Power meters changed cycling because they made effort measurable. They didn't replace the thing they made measurable. They sit alongside it. The serious cyclist who only watches power is leaving information on the table; the serious cyclist who only watches RPE is leaving precision on the table. The pair is the system.
This piece is the working framework for using power and RPE together — what each one tells you, when each one is the better real-time tool, and how the divergence between the two reveals the things neither one would catch alone.
What each number is actually measuring
Power is mechanical work delivered to the cranks per unit of time, expressed in watts. It is an objective measure of what the rider is doing. A 300W effort is 300W whether the rider feels great or terrible, whether they're caffeinated or under-slept, whether the room is cold or warm.
RPE — rate of perceived exertion — is a subjective measure of how hard the work feels to the rider. The dominant scale is the Borg CR-10 (a 0-10 scale with verbal anchors), used in endurance research for fifty years. Anchored properly, it is reliable across riders and stable within a rider over time.
The two numbers do not measure the same thing. Power measures the work. RPE measures the cost.
The rider whose power is on target but RPE is two points higher than usual is doing the same work at a higher cost. The rider whose RPE is on target but power has dropped 5% is paying the same cost for less work. Both are signals; both matter; neither one is fully visible without the other.
The 10-point RPE scale, anchored for cyclists
The CR-10 scale, lightly adapted to be useful for cyclists in the field:
- 0 — Rest. Lying still. No effort.
- 1-2 — Very easy. Recovery spin. Could hold a phone call.
- 3-4 — Easy / endurance. Zone 2. Full conversation possible. The cardiovascular effort is real but the legs are not under stress.
- 5-6 — Moderate / sweet spot. Conversation in short sentences. The effort is "comfortable hard." The rider could hold this for an hour but would prefer not to.
- 7 — Threshold. Spoken sentences become harder. Breathing is heavy and rhythmic. The effort is sustainable for 20-60 minutes depending on training state.
- 8 — VO2 / hard. Single-word answers. Cannot be sustained beyond 4-8 minutes. The effort feels at the edge of control.
- 9 — Very hard. Cannot speak. Sustainable for 1-2 minutes at most. The body's compensatory mechanisms are fully engaged.
- 10 — Maximal. Sprint. Sustainable for seconds. Nothing left in reserve.
A rider learns this scale in 6-8 weeks of structured training, faster if they have a coach calling out "this should feel like a 7" mid-effort. The numbers stabilise once the rider has felt each anchor enough times.
The four divergences worth catching
The reason to track both power and RPE is not that they always agree. It is that the times they disagree are the times most worth listening to.
Divergence one: power on target, RPE creeping up. The most common and the most diagnostic. The rider's prescribed sweet spot at 250W felt like a 6 four weeks ago. This week it felt like a 7-8. The watts are still there. The cost is rising. This is the early signal of decoupling — the rider is accumulating fatigue faster than they are recovering from it. Run another two weeks at the same load and the power will start dropping; catch it now, deload for a week, and the trajectory often re-flattens before the file admits anything was wrong.
Divergence two: power below target, RPE on target. The rider feels exactly as they should — the threshold session feels like a 7 — but the wattage is 5-8% below target. Several explanations are possible: cumulative fatigue, dehydration, inadequate fuelling, life-stress load, or a power-meter calibration drift. The rider needs to shorten the session to keep the quality, and the post-ride question is what changed in the system. The worst response is to grind out the prescribed wattage at a higher RPE — that's the path to overtraining.
Divergence three: power high, RPE low. Less common, more interesting. The rider hits 290W on a 280W prescribed effort and it felt like a 5-6 instead of a 7. This is the file that says fitness has moved. Ride two more sessions at the new prescribed wattage; if the RPE-power relationship holds at the higher level, your FTP has shifted. Re-test or recalibrate the zones.
Divergence four: race day mismatches. A race or hard group ride where power is high (or strangely low) and RPE is meaningfully different from training. This is usually a pacing issue — the rider went too hard early and is over-extended (high power, climbing RPE), or sat in the group too long and arrived at the finish without having actually raced (low power, low RPE). The post-race review is more useful than any single number.
How to log RPE without making a project of it
The friction is the problem. Riders who try to log RPE on a 1-decimal-place precision scale, mid-ride, give up inside two weeks. Riders who log a single integer post-ride for the session as a whole sustain it for years.
The minimum viable system:
- One RPE number for the session as a whole, logged post-ride.
- For interval sessions, a second number for the average RPE of the work intervals.
- That's it.
TrainingPeaks Premium has the post-session RPE field built in. TrainerRoad, intervals.icu, Final Surge, and most other serious platforms support it as a standard field. The logging takes ten seconds and the data accumulates into a useful trend across weeks.
The riders who do this consistently end up with a TrainingPeaks calendar where every session has both a power-derived TSS and a self-reported RPE. The combination produces a much richer view of training load than either alone — the published research on session-RPE-derived training load (sRPE × duration) tracks closely with power-derived TSS for most rider populations, and where they diverge is where the interesting questions are.
When RPE is the better real-time tool
Three situations.
Group rides and races. Efforts vary too much, recovery between efforts is too short, and the rider's job is to ride the race in front of them, not the file. Power becomes the post-ride analysis tool. RPE becomes the in-the-moment pacing input. The rider who keeps glancing at the head unit during a sprint or a paceline is the rider who's about to get dropped.
Sessions where terrain dominates. A 25-minute climb at threshold, ridden outdoors on a real ascent, is a session where the gradient changes constantly. Trying to hold a fixed wattage on shifting terrain produces a worse file than letting the power vary within an RPE band. The objective is "this should feel like a 7 throughout"; the watts will land where they land.
When the power meter is unavailable. Calibration drift, broken meter, borrowed bike. The session still has to be ridden and RPE is the working substitute. Most riders find this is a useful exercise in itself — three or four months of training with RPE only typically sharpens the rider's internal sense of effort meaningfully.
When power is the better real-time tool
Three situations.
Short, precise intervals. A 4-minute VO2 effort or a 90-second over-threshold push. The duration is too short to settle into a stable RPE; the wattage is the cleaner signal. Hold the watts and let the RPE be whatever it is.
Indoor structured workouts. ERG mode is doing the work; the rider's job is to stay on the wattage. RPE becomes the post-session calibration check rather than the in-ride decision tool.
Pacing a long, sustained effort where you don't know the terrain. A long time trial on unfamiliar roads. The power target is your defence against blowing up in the first ten minutes. RPE is the second-channel verification.
What to avoid
Two failure modes worth naming.
Reading too much into a single bad day. A single session with high RPE and low power can mean almost anything. The signal is in the trend over 14-28 days. Look at the rolling average; ignore the daily noise.
Letting RPE override prescribed effort on every hard day. "I felt tired so I skipped the session" is a problem if it becomes the default. The rider needs to ride into a hard session believing the effort will land, then evaluate honestly afterwards. The line between "I'm appropriately under-recovered, this should be tomorrow" and "I'm avoiding effort because hard work feels hard" is real and depends on track record.
The two-channel system works because the channels mostly agree. The watts you'd expect at the RPE you'd expect, at the volume you'd expect, are the file of a rider on track. The interesting questions live in the divergences. The cyclist who logs both numbers consistently has the data to ask those questions; the cyclist who logs only one is guessing about the other.
The work is on the bike. The interpretation runs on two channels. Use both.