Learning Curve Cost Estimator
Estimates the direct labor hours and full bid price for a follow-on production lot using learning curve analysis — under both the Wright (cumulative average) and Crawford (unit) models — then converts those hours into labor cost, material cost, and a recommended bid price at your target margin. Also shows how much you'd overprice the same lot by assuming no further learning. Built for cost estimators and program managers in aerospace, defense, shipbuilding, and repetitive manufacturing who need to price a follow-on order, not just calculate an abstract hours-per-unit figure.
Download This Calculator
Get the Excel spreadsheet behind this calculator to use offline, customize for your needs, and publish as a web tool using Sheetflow.
Lot-Specific Hour Estimates
Estimates hours for units N through M — the follow-on lot you're actually bidding — by subtracting cumulative hours through prior units, not by averaging the whole program.
Wright and Crawford Side by Side
Calculates both the cumulative average (Wright) and unit (Crawford) models from the same inputs, and shows the hour difference between them so you can see what the model choice costs.
Bid Price, Not Just Hours
Converts lot hours into labor cost, material cost, and a recommended bid price at your target margin — then shows how much a flat-rate bid would overprice the same lot.
Frequently Asked Questions
What is a learning curve, and how do I pick the right percentage?
A learning curve captures the well-documented pattern that direct labor hours per unit fall by a constant percentage every time cumulative production doubles. An 85% learning curve means that when output doubles, the relevant hours figure drops to 85% of its prior level — a 15% reduction per doubling.
The percentage depends heavily on how much of the work is human versus machine. NASA's widely cited guidance suggests roughly 80% for processes that are about three-quarters hand labor, 85% for an even mix of hand and machine work, and 90% for processes that are mostly machine-driven — the intuition being that machines don't learn, so automation-heavy processes show less improvement from repetition. Using the calculator's defaults, an 85% curve with 100 hours for the first unit produces an exponent of −0.2345, which drives every downstream calculation.
What's the difference between the Wright and Crawford learning curve models?
They apply the learning percentage to different things, and this genuinely matters. The Wright model (cumulative average theory) says the cumulative average hours per unit drops by the learning percentage each time output doubles. The Crawford model (unit theory) says the hours for an individual unit drop by the learning percentage each doubling. Same input percentage, different math, different answer.
Using the calculator's defaults — 100 hours for unit one, an 85% curve, 200 units already built, and a 300-unit follow-on lot — Wright estimates 5,871 hours for the lot while Crawford estimates 7,666 hours, a difference of about 1,795 hours. At a $95 loaded rate, that's roughly $170,000 of labor cost riding on which model you use. Wright is more common in aerospace and defense estimating; Crawford appears more often in shipbuilding and some government contracting contexts. Confirm which one your contract, customer, or internal estimating standard expects before submitting a bid.
How do I calculate hours for a follow-on lot rather than a whole program?
Calculate cumulative hours through the end of the new lot, then subtract cumulative hours through the units you'd already built — the difference is the hours attributable to just the new lot. This matters because a follow-on lot benefits from all the learning that already happened on prior units, so it should be meaningfully cheaper per unit than the original production run.
Using the calculator's defaults: cumulative hours through 500 total units comes to 11,645, and cumulative hours through the 200 prior units comes to 5,775. The difference — 5,871 hours — is what the 300-unit follow-on lot should actually take, averaging 19.57 hours per unit. Compare that to the 100 hours the very first unit required, and you can see why estimating a follow-on lot from first-unit data (or from a program-wide average) produces badly wrong numbers.
What happens if I ignore the learning curve when bidding a follow-on order?
You overprice the bid, often enough to lose it. The most common shortcut is assuming the new lot will run at your current cumulative average rate — reasonable-sounding, but wrong, because learning continues as you build more units, and the new lot's average will be lower than your running average to date.
Using the calculator's defaults: the current cumulative average is 28.87 hours per unit, so a flat-rate estimate for 300 more units gives 8,662 hours and a bid of $1,197,342. Properly applying the learning curve gives 5,871 hours and a bid of $865,916 — meaning the flat-rate approach overprices by $331,426, or 38.3%. On a competitive procurement, a 38% overbid is not a close call; it's the difference between winning and not being in the conversation.
When does learning curve analysis stop being reliable?
The theory assumes reasonable continuity in the production environment — no major redesign, no change in production process or tooling, a stable workforce, and no long gaps between production runs. When any of those assumptions break, learning can stall or even reverse, and a curve fitted to prior data will overstate future efficiency.
Extended production breaks are a particularly common trap, since workforce turnover and skill decay during a gap can effectively reset part of the accumulated learning — some estimating practices apply a "forgetting" or retrograde adjustment for exactly this reason. Similarly, learning curves flatten at high cumulative quantities: the improvement from unit 1 to unit 2 is dramatic, while the improvement from unit 5,000 to unit 10,000 is often negligible in practice even though the math predicts the same percentage drop. For very high-volume mature programs, validating the curve against recent actuals matters more than trusting the original curve fitted years earlier.
Transform Your Excel Models into Web Tools
Turn your complex Excel calculations into online calculators, web forms, and APIs. No coding required — upload your spreadsheet and publish your calculations instantly.
Calculations are for estimation and planning purposes. Users should verify important results for their specific situations. No signup required. Calculations performed securely.