A form fix looks like a plateau unless your log knows the date.

The squat that got lighter on purpose

A lifter films a top set of squats from the side, watches it back between sets, and sees what everyone else at the gym already knew: the hip crease never got below the knee. So they fix it. Next session, same bar, honest depth. The bar feels twice as heavy and the reps die early.

Six weeks later the log tells a clean, wrong story. Top set down. Estimated 1RM down. If the app runs plateau detection, it's been flagging squat for a while now.

And the lifter starts to wonder if the old squat was fine after all.

A form change is a dated intervention

Call it the dated intervention. When you deliberately change how a lift is performed (deeper squat, stricter row, a new bench grip), you've changed the independent variable in your own N=1 experiment. The load that follows belongs to a different lift with the same name.

Your log can't know that unless you tell it. Everything after the change gets read against everything before it, so a drop that's entirely expected gets scored as regression, and a stall flag fires on a lift that's doing exactly what a freshly fixed squat should do.

The remedy is small. One timestamped note, on the day of the change, saying what changed. That note turns one confusing chart into two readable ones.

The load that follows belongs to a different lift with the same name.

Why 'be patient with your form' doesn't help the log

The standard advice after a technique fix is patience. Drop the weight, groove the pattern, the numbers come back. Fine as far as it goes. But it's advice for the lifter's mood, and it says nothing about the record.

Most tracking treats a lift as a constant. Squat is squat across five years of history, so a chart can draw one line and a stall check can compare this month's top set to last month's. That assumption holds for most of a training career. It breaks on exactly the day you get serious about technique.

So the generic advice and the generic log fail together. Patience tells you to wait. The chart keeps showing a decline, and charts are louder than advice. Weeks in, with nothing in the record explaining the drop, plenty of lifters quietly drift back to the old groove because the old groove had better numbers.

Why do numbers drop after a form check?

Range of motion and load capacity trade against each other, and squat depth shows it plainly. Squat deeper and the bar gets lighter, because the same legs have a longer, harder path to cover. Cut the depth and the load can climb.

Lighter bar, better position. That's the trade a depth fix makes on purpose.

The same mechanics show up across the other common fixes. Changing bench grip width changes the range of motion and which muscles carry the bar, so even a move toward a stronger position usually costs load while the pattern settles. A row with the torso held still loses the hip drive that was moving the last few reps. A paused squat loses the bounce. Each one produces an immediate, predictable drop that says nothing about whether you got weaker.

On a single top-set chart the two situations look identical. Session to session, they barely resemble each other.

One squat, two histories

Picture the lifter from the opening, on a standard weekly progression. Before the change, the squat history shows the top set creeping up across a block, then going flat for a few sessions. That flat stretch is exactly what a plateau flag exists to catch.

Then the depth fix. Session one at honest depth: a plate or two lighter, fewer reps. Session two, same load, one more rep. A few sessions later, a small load jump. Read across the whole history, this is a dip below a line that had already gone flat. Read from the change forward, it's a fresh progression climbing from a new floor, and it's climbing, which the old squat had stopped doing.

Same sets. Same dates. Two opposite conclusions, and the only thing that decides which one is correct is whether the log knows where the change happened.

With a note on that session ("depth to below parallel, filmed, confirmed"), the lifter compares post-change sessions to each other and sees progress. Without it, they compare against a squat they no longer do, and they see a problem.

What changes once the intervention is on the record

Three things follow.

First, the fix gets judged on its own window. Compare post-change sessions to each other for at least one full block of your program before asking whether the change worked. The old numbers stay in the record as reference, never as the benchmark.

Second, you choose between a note and a new lift. A small change (grip a finger wider, a slightly longer pause) fits a note on the existing lift. A big one, like moving from a squat cut above parallel to a full-depth squat, often deserves its own exercise name so the charts never mix. Either works as long as the boundary is dated.

Third, a stall after the change means something again. Once the new pattern's rebound runs out and the top set goes flat at the new depth, that's a real plateau, and it calls for the usual programming response. The note doesn't excuse a stall forever. It just stops the change from being mistaken for one.

That third point protects the fix itself. The worst outcome of an unannotated form change is a lifter who reverts to a worse squat because the record argued for it.

Where the note field stops being decorative

Most loggers treat notes as a diary margin: optional, unsearchable, ignored by everything else in the app. For this problem the note is the most load-bearing field in the record, because it's the only place a deliberate intervention can live.

Platepusher keeps notes as part of the training record, so the date you changed your depth sits in the same history the plateau flag reads from. Notes attach at three levels: a whole session, an individual exercise (including a separate setup note), and a single set. They show in the History session detail and when you edit a past session, and a note on a set from a recent session can feed the Today signals. The lift's progression chart and plateau flag don't display notes next to them. The flag still does its job: it describes what the math sees when the top set stops moving. The note tells you which explanation fits. Signal from the instrument, context from the lifter.

That split matters. An app that silently decided your drop was "just a form change" would be guessing. An app that ignored the change would be wrong. A plain record with a dated note lets you make the call, which is where the call belongs.

If you're bringing years of history over from another tracker, the same logic covers old technique changes. Write down the ones you remember. The [argument that your training history is the experiment](/articles/your-history-is-the-experiment) only holds if the experiment's conditions are on the record.