Unit Learning Curve Calculator: How to Use It

This calculator works out the time or cost per unit after a given number of production units, based on the learning curve model. You enter the time for the first unit, the learning rate (as a percentage), and the unit number you want to estimate; you get back the predicted time for that unit and the cumulative average time. It uses the standard Crawford cumulative average model, which is the most common in manufacturing and project cost estimation.

Formula

The learning curve is expressed as:

Y_n = a × n^b

Where:

• Y_n = time (or cost) to produce the nth unit, in the same unit as a (e.g., hours, dollars).
• a = time to produce the first unit (your input t).
• n = the unit number (your input n).
• b = the learning curve exponent, calculated as b = log(r) / log(2), where r is the learning rate expressed as a decimal (e.g., 80% = 0.80).

The exponent b is negative for learning rates below 100%, meaning each doubling of cumulative production reduces the unit time by a constant percentage. For an 80% learning curve, b = log(0.80)/log(2) ≈ -0.3219.

The cumulative average time (the second output) is the average time per unit over the first n units. The formula used here is the Crawford cumulative average: Average = a × n^b / (1 + b). This is an approximation that works well for n ≥ 10 and is standard in cost estimating.

Worked Example

Using the default values: first unit time = 100 hours, learning rate = 80%, unit number = 10.

1. Convert learning rate to decimal: 80% = 0.80.
2. Calculate b: b = log(0.80) / log(2) = -0.09691 / 0.30103 = -0.3219.
3. Calculate time for unit 10: Y₁₀ = 100 × 10^-0.3219 = 100 × 0.4765 = 47.65 hours.
4. Cumulative average: Average = 47.65 / (1 – 0.3219) = 47.65 / 0.6781 = 70.27 hours.

The calculator shows 47.65 hours for unit 10 and 70.27 hours as the cumulative average. That matches the formula outputs exactly.

When to Use a Learning Curve Calculator

Learning curve analysis is used in manufacturing, project management, and cost estimation whenever repetitive tasks are performed by humans or teams. Typical applications include:

• Estimating labour hours for aircraft, ship, or electronics production.
• Forecasting assembly time for new products as workers gain experience.
• Setting piece-rate wages or production targets.
• Comparing the effect of training programs on productivity.

The model assumes that each time cumulative production doubles, the time per unit drops by a fixed percentage. That percentage is the learning rate. An 80% learning rate means that when you double the total units produced, the time for the latest unit is 80% of what it was before the doubling.

How to Choose the Learning Rate

The learning rate is not a fixed number; it depends on the industry and the nature of the task. Common values are:

• 70–75% for complex, manual tasks with little mechanisation.
• 80% for typical assembly work with some machine assistance.
• 85–90% for highly automated processes where the human factor is small.

You can estimate the learning rate from historical data by plotting unit times against cumulative units on log-log paper (or a spreadsheet) and fitting a straight line. The slope of that line is the exponent b, and the learning rate is 2^b. If you don’t have data, use an industry benchmark or start with 80% and adjust as actual results come in.

Common Mistakes and Limitations

One frequent error is using the wrong formula. There are two main models: the Crawford unit model (used here) and the Wright cumulative average model. The Wright model calculates the average time for all units directly, not the time for a specific unit. Mixing them up gives different results.

Another mistake is forgetting that the learning rate applies per doubling of cumulative production, not per unit. A 10% improvement per unit is not the same as an 80% learning rate.

Rounding too early can also skew results, especially for high unit numbers. Keep at least four decimal places in the exponent b.

The model has limits: it assumes a constant learning rate, which real processes rarely have. It also breaks down for very small n (n=1,2) – the formula is derived for n≥10. For n=1, the formula gives Y₁ = a, which is correct, but the cumulative average formula is not accurate for small n.

Practical Example: Aircraft Production

Suppose a company builds a new fighter jet. The first jet takes 500,000 hours to produce. With an 80% learning curve, the 100th jet would take:

Y₁₀₀ = 500,000 × 100^-0.3219 = 500,000 × 0.2271 = 113,550 hours.

The cumulative average for the first 100 jets would be 113,550 / (1 – 0.3219) = 167,500 hours per jet. That means the total labour for 100 jets is 16.75 million hours. This kind of estimate helps in pricing contracts and planning workforce needs.

Frequently Asked Questions

What is the difference between unit and cumulative average learning curve?

The unit model predicts the time for a specific unit (e.g., the 50th unit). The cumulative average model predicts the average time per unit over a range of units (e.g., units 1 to 50). The calculator gives both, because they answer different questions. For cost estimation, the cumulative average is often more useful because it gives total labour hours.

Can I use this for cost instead of time?

Yes. The learning curve applies equally to direct labour cost, material cost, or total cost, as long as the cost follows the same percentage improvement pattern. Simply enter the first unit cost instead of time, and the output will be in dollars (or any currency) instead of hours.

What does a 90% learning curve mean?

It means that every time cumulative production doubles, the time per unit falls to 90% of its previous value. For example, if unit 1 takes 100 hours, unit 2 takes 90 hours, unit 4 takes 81 hours, unit 8 takes 72.9 hours, and so on. A higher learning rate means slower improvement.

How do I find the learning rate from real data?

Take the log of the unit number and the log of the time per unit for several units. Plot log(time) vs log(unit number). The slope of the best-fit line is b. Then the learning rate = 2^b. For example, if b = -0.3219, the learning rate is 2^-0.3219 = 0.80, or 80%.

Does the learning curve apply to software development?

Yes, but with caution. Software tasks are less repetitive than manufacturing, and the learning rate varies widely. Some studies suggest a 70-80% learning curve for coding tasks, but team composition and technology changes can break the pattern. Use it as a rough guide, not a precise prediction.

What is the difference between learning curve and experience curve?

The learning curve typically refers to direct labour time per unit, while the experience curve is a broader concept that includes all costs (materials, overhead, etc.) and applies to the entire industry, not just one firm. The experience curve often has a lower percentage (e.g., 70-80%) because it captures more improvement sources.

Related

• Pool Turnover Rate Calculator: Formula & Guide
• Advanced Earth Curve Calculator: See the Horizon
• Monthly Turnover Rate Calculator & Formula

in t | Time for first unit | hours | 100 | 0
in r | Learning rate | % | 80 | 1
in n | Unit number | - | 10 | 1
out time | Time for unit n | hours | t * (n ^ (log(r/100) / log(2))) | 2
out avg | Cumulative average time | hours | (t * (n ^ (log(r/100) / log(2)))) / (1 + (log(r/100) / log(2))) | 2