After every workout, your email has a button on it: Rate Workout, thirty seconds. Tap it and you get two things to set. How hard the session felt, on the standard 1 to 10 RPE scale. And how you felt doing it, from Strong through to Weak. The page shows you the range the session was designed for, so you are rating against the plan rather than into the void.
Most athletes assume this ends up in a spreadsheet somewhere and disappears. It does not. That rating is one of the few signals the system cannot derive from your files, and it changes what you are given next. Here is what happens between the rating and the next session.
Why objective numbers are not enough
CTL, ATL, TSB, ACWR: these all derive from TSS, and TSS derives from power or heart rate. They do a good job of describing what you did, but they cannot tell the full story of how it felt.
Two athletes with identical TSS for the week can be in completely different states. One slept eight hours a night, ate well, and cruised through. The other worked 60 hours, fought a cold, and ground every session out. The numbers look the same. The bodies do not.
The gap between how a session should have felt and how it actually felt is where RPE lives. If a Z3 tempo interval keeps feeling like a Z4 threshold effort this week, the thing that changed is usually you.
How the system uses it
Your rating goes back to your Intervals.icu activity, where it sits alongside the file. The post-workout analysis then compares how the session felt against how a session of that type should have felt, and records the verdict: easier than expected, as expected, or harder than expected. That verdict is what the rest of the system reads.
A single hard rating does not do much on its own. Bad days happen, and a coach who panics at one of them is not much of a coach. Patterns do three specific things.
The coach stops giving you that workout type. This is the sharpest of the three and it works per stimulus rather than across the board. If a sweet spot session rated harder than expected in the last seven days, the selector avoids sweet spot. Between seven and fourteen days ago, it keeps the type but suggests a gentler shape, usually shorter intervals. Your threshold work is untouched by a bad sweet spot week, because the problem was specific and the response is too.
Three hard sessions in a week swap the stimulus. Once three of your last seven days rate harder than intended, the rules engine raises a signal and the day's session gets swapped toward endurance. Note that this is a count, not a streak: three scattered hard-feeling days across the week trip it just as surely as three in a row.
High RPE holds you back from a harder phase. Training phases advance on readiness rather than calendar dates, and your average recent RPE is one of the gates. An average at 8 or above while you are carrying fatigue blocks the promotion outright, so an incomplete Build phase gets extended instead of rolling into specialty work. At 7 or above with any negative TSB it logs a caution rather than blocking. The reasoning is out of Foster's 1998 session-RPE work: internal load is real load, and the objective numbers do not always see it.
Whichever fires, it shows up in your coaching note the next morning, so the change arrives with its reason attached. That is the same principle the whole daily adaptation pipeline runs on.
What the ratings actually mean
Two numbers, and they are not measuring the same thing.
RPE, 1 to 10. How hard the work was. This is the standard Borg-derived session scale, and the rating page shows you the band the session was built for before you pick: an endurance ride and a VO2max set carry very different expectations, and rating a VO2max session a 7 means something quite different from rating an endurance ride a 7. What the system acts on is the gap between the two, never the raw number.
Feel, Strong to Weak. How your body was, which is a separate question from how hard the work was. A brilliant day can still hurt. The page suggests a value based on your recovery score, and you are free to disagree with it; the disagreement is itself informative.
You do not need to calibrate yourself against anyone else. Because every rating is judged against what that particular session was designed to feel like, your 7 and somebody else's 7 never need to mean the same thing.
How to rate well
A rating the coach can act on is worth more than a fast one, and rating well is a skill. It takes about a month to settle. Four things make the difference.
Rate the session, not the worst minute of it. Session RPE is a single verdict on the whole thing, which is what makes it comparable week to week. The convention from Foster's work, where this scale comes from, is to give it roughly half an hour after you finish. Rate in the first two minutes and the last interval dominates everything; rate the next morning and you are rating your sleep. Half an hour out, the session has settled into one impression.
Rate the effort, and let the result go. The question is what the work cost you. A session you executed beautifully can still have cost a great deal, and a session you abandoned halfway can have cost very little. Athletes tend to shade their rating toward how pleased they are with themselves, which quietly makes the signal about mood.
Use the labels as anchors. The scale is not arbitrary and the app names every point on it:
| RPE | What it means |
|---|---|
| 1–2 | Very easy to easy. Conversation is completely unaffected. |
| 3 | Moderate. You are working, but you could hold it a long time. |
| 4 | Somewhat hard. Talking is in short sentences now. |
| 5–6 | Hard. Sustainable, but it demands your attention. |
| 7–8 | Very hard. A few words at a time. |
| 9 | Extremely hard. You are counting down. |
| 10 | Maximum. Nothing left. |
Remember what your rating is compared against. Every session type carries a band it was built for, and the rating page shows you that band before you choose:
| Session type | Expected RPE |
|---|---|
| Recovery | 1–3 |
| Aerobic | 2–4 |
| Endurance | 3–5 |
| Tempo | 4–6 |
| Sub-threshold | 5–7 |
| Threshold | 6–8 |
| VO2max | 7–9 |
| Anaerobic and neuromuscular | 8–10 |
This is the part worth internalising, because it changes what a number says. An 8 on a VO2max session tells the coach almost nothing: that session was built to land between 7 and 9, so an 8 is the plan working. An 8 on an endurance ride is one of the loudest things you can say, because that ride was built for a 3 to 5. The absolute number is not the signal. The distance from the band is.
Which leads to the one mistake that costs you the most: rating everything in the middle. If a hard week goes in as a string of comfortable fives because you did finish every session, you have made yourself invisible to the thing whose entire job is to notice. Nobody is marking your effort. An honest 9 on a session that should have been a 6 is not a confession of weakness, it is the most useful two seconds of input you will give the coach all week.
Why this matters for AI coaching
A lot of AI coaching tools treat training as an open-loop problem: here is your plan, execute it, see you next Sunday. The feedback they collect is usually ornamental: nice charts, maybe a weekly summary, none of it wired back into what gets generated next.
The RPE button is one of the places where IntervalCoach closes the loop. Your subjective experience is a signal the system listens to, and listening to it means acting on it. It acts with a specific, bounded change to the workout you are about to do.
Use it
If you are going to build one habit around IntervalCoach, build this one. Half a minute after a session, judged honestly against the band it was designed for, buys you a coach that can tell the difference between a plan that is working and a plan that is grinding you down. There is no other input with that ratio.
If you have turned off post-workout emails and still want to rate, the iOS and Android apps let you set both values on an activity, and anything you enter on the activity in Intervals.icu itself is read straight back. The coach does not mind where the number comes from.