't say what happens to beginners, because it didn't include any.
So if all three groups knew the lifts and held the same plan, what was different between them?
Coach, Then PDF, Then App: A Path the Industry Already Took
None of the three options at the desk is new. One of the best-known home-workout programmes went through all three, in this same order, as it grew.
Kayla Itsines trained people in person from 2009. By 2014 she couldn't keep up with demand, so a downloadable 12-week Bikini Body Training Guide followed: a PDF that, according to Forbes Australia, was sent around more than purchased. In 2015 came the app, Sweat with Kayla, a different app from the one tested here.[1]
The moves were made to reach more people. This trial asks a different question: run one plan through all three formats, and how many people finish each? The answer lines up with one small-sounding difference.
Who noticed when you skipped a session.
With the coach, a certified trainer was there at every session, one trainer to between one and four lifters. The trainer tracked attendance and corrected technique on the spot. With the app, TrueCoach sent reminders, tracked progress, showed technique videos and gave technique feedback through video, but later, not in the moment.
With the PDF, nobody checked anything until the training logs came back after ten weeks.
Put those next to the headcount and they fall in the same order. The researchers read it as accountability, the sense that someone will notice, but that is their explanation, not something the trial tested.
The app kept almost as many people as the coach. So did the people who finished with it get almost the same results?
App vs PDF: Almost Identical Once They Finished
No: once they finished, they looked almost exactly like the PDF group.
Strength was measured as each lifter's one-rep max, the heaviest weight they could lift once. Among finishers, the app group's squats rose 19.2 kg and the PDF group's 19.4 kg, while their bench presses rose 8.2 kg and 7.7 kg.
The app's edge over the PDF wasn't bigger gains. It was more people reaching the end.
One honest footnote: "almost identical" here is a plain comparison of averages. It covers only people who made 26 of 30 sessions.
Which leaves the question anyone weighing a trainer really wants answered. If you'd finish anyway, what does the person in the room add?
Squat vs Bench: The One Lift Where the Coach Showed Up
In this trial, the answer is one lift.
Among finishers, the coached lifters added 26.6 kg to their squat, against 19.2 kg with the app and 19.4 kg with the PDF: about 7 kg more. The gap was statistically significant against both, meaning it's unlikely to be a fluke.
On the bench press, the coached group's gain sat right alongside the other two, and the statistics couldn't tell the three groups apart.
The researchers think the lifts explain it. A squat needs many joints and muscles to work together in precise order, so feedback in the moment pays off. Pressing a bar off your chest is simpler, and experienced lifters seemed to manage it on their own. The coach gave that feedback on the spot, and the app's came later by video.
Before crowning the coach, meet the group. It was small, with 15 lifters analysed, and younger, at about 26 years old on average against about 32 in the other two. It also had fewer years of lifting and the lightest starting squats, and the analysis adjusted only for where each lifter's squat started.
Two other trials lean the same way on the squat. Only one agrees about the bench. In a 2023 trial in trained lifters, supervision tipped squat gains its way while bench gains came out similar.[2] In a trial from 2000, the trainer's lifters gained more on both lifts.[3]
That covers the lifters who finished. The more telling story may be the 22 PDF lifters who didn't, because each of them left a reason.
The Exit Notes: Who Left, and the Reasons They Gave
Nobody simply vanished. Every dropout left a reason, and the researchers listed them by group.
In the PDF group, 12 of 47 left citing illness, about 1 in 4; in the coached group, 1 of 17 did.
Personal reasons, the lack-of-time and family-commitments kind, took 6 lifters from the PDF group and 2 from the app group. They took none from the coached group.
You can probably hear the text message already. So could the researchers: they wonder aloud whether "sick" was sometimes the polite way out, especially where nobody was checking.
Then, in the same breath, they say the pattern may partly be a matter of chance. The numbers are tiny, the groups were different sizes, and the paper reports no statistical test on any of it. Their suggestion for next time is anonymous exit surveys.
It's a funny detail built on very small numbers. Which is exactly why the next question matters: how far can anyone trust this headcount?
Among lifters who reached week 10, the app group's and the PDF group's average gains on the squat and the bench were almost exactly the same. The app's advantage was getting more of them there.
Why Other Trials Found Smaller Gaps
Here's the honest version: the headcount belongs to one trial, and it isn't a law.
A 2022 review gathered 12 trials comparing supervised and unsupervised lifting, mostly in people who were new to it. It found people kept to their planned training about as often either way: 91.5% with supervision, 87.1% without.[4]
Across trials in adults aged 60 and over, people turned up for 81% of their planned sessions on average, with supervision or without.[5] Those authors warn that most unsupervised attendance came from self-kept diaries, and that 21 of their 34 studies gave the "unsupervised" group some supervision anyway.[5]
None of these sit on one scoreboard with this trial, which counted something else again: people who cleared 26 of 30.
So why was this gap so big? One difference stands out, as a possibility nobody has tested: this PDF group had nobody checking in until the very end. In the review's trials, several "unsupervised" groups had instruction, check-ins or gym staff around, and in the 2000 trial the trainer was even in the room.[4][3]
Two more limits are worth knowing. Only people who cleared the 26-session bar were counted in the results, and the authors warn that this may tilt the results toward more motivated lifters.
And again, every lifter here had at least a year of training. The authors cite other work suggesting in-person feedback may matter especially for beginners. That's their reading of other studies, not something this trial found.
The trial also measured bodies, which raises the question a scale-watcher came here with: did the coach change anyone's?
Did the Coach Change Their Bodies?
On one measure, only the coached group moved.
Their fat-free mass, meaning everything in the body that isn't fat, rose by about 1.4 kg, and their scale weight by 1.8 kg, which the authors say was likely because of that fat-free gain. The app and PDF groups showed no measurable change in fat-free mass.
Fat-free mass isn't the same thing as muscle, though. This 1.4 kg came from a test that sends a weak electrical current through the body, not from a lab scan, and the authors call it less precise. They also ask for caution on the muscle numbers it produced.
And this wasn't a weight-loss study: everyone was asked to keep eating as usual.
A 2022 review pooled the body-makeup results from 6 trials. It found supervision made almost no measurable difference.[4]
One trial's electrical test against a pooled review: they don't agree. Whether a coach changes your body, not just your attendance and your squat, is still an open question, and it's bigger than one 10-week trial can settle.
After Week 10: What a Year Looks Like
There's one thing a 10-week headcount can't show: what happens after the trial ends.
One app's data gives a glimpse. Researchers followed 389,481 users of the Fitbod app, counting only people who had logged at least one workout a week through their first four weeks.[6]
A year later, about 1 in 10 of the self-described beginners were still logging weekly workouts. Among those with four or more years of lifting, it was about 1 in 4.[6]
It's observational: the researchers watched what people did rather than testing a change. It counts logged workouts rather than all exercise. One of its authors was employed by the app company, and it's a different app with a different definition of sticking with it. But one pattern stood out: consistency in those first four weeks was the strongest predictor of who kept going.[6]
Ten weeks may be the easy part. Which changes what the question at the sign-up desk really is.
Back at the Sign-Up Desk
Standing at the desk, it feels like a choice between paying more and getting more.
In this trial, the formats differed most in something else: how many people were still doing the plan at week 10. Among the people who got there, the coach's measurable extra was the squat, and the app matched the PDF on gains while nearly matching the coach on finishing.
What the trial can't know is you. Your record with plans, whether someone checking in changes how often you show up, and whether your goal is a lift where a second pair of eyes could matter: those answers are yours.
Will I get to week 10? And is my goal a lift that rewards someone watching my technique? This trial can't answer those two questions for you, but it shows why they're worth asking before you choose. Asked of a meal box or a weight-loss group instead, the same two questions get different answers.
The researchers closed their paper with advice for coaches, not customers. It reads nothing like a sales pitch. They suggest matching the amount of supervision to two things: how complex the lift is, and how much a person needs someone holding them to account.
So supervision, in their eyes, isn't all or nothing. It's something to turn up for the technical lifts and for people who drift without a check-in, and to turn down where neither applies.
Their other idea is a mix: an app for most of the training, with a coach checking in now and then, in person or on a video call. None of the three groups got that mix, so it's their suggestion, not something this trial measured.
That shifts what's worth finding out about any offer on the table: who actually notices a skipped session, and whether anyone watches the lifts that need watching.
What other research found
Before you change anything
About half the lifters were women: 6 of the 15 coached lifters who were analysed, 20 of 39 with the app and 12 of 25 with the PDF. The trial never looked at men and women separately, so it can't say whether the formats worked differently for either.
The app here was one online coaching app with a coach reviewing technique videos later, and the in-person coach had at most four lifters at a time. An app with no coach behind it, or a coach splitting attention across a bigger group, wasn't part of this trial.
Everyone was 18 to 58 and, like every lifter here, had at least a year of training. In a 2024 review of trials in adults aged 60 and over, the one supervision benefit that held up under the authors' re-checks was leg strength, not grip strength. It's a pattern also seen in older adults, not evidence that this trial's squat result applies to them.
The programme ran for 10 weeks, and everyone was asked to keep eating as usual, so the trial says nothing about longer programmes or about training paired with a diet.
Nobody could compare how heavy each group trained. The app couldn't export its training data, the authors explain, so there's no load record to show whether the coach's squat edge came from heavier training, better technique or both.
The groups started at very different sizes: 17 people with a coach, against 48 with the app and 47 with the PDF. The paper doesn't say why, or describe how people were assigned beyond "randomly".
The paper doesn't say whether the strength testers knew who was in which group. The research team also strongly encouraged every lifter out loud during the max attempts. And it never says where the app and PDF groups trained, so whether they had the same gym and equipment as the coached group is unknown.
Those three gaps are FitChef's reading of the methods. The authors themselves flag two points about the tests: most body tests weren't done first thing in the morning on an empty stomach, and nobody measured how repeatable the tests were.
How much weight this deserves depends on which result is in question. The squat edge was tested and passed, but it rests on 15 coached lifters. The finishing shares and the exit reasons were never tested at all, and the body result is one electrical test set against a pooled review that found almost nothing.
On the bench press, the trial found no difference big enough for groups this size to catch. That isn't the same as proof of no difference.
The outside studies that agree aren't fully independent of each other. The 2022 review counts the 2000 and 2014 trials among its 12, its first author co-wrote the 2023 trial, and the 2008 study's first author was on the 2000 team.
And the paper's own summary says more than its results do. It calls supervision better for well-being, but when the three groups' well-being changes were tested against each other, no clear difference showed up. And satisfaction with coaching was only reported for the coach and app groups.
Two of its numbers don't add up either. The finishing percentage it prints for the PDF group doesn't match its own headcount, which is why this page uses the counts. And one row of its muscle table doesn't reconcile, so the authors' caution on muscle numbers stands.
The body question is the one thing this page can't close by itself: the trial and the wider research point different ways, and one 10-week test can't settle which to believe. Settling it means weighing every trial that actually measured bodies, side by side. That is the whole job of the page on whether paying for a trainer changes what happens in the mirror or mostly gets people through the door. And if the choice at the desk is a group programme rather than one-to-one coaching, seven trials set those two side by side.
What This Study Found
All findings from this paper, in plain language.
- Among lifters who made at least 26 of 30 sessions, those coached in person added the most to their squat, clearly more than the app or PDF groups.
- All three groups got stronger on the squat, whether they trained with a coach, an app or a PDF.
- All three groups also got stronger on the bench press, with every format producing a clear gain.
- On the bench press, the three formats ended up about even, with no difference the trial could detect among lifters who finished.
- Only the coached group gained fat-free mass, meaning everything in the body that isn't fat, and it gained more than the other two groups.
- The coached group's scale weight went up, more than the PDF group's, which the authors put down to the fat-free mass it gained.
- The body test showed muscle gains in the coached and PDF groups, but the authors themselves say these muscle numbers should be read with caution.
- Fat mass crept up in the app group, and no two groups differed clearly, in a trial where everyone kept their usual diet.
- Well-being scores rose in the coached and PDF groups but not the app group, though the groups' changes didn't differ clearly from each other.
- Lifters coached in person were more satisfied with their coaching than app users were, and the paper reports no satisfaction result for the PDF group.
- 15 of 17 coached, 39 of 48 app and 25 of 47 PDF lifters reached 26 of the 30 sessions, shares the paper never tested statistically.
- PDF lifters left most often for illness, injury and personal reasons, and the researchers wonder about polite excuses while allowing it may partly be chance.
- Where each lifter started strongly shaped where they finished, on every strength, body and well-being measure the trial ran.