Skip to content

Free template

Updated

Learning curve and ramp-up workbook

A learning curve says that each time cumulative output doubles, the time per unit falls to a fixed share of what it was. On a 90% curve, unit 2 takes 90% of the time of unit 1 and unit 4 takes 90% of unit 2. This workbook turns that into planning numbers. Enter the first-unit hours and a learning rate. It gives the time per unit, the running total and the running average for up to 500 units under both textbook models, unit-time (Crawford) and cumulative-average (Wright). It can fit the rate from your own observed hours, with R squared and a warning when there are fewer than 5 points or the fit is poor. Then it answers the planning questions: the hours for the next batch, the unit at which the time reaches a target and a weekly ramp plan, with a plateau at your standard time and a sheet for new hires on different start dates. The curve models repetitive, stable work, and a rate you assume is a guess, so replace it with your own data. The workbook quotes no typical rates.

Free to use: print it, copy it and edit it for your team. Enter your name and work email once to download.

Page 1 of the learning curve reading guideThe definition of a learning curve, the exponent b = ln(rate) / ln(2), the unit-time and cumulative-average formulas, two charts of a 90% curve on linear and log-log axes, and three things to know before using it.
Unit-time and cumulative-average modelsTable to 500 unitsPlateau at standard timeFit from your own dataR squared and warningsHours for units a to bUnit that reaches a targetWeekly ramp planHire cohort sheetWorked example

When to use it

When to use a learning curve workbook

  • When you launch a new product, line or cell and need the hours for the first batches and the unit at which the time per unit reaches standard.
  • When a group of new hires starts on different dates and you want the expected weekly output of the group.
  • When you have the hours for a few early units and want the rate your own data implies, not a rate from a book.
  • When you schedule or quote a repeat job and need the hours for a batch from unit a to unit b.
  • Not for one-off jobs, work set by a machine's cycle, or a design that keeps changing. The limits are listed on page 4 of the PDF and on the Read me sheet.

How to fill it in

  1. 1

    Pick the model that matches what you measured

    If you timed individual units, use the unit-time model. If you keep a running average, use the cumulative-average model. A rate fitted to one model is not the rate for the other.

  2. 2

    Enter the first-unit hours and a rate, or fit one

    On the Curve sheet type the hours for unit 1 and a rate above 50% and below 100%. Or put your observed points on the Fit sheet and switch the source to the fitted rate.

  3. 3

    Set the number of units and a standard time

    N can be up to 500. A standard time makes the curve flatten there, because a unit does not go on getting faster than standard.

  4. 4

    Fit from your own data

    Enter up to 40 points, a unit number and the hours you measured. The Fit sheet gives the implied rate, the fitted first-unit hours and R squared, and warns on fewer than 5 points, a poor fit or hours that are not falling.

  5. 5

    Read the planning answers

    The Plan sheet gives the hours for units a to b and the unit at which the time reaches a target, from a closed form checked against the table, with the total hours to get there.

  6. 6

    Build the weekly ramp

    Enter people and productive hours per person per week. The sheet shows the units built each week and the time per unit at the start of the week. The Cohort sheet does the same for each operator's start date.

  7. 7

    Re-fit as the data comes in

    Replace the assumed rate with your own as soon as you have observed hours, and again as the run goes on.

The two models

Unit-time or cumulative-average?

Both models use b = ln(rate) / ln(2). They differ in what T1 x^b describes. The figures are for a first unit of 10 hours and a 90% rate, worked from the formulas.

  • T1 x^b is

    Unit-time model (Crawford)
    the time of unit x
    Cumulative-average model (Wright)
    the average time of units 1 to x
  • Unit 2

    Unit-time model (Crawford)
    9.00 hours
    Cumulative-average model (Wright)
    8.00 hours (average of the two is 9.00)
  • Unit 8

    Unit-time model (Crawford)
    7.29 hours
    Cumulative-average model (Wright)
    6.24 hours (average of the eight is 7.29)
  • Unit 64

    Unit-time model (Crawford)
    5.31 hours
    Cumulative-average model (Wright)
    4.51 hours (average of the 64 is 5.31)
  • GAO says it suits

    Unit-time model (Crawford)
    Production that starts well prepared in tooling, suppliers and lead times
    Cumulative-average model (Wright)
    Starts with prototype tooling, a thin supplier base or early design changes, where averaging smooths the swings
  • Use it when

    Unit-time model (Crawford)
    You timed single units
    Cumulative-average model (Wright)
    You keep a running average

A filled-in example

Illustrative, not a benchmark

An example (illustrative, made-up numbers): a new bracket assembly cell. The first unit takes 9 hours, the team plans on an 85% unit-time curve, the standard time is 3.5 hours and the crew is 2 people at 30 productive hours a week each. The Example sheet and PDF page 4 hold the same figures.

  • Batch: units 21 to 40 take 81.2 hours, an average of 4.06 hours a unit, against 180 hours if every unit took 9.
  • Target: the time per unit reaches 4.5 hours at unit 20. The closed form (target divided by T1, raised to the power 1 over b, rounded up) and the table check both give 20, and building up to it takes 111.6 hours.
  • Plateau: the curve reaches the 3.5 hour standard at unit 57, after 254.4 hours in total.
  • Ramp: the crew builds 9, 12, 15, 16 and 18 units in weeks 1 to 5, then 17 a week, against a standard output of 17.1 a week. Unused hours carry over, so week 5 is above standard.
  • Fit: eight observed points (illustrative) give a fitted rate of 84.1% with R squared 0.994, close to the 85% the team assumed.

With these made-up numbers the cell needs about 111.6 hours to get to a 4.5 hour unit and reaches standard output in week 5 or 6. Your numbers will differ. The point is that the Fit sheet lets your own hours replace the 85%.

Common mistakes

  • Assuming a rate and never checking it

    A few points of rate change the hours a lot: moving from 85% to 90% adds about 33% to the hours for the first 100 units. Fit your own.

  • Mixing the two models

    A rate fitted to unit times is not the rate for the cumulative-average model. Choose the model that matches what you measured and keep to it.

  • Trusting R squared from two points

    A line through two points always fits exactly, so R squared is 1 and means nothing. Five or more points, spread across the run, are better.

  • Using it for work that is not repetitive

    A one-off job, a changing design or a new method partway through breaks the pattern. NASA's guide says a stated slope is an estimate, not a discrete value.

  • Leaving out the plateau

    Without a standard time the curve keeps falling below anything you can achieve. Set the standard time so the curve flattens there.

  • Planning on part-built units

    A week's output is whole units. The plan counts only completed units and carries unused hours into the next week.

Download the template

Free to use: print it, copy it and edit it for your team. Enter your name and work email once to download.

Run this template in LeanSuite

If new hires are your ramp-up, LeanSuite records each on-the-job training sign-off on a tablet, time-stamped, and the person's level updates on the skills matrix.

See it in a 45-minute demo

FAQ

Learning curve and ramp-up workbook: common questions

  1. The pattern that the more often a task is repeated, the less time each repeat takes. The usual form says each time cumulative output doubles, the time per unit falls to a fixed share, the learning rate. A 90% curve with a 10 hour first unit gives 9 hours for unit 2 and 8.1 for unit 4.

  2. They use the same formula but describe different things. In the unit-time model (Crawford) T1 x^b is the time of unit x. In the cumulative-average model (Wright) it is the average time of units 1 to x. The same rate gives different unit times, so pick the one that matches how you measure.

  3. Fit it from your own hours. Plot the hours for several units against the unit number on log-log axes and fit a straight line; the learning rate is 2 raised to the slope. The Fit sheet does this from two or more points and shows R squared. Five or more points spread across the run are better than two.

  4. We do not give one. A rate depends on the task, the people and the process, and we have no verified source for typical figures. NASA's own guide says a stated slope is an estimate, not a discrete value, and that a slope from a company's own history on work it knows can be accepted for that situation. Fit yours.

  5. For the unit-time model, the unit number is the target time divided by the first-unit hours, raised to the power 1 over b, rounded up. The Plan sheet works it out and checks it against the table. If the target is below your standard time the curve never gets there.

More free templates

All templates
Browse all 205 templates

Can't find the template you need?

Tell us what you're looking for. We'll make it and send it to your inbox, free.

A shift handover sheet, a red tag log, a changeover checklist for one machine: if your team uses it on the floor, ask.

By sending this form you agree to our Terms of Service and Privacy Policy.

Pass it on

We want LeanSuite to be the best place on the internet for lean help. If this was useful, send it to someone on your team or in your network who needs it.