Feedback is a response to some kind of action or activity. This response helps a system or actor to correct its mistakes when taking further actions, as well as to reinforce the steps that lead to its successes.
Feedback comes in many types depending on the context. For example, input feedback confirms whether or not the input we provided was valid. When we press a computer key and a single letter appears, that is input feedback, and our brain takes it as confirmation that we pressed the key hard enough and fast enough. The next key we press is thus pressed with a similar level of force and speed. If the next letter shows up, this is further feedback reinforcing that input action for the subsequent keypresses.
In this way, frequent small actions are constantly reinforcing themselves, enabling the training of some skills through practice alone.
But the more time there is between taking an action and receiving feedback for it, the less the brain associates that feedback with the original action, and the harder it is to apply the correction or reinforcement to the relevant behavior.
For example, if you make a work of art and then get negative feedback on it many months later, you may take it as a criticism of your current skill (even if you have since improved) or your identity / personhood. Whereas getting that same feedback from a teacher in the middle of the piece is more likely to be taken as criticism of a particular technical procedure, allowing it to be corrected in practice.
This can be partly mitigated by having a layer of abstraction (https://lib.hol.ink/index.php?topic=28.0) inbetween, such as a written process you can assess and apply the corrections to for longer activities. This can then be tracked and dated, or accompanied by a process log, so you can check exactly what procedure was actually used when a praised or criticized work was made and how it compares to your current process.
By breaking complex skills down into smaller activities that are easier and faster to correct, we can also gain the ability to progress in that area through practice alone again, because the feedback is close to immediate. This is one of the big benefits of training with exercises (https://lib.hol.ink/index.php?topic=20.0) rather than just finished pieces.
Scaffolding up the complexity of our process and doing speed training of skills we've already learned can also help to make feedback faster when we are doing finished pieces. Because it becomes painfully clear what we're struggling with compared to what we've already mastered. Jumping the gun and trying dozens or hundreds of new skills all at once doesn't afford that kind of clarity.