Two lifters bench 100 kg for 5 reps. One of them had three more reps in the tank. The other had none. Same bar, same reps, same line in the log, completely different meaning. The number that tells them apart is RPE.

What RPE actually is

RPE (rate of perceived exertion) is a 1-10 scale for how hard a set felt, and in lifting it usually maps to reps in reserve (RIR):

RPE 10: no reps left. The last rep was a genuine grind or a miss.
RPE 9: one rep left. You could have done one more, not two.
RPE 8: two reps left.
RPE 7: three reps left. Comfortable and controlled, a rep or two below your ceiling.
RPE 6 and below: four or more reps left. Warmup or recovery territory.

It is a subjective scale, which is exactly the point. Weight and reps tell you what you did. RPE tells you how much you had left when you finished.

Why the same weight means different things

Weight alone is an incomplete data point. 100 kg for 5 at RPE 7 means you finished with headroom, your body handled that load easily, and next session is a reasonable candidate for more weight or another rep. 100 kg for 5 at RPE 9.5 means you were close to your ceiling, the set cost you something, and repeating it, let alone adding to it, is a real ask.

Progressive overload has several dials: load, reps, sets, and proximity to failure. RPE is how you read that last dial. Without it, "100 kg for 5" looks identical on both days. With it, one of those days is progress and the other is barely holding on.

How to calibrate it honestly

RPE is a skill, not an instinct, and most people are wrong about their own number for the first few months of tracking it.

Know what 10 actually feels like. Occasionally take a set to a genuine last-rep grind on a safe, spotted or machine-based exercise so you have a real anchor. Without that anchor, most "RPE 9" sets are actually 7s.
Log it right after the set, not at the end of the workout. Memory flattens effort. By the time you are writing up set four from an hour ago, sets one and four start to look the same.
Do not let a hard session bias a fresh set. The first working set of the day and the last one carry different fatigue, and RPE should reflect the set in front of you, not the session as a whole.

The mistakes that make RPE useless

Logging everything at RPE 9 or 10. If every set is "basically maxed out," the number stops carrying information. You cannot tell a genuine grind from a normal working set, and you cannot tell whether next week is harder or easier than this one.
Chasing a number instead of a process. Deciding the weight has to go up this session, then reporting whatever RPE makes that decision look justified, defeats the purpose. The set happened at whatever effort it happened at, regardless of what the plan wanted.
Redefining the scale week to week. If RPE 8 means something different on a good week than on a bad one, the trend across weeks is noise, not signal.

What logging effort actually enables

Weight and reps on their own can only tell you what happened. Add RPE, honestly logged, and you can start telling grinding apart from progressing.

Zirv logs RPE per set, right alongside the weight and reps for that same set. It is optional, so nothing forces a number you have not calibrated yet, but every set that includes one becomes part of the pattern the app can read later.

That pattern matters most when it moves in the wrong direction. As covered in how to plan a deload week, one of the signals behind Zirv's overtraining detection is effort creeping up at a weight that has not changed: specifically, RPE rising by a full point or more while the weight you are using on that exercise stays within about 5% across your recent logged sessions. On its own that is one signal among several, and it only gets checked once an exercise has enough logged history behind it. But it is a pattern a spreadsheet or your memory will not catch: the bar looks the same, the reps look the same, and only the effort column shows the cost creeping up.

The takeaway

Weight is what the bar did. RPE is what it cost you. Log both, honestly, every set, including the ones that felt worse than the plan said they should. A few weeks of that and the number that used to look identical on two very different days stops being a guess and starts being data you can actually use.