How long will the next batch take?
Build times fall as a team learns, but not by a constant amount, so a straight line through them misleads you. A learning curve says time drops by a fixed percentage every time output doubles, and taking logs turns that curve into a straight line you can fit with ordinary regression. The useful part is what the slope becomes: a quotable number of hours.
Analytics Intermediate Regression free
After you install, this is the model to open.
Learning Curve: How Long Will the Next Batch Take?
- In your spreadsheet, click the Sortia icon in the strip of icons down the right-hand edge. No strip? Click the arrow at the bottom-right to open it. You can also use Extensions, then Sortia, then Open Sortia.
- Click Start from a template and put that name in the search box.
- Pick the card with that name and click Load this template. It arrives on a new tab with real numbers already in it.
The answer
The regression gives the slope; the fit block turns it into hours you can put in a quote:
- Learning rate
- 79.6% close to an 80% curve
- Fit quality
- 0.985 R squared
- Next 26 units
- 907 hours, fitted
- Quoting the last unit
- 1,108 hours, 18% high
The slope of -0.328 means two to that power is 0.796, so this is close to a textbook 80% learning curve: every doubling of cumulative output cuts the time per unit by about 20%. An R squared of 0.985 with a standard error of 0.012 says that is a real pattern rather than a line drawn through noise. The fitted time for unit one is 101.8 hours against the 100.1 actually recorded, which is a good sign the curve fits the early units as well as the late ones. Then the part that matters commercially: quoting the next 26 units at the most recent observed time of 42.6 hours gives 1,108 hours, while the fitted curve gives 907. That is 200 hours, about 18%, of difference, and the fitted figure is the one you can defend, because it comes from the whole history rather than from whichever unit happened to be built last.
The model
Fourteen recorded build times, log-transformed in two helper columns, with the regression run on the transformed data. A second block converts the fitted slope into a learning rate, an implied time for unit one, and the total hours for the next twenty-six units.
| Units recorded | 14, from 100.1 hours down to 42.6 |
| Transform | ln(hours) against ln(unit number) |
| Fitted slope | -0.328, standard error 0.012 |
| Fit quality | R squared 0.985 |
| Learning rate | 2 to the power of the slope |
| Forecast | units 15 to 40, summed |
Once it is in your sheet
- The model arrives with real numbers in it and runs as it stands, so you can press the button first and understand it second.
- Change the numbers to yours. The sheet marks which cells are inputs and which hold formulas, and most labels carry a note explaining the row.
- Press the run button at the bottom of the panel. It is labeled for the tool you are in, and the result lands on its own tab, with a written reading of it beside the figures.
Never used Google Sheets? Start here goes the whole way, in seven steps, and assumes nothing.
Next question
- What is your average hiding?One Table That Describes Your Numbers
- You guessed min, likely, max. What does history say?Which Distribution Fits Your Data?
- What season length should your forecast use?Does Your Revenue Have a Season?
- Is that vibration random noise or a real rhythm?Find the Hidden Vibration Frequencies
- Can you get the same random numbers back?Reproducible Random Draws, With a Seed
- Did the new format really rate better?Two Groups, Ranked Feedback, One Answer
Every model like this one, and the method behind them: Statistics in Google Sheets.