Two flat weeks is rounding error. A real plateau is six to eight weeks of flat e1RM at the same RPE, and you can only call it if your log captured the data.
You moved the same weight twice. That isn't a stall.
Eight days ago your bench top set gave you 102.5 for 3 at RPE 8. Today, same block, same cue to call it at an 8, the bar gave you the same 102.5 for 3. Two sessions, no movement, and the tab already open is the deload calculator, or worse, a fresh program template. Before you re-zero a block that might be working fine, look at how short that window actually is. Two flat weeks on a primary lift is not a plateau. It is the resolution limit of the measurement. The bar loads in increments your day-to-day strength variation swamps, and reading two data points as a trend is the most common way a serious lifter kills a program that was still delivering.
The noise floor
Here is the handle: the noise floor. Below it, nothing your log shows you is signal. The actionable threshold for a real plateau is six to eight weeks of flat-or-falling estimated 1RM, measured at consistent RPE, on the same primary lift variant, with weekly volume held roughly steady. Anything shorter sits inside the band where normal biological variance lives: a short night, a stressful week, a slightly heavier conditioning day, the half-point of RPE drift that creeps in when you are tired. The number on the bar can stall for two or even three weeks while the underlying trend is still climbing. The threshold is not arbitrary toughness. It is the point where the flat line is finally long enough to outrun the noise underneath it.
Both panic and denial come from the same hole in the log
The dominant framing is that any week without a PR is a warning light. It produces two opposite failures, and the same lifter usually commits both in the same year. One is intervening too early: the program asks you to sit at a working weight for three or four weeks before it pushes, you read the holding pattern as a stall, and you bail before the planned overload ever arrives. The other is intervening never: the bar has genuinely not moved in two months, but because no single week felt dramatic, you keep grinding the same load at a creeping RPE and call it consistency. Treat them as a discipline problem and you will keep missing. They are a measurement problem. Both happen because the log did not hold enough clean data to tell signal from noise, so the lifter substitutes a feeling for a reading.
Your e1RM is a calculation, and calculations have error bars
Estimated 1RM is not something you measured, it is something a formula returned. Standard rep-max estimators (Epley, Brzycki, and their cousins) take a weight, a rep count, and increasingly an RPE, and project a one-rep max from them. That projection inherits every bit of slop in its inputs. A single grindy rep you logged at RPE 8 that was really a 9. A bar that felt fast because you were caffeinated and slept nine hours. The result is that your e1RM on any given day carries a band of a few percent around the true value, and strength itself oscillates week to week even on a block that is working. Stack a noisy estimate on top of a naturally wavy signal and a two-week dip tells you almost nothing. You need a window long enough that the trend line, not the wobble, is what you are reading. Six to eight weeks is that window for most intermediate and advanced lifters on a single primary variant.
Walking one block, top set by top set
Take a 5/3/1-style bench progression, last set logged as an AMRAP, e1RM computed each week. Week 1: 104. Week 2: 101. Week 3: 105. Week 4: 103. To the panic reflex, week 2 was a stall and week 4 confirmed it. Read across the four points and the trend is flat-to-up inside a 4-point band. That is noise, and switching here throws away a working block. Now a different read: weeks 1 through 8 come in at 105, 104, 105, 103, 104, 102, 103, 102, every top set called honestly at RPE 8, volume unchanged. No single week screams. But the line bends down and never recovers across eight sessions, with RPE held constant the whole time. That is a plateau, and it is only visible because RPE was captured every session. Strip the RPE out and the second log is indistinguishable from the first: just weights that bounce around. The decision lives entirely in whether the log recorded the inputs.
The question is really about what your log holds
If the threshold is six to eight weeks, then 'is my plateau real' is not a question about your body, it is a question about your records. You can only run the comparison if the log captured the inputs that make it valid: RPE on every working set, the exact variant (low-bar versus high-bar, paused versus touch-and-go), the set and rep scheme, and enough of a volume note to confirm you held it steady. A log of bare weight and reps, no RPE, cannot distinguish a true stall from three weeks of poor sleep, because it never recorded the effort that separates them. The lifters who make this call cleanly are not tougher or more patient. They simply kept a record dense enough to be read six weeks later, and they read the whole window instead of the last two points.
An instrument that holds the window for you
This is where a logging tool stops being a notepad and starts being an instrument. A log that records RPE alongside weight and reps, computes e1RM per session, and holds the trend on a single variant turns the six-to-eight-week window into a readable line instead of a thing you reconstruct from memory at the worst possible moment. You stop eyeballing the last two sessions and start seeing the slope. The math runs while you warm up; it describes what the bar has been doing and gets out of the way. It does not tell you to deload or to push. It hands back the line your own training already drew and lets you, the lifter who knows the block, make the call. Recognition, not motivation. Math, not a chatbot.
What we're tracking next
The open question is how short the valid window gets as a lifter advances. A novice on linear progression can read a stall in two or three weeks because the expected weekly gain is large relative to the noise. The deeper into a training career you go, the smaller real progress gets per block, and the longer the window has to stretch before flat genuinely means flat. The same noise floor, scaled to how much signal you should expect, is the next thing worth pinning down for the advanced cohort.
Keep a log that records RPE, computes e1RM per session, and holds the trend on one variant, so a plateau is a line you can read, not a two-week guess.
This is the read serious lifters do by hand and get wrong under pressure: a flat fortnight feels like a verdict, so the program dies before its overload ever lands. The fix is not discipline, it is a record dense enough to read six weeks deep at matched RPE on one variant. That record is the whole job of the log.
