Welcome • User Guide
ToS • Privacy • Canary
Donate • Bugs • License

©2026 Poal.co

667

I posted this question in the math forum on redddit, but it was deleted because it is not a math question.

I am hoping someone here can enlighten me.

In my life as an estimator for a machine tool builder, I have always used a learning curve model to estimate future repeat production.

That model is: Every time a quantity of parts is ran, the time to complete should improve; that improvement will be seen every time the quantity is doubled. In other words - with a 95% learning curve, a 5% improvement (100-95) will be seen between the first and second part, and again from the 2nd to the 4th, from the 4th to the 8th, etc. Every time the quantity doubles, the same percentage improvement is realized. (There is a little more to it, batch sizes and a 'forget factor' and so on, but I think you get the point.)

Every time a job is repeated, it is expected to go faster/cheaper because of experience and lessons learned.

If a job is able to be improved considerably every time it is re-ran, then the cost/time goes down considerably on subsequent runs. That is, a plot of cycle times or costs vs quantity, would show a very big drop from part 1 to part 2, from part 2 to part 4, from 4 to 8, etc. A big drop = a steep curve. Going down that curve happens quickly.

If a job is very difficult to improve upon, then the curve is considerably shallower, because the time to produce does not get better over time = shallow learning curve.

I have always used 'steep learning curve' to denote 'rapid improvement over time' but am now being told that is the opposite; apparently my definition is backward, and a SHALLOW learning curve means rapid improvement over time.

Could someone please set me straight on this?

I posted this question in the math forum on redddit, but it was deleted because it is not a math question. I am hoping someone here can enlighten me. In my life as an estimator for a machine tool builder, I have always used a learning curve model to estimate future repeat production. That model is: Every time a quantity of parts is ran, the time to complete should improve; that improvement will be seen every time the quantity is doubled. In other words - with a 95% learning curve, a 5% improvement (100-95) will be seen between the first and second part, and again from the 2nd to the 4th, from the 4th to the 8th, etc. Every time the quantity doubles, the same percentage improvement is realized. (There is a little more to it, batch sizes and a 'forget factor' and so on, but I think you get the point.) Every time a job is repeated, it is expected to go faster/cheaper because of experience and lessons learned. If a job is able to be improved considerably every time it is re-ran, then the cost/time goes down considerably on subsequent runs. That is, a plot of cycle times or costs vs quantity, would show a very big drop from part 1 to part 2, from part 2 to part 4, from 4 to 8, etc. A big drop = a steep curve. Going down that curve happens quickly. If a job is very difficult to improve upon, then the curve is considerably shallower, because the time to produce does not get better over time = shallow learning curve. I have always used 'steep learning curve' to denote 'rapid improvement over time' but am now being told that is the opposite; apparently my definition is backward, and a SHALLOW learning curve means rapid improvement over time. Could someone please set me straight on this?

(post is archived)

You are currently inside a comment thread.
Click here to see all the comments (26).
[–] • 0 pt

Learning is exponential. Experiments with neural nets show that learning is exponential. Also it depends on how you plot performance over time, as that will shape the curve. Then you have to define the behavior, or behavior group. Learning curves are like climbing a mountain, the steeper it is the harder it is to make progress. Take for example a martial art like Bagua, or something like that. Those arts have very steep learning curves because it takes a lot of effort and time to make progress, vs something like boxing where you can make progress quickly, that means the curve is less steep. A steep curve takes a lot of effort to make progress, but when you do make progress, it's a more dramatic improvement.

Another example, let's say you are practicing a jab vs a side kick. You can repeat the jab much faster and it takes less energy and the difference between the first and 1,000th jab won't be very high. Whereas it takes much more energy and effort to throw a side kick, and the difference between the 1st and 1,000th kick will be much more noticable.

[–] • 1 pt

I agree. In this one horizontal would be progress, and vertical the accumulated effort spent. For a steep task, you put a lot of effort in and see little progress (horizontal axis). This makes people drop out before they reach the top. Once they reach the top, they suddenly can reach the right side. Whereas something more gradual gives continual progress from effort, so it's easier to stay in the game. The curve is thus illustrating psychological difficulty, not reward for effort.

[–] • 1 pt

BOOM goes the dynamite!

[–] • 1 pt

No, while parts of what you're talking about may be true, they have nothing to do with a graph showing learning (skills acquisition) over time. Just go through the task of building a graph that represents how well something has been learned over some period of time. Look at the different curves you've drawn, and try to think about what they'd mean.

[–] • 1 pt

They're all either logarithmic or exponential. Learn to cooperate instead of competing for the glory of being right.