RPE is easy to prescribe and hard to collect. In a gym you write "5 at RPE 8", watch the set, and correct the client on the spot when the bar is still flying at rep five. Online you get a number in a log from someone who has never trained to failure in their life, and you have to decide whether to program off it.
Everything difficult about RPE with online clients is calibration you cannot supervise. This covers how to teach the scale over the first month or two, how to check the ratings you get back against the loads the client actually lifted, and what to do with the clients whose numbers never come right. For the scale itself, what RPE is and how it works has it.
Why RPE breaks down at a distance
An RPE rating is a comparison. The client is asking themselves how close that set was to one they could not have finished, and the answer is only as good as their idea of what that feels like. Most clients have trained hard, have felt terrible during a set, and have never once found the rep that would not move.
So the number they send you rates discomfort, not proximity to failure. Discomfort peaks early and then plateaus, which is why a new client rates a set of ten at 8 and then rates the same ten with five more kilos at 8 as well. Both felt awful. Neither was close to failing.
In person this self-corrects within a couple of sessions because you are there to say "that was a 6, add weight". Remotely there is nothing to correct against, so the program adjusts itself around a number that is a point or two low, and four weeks later the client is training at RPE 6 while you both believe it is an 8.
Plan for a teaching period. Building a program that depends on a new client's ratings in week one is a design error rather than a client problem.
The four-to-six week teaching sequence
Teach the scale in stages, and do not ask for a rating you have not given the client a way to produce.
| Weeks | What you prescribe | What you ask the client for | What you are doing |
|---|---|---|---|
| 1–2 | Fixed loads and reps | Reps completed, plus "could you have done more?" | Establishing the log habit and a baseline of honest reps |
| 2–3 | One set to genuine failure on a machine | Video of that set and the rep count | Giving them a real 10 to compare against |
| 3–4 | RPE ranges (7–8, 8–9) | A rating per working set | Letting them be half a point wrong without being wrong |
| 5–6 | Single RPE values on main lifts | A rating per working set | Normal programming, with ratings you can use |
Two things make this work. The failure set comes early, because every rating before it is a guess against a reference the client has never experienced. And ranges come before single values, because a client asked for "RPE 8" who is only capable of "7 or 8" will pick a number and defend it rather than admit they are unsure.
Six weeks is the outside. Clients who have used RIR or AMRAP sets before often need nothing but the failure set and a fortnight of ranges.
The supervised set to true failure
This is the highest-value thing you will do for RPE accuracy, and most coaches skip it because arranging it is awkward.
Pick a self-limiting movement: leg press, chest press, machine row, seated shoulder press, a dumbbell press where the dumbbells can be dropped. Never a barbell squat or bench, never anything a failed rep pins the client under. Prescribe a load they can do for twelve to fifteen reps, tell them the goal is the rep that does not move, and ask them to film it end to end.
Then watch it. You are checking that they actually reached failure — the last complete rep visibly slow, a failed attempt after it — and where their own estimate landed. Ask what they would have rated the set two reps earlier. Most clients say 9 or 10 for a set they can then see was an 8.
That video becomes the reference for the rest of the relationship. When a rating looks wrong in week nine, you are not arguing about a definition, you are pointing at the day they found out what a 10 was. The practical side of getting footage you can read is in running remote form checks.
Repeat it about once a quarter, or whenever a client's strength has moved a long way.
Prescribe ranges before you prescribe numbers
A single RPE value is a precision claim. "RPE 8" tells the client you want two reps left, and a client genuinely at 7.5 or 8.5 can only round, which throws away the information you wanted.
Ranges fix that. "Take these to 7–8" is a question a new client can answer honestly. Write them wide early — 6–8 for accessory work, 7–8 for main lifts — and narrow them as the ratings start agreeing with the loads.
Ask in reps, not ratings. "How many more could you have done?" is a factual question about a set that just happened. "Rate that out of ten" asks for a judgment, and people answer it with how they feel about themselves that day. Collect reps in reserve and convert it yourself if your programming is written in RPE — RPE vs RIR covers when to prescribe each.
Tell clients what the rating is for, too. A client who thinks it is a report card rates high to show effort; one who thinks it measures strength rates low. The number decides next week's weight, so a wrong one costs them progress rather than face.
Audit ratings against the loads
A client's RPE and their logged load are two views of the same set, and when the two disagree, the ratings are the part that is wrong. Take one exercise over three or four weeks and ask whether the ratings moved when the loads did. Here is a client's dumbbell bench press, three sets a session, prescribed at RPE 8:
| Week | Load | Set 1 | Set 2 | Set 3 |
|---|---|---|---|---|
| 1 | 22.5 kg | 10 reps @ 8 | 10 reps @ 8 | 9 reps @ 8 |
| 2 | 25 kg | 10 reps @ 8 | 10 reps @ 8 | 10 reps @ 8 |
| 3 | 27.5 kg | 10 reps @ 8 | 10 reps @ 8 | 10 reps @ 8 |
Two things are wrong here, and neither is visible in any single session.
The load went up five kilos and the reps went up with it while the rating never moved. A client who adds five kilos and holds their reps should report a harder set, or you should see the reps fall instead. Second, all three sets within a session are identical. Fatigue accumulates across sets, so the third set at a fixed load is almost always harder than the first. A client reporting 8, 8, 8 is rating the exercise, not the sets.
This client rates everything 8 because 8 is what you asked for. The fix is not a lecture about the scale. Give them a rep target for a fortnight — "as many as you can with good technique, stop when the bar slows" — and watch what the honest rep counts do. If week one's "RPE 8" set turns out to be fourteen reps, you both now know what the ratings were worth.
None of this audit exists unless the rating is stored against the set. RPE and RIR are per-set tracking columns in Fitsly's training tools, logged in the app next to the weight and reps and kept in the client's per-exercise history, which makes a three-week comparison a thirty-second job rather than a scroll back through chat.
When the rating and the video disagree
You will get sets rated 9 where the last rep is fast and the client racks the weight and reaches for their phone. Treat the video as the fact and the rating as the claim, but do not open with that.
Ask first, because there are honest explanations. Technique broke down before the muscle did and the client rated the set where they lost position. Something hurt. They stopped because it was the last set and they were done. Any of those makes a 9 defensible even though the bar looked quick, and finding out which it was tells you more than the rating did.
If none of them apply, correct it against the reference you built. "Compare this to the leg press set in week two — how many more do you think were there?" That is a question about a specific memory, and far easier to answer accurately than an abstract one. Then represcribe: same exercise next week, more load, and a rep-in-reserve count rather than a rating.
Do that two or three times and most clients recalibrate. What does not work is silently adjusting the program around ratings you have privately decided are wrong — the client keeps sending the same numbers and neither of you fixes the thing underneath.
The inflater and the deflater
Two client types produce bad ratings for reasons that have nothing to do with the scale.
The inflater rates high because the rating is how they show you they worked. Every set is a 9, often with a comment about how brutal it was. You back the loads off and they get under-loaded. The tell is ratings that stay high and stable while performance keeps improving.
The deflater rates low because a low number reads as strength. Every set is a 6 or a 7, including the ones where the reps fell off. You add load into fatigue, and it ends in a stalled lift or a tweak. The tell is ratings that stay low while reps or loads go backwards.
Both are fixed by changing what the number means socially rather than by re-explaining it. Ask for a rep count instead of a rating, which is much harder to editorialise, and praise the accuracy rather than the effort — "that estimate was spot on" — so the client learns that a good rating is a correct one. With the inflater, be direct about the cost: a set rated 9 that was really a 7 buys them a lighter week they did not need.
Separate from both: ratings that genuinely climb at a fixed load, week after week. That is not mis-calibration, it is accumulated fatigue, and the deload is due earlier than the program says — programming deload weeks covers what to do about it.
When to stop using RPE with a client
Some clients never calibrate, and programming off their ratings is worse than not asking for them.
Give up when eight to ten weeks have passed and the ratings still do not move with the loads; when the client is a beginner whose technique changes enough between sessions that comparable sets do not exist; or when the ratings have become a source of anxiety, which happens more often than coaches expect with clients who read the number as a grade.
Then program the old way: fixed loads with a defined progression and rep targets, and let the reps carry the information the rating was meant to — a client given three sets of eight who logs 8, 8, 6 has told you plenty. You set the load and the sets and you manage the total, which is what how much training volume clients need is about.
Keep one softer signal: a whole-session effort rating is much easier to give than a per-set one and still tells you whether the week landed. Then try RPE again in six months, after another failure set. A client who could not use the scale at twelve weeks of training often can at nine months.
