Finding the delayed effect of a change in input

I am trying to figure out the delay between changing the speed of a pump that pumps a modifier into a process and the change in Amps drawn by an extruder at the end of the process. The amps drawn are changing constantly and are effected by other variables but the amps are held within a range by changing the speed of the modifier pump. Because the amps drawn are constantly changing you can't just look at the trend line and see time delay accurately. It is generally accepted that the delay is between 10 and 20 minutes but I want to try to find out more accurately how long it is so that I can see if it is affected by other variables in the process.

I have logs of the percentage speed of the pump and the amps drawn by the extruder, I have extracted the pump % speed and extruder amps every minute for a couple of months and that is the data set I am working with.

Now my idea is that I can find the delay by taking the change in modifier input as t=0 and check every minute if the extruder amps changed in the same or opposite direction of the modifier(out to 20 minutes), if it changed in the same direction I set the value to 1 and if changed in the opposite direction I set the value to -1. My thought is that by iterating through tens of thousands of changes in modifiers like this and then adding up all the results at 1 minute, 2 minutes, 3minutes... I will see a spiking value between 10 and 20 minutes somewhere where the modifier actually had an effect on the extruder amps while the other time periods where it doesn't have an effect it should be random so the -1's and 1's should stay around zero. That is not what I see when I try to trend the time periodes though, it is a noisy mess that doesn't make sense.

My expected result would be something like this

Is there something fundamentally wrong with my thought process? and are there any words or data science methods I can google that could help with problems like these?