10.3 Learning Curve Theory (Wright Cumulative vs. Crawford Unit Models)

10.3 Learning Curve Theory (Wright Cumulative vs. Crawford Unit Models)

In repetitive manufacturing, modular prefabrication, and serial construction operations (such as multi-story buildings, tract housing, pipeline spreads, and solar farm installations), the labor hours required to perform a task decrease as workers repeat the identical operation. This phenomenon is modeled in cost engineering through Learning Curve Theory (also known as the Experience Curve, Progress Curve, or Improvement Curve).

First quantified by Theodore Paul Wright in 1936 during his studies of aircraft manufacturing costs at the Curtiss-Wright Corporation, learning curve analysis is a vital tool for the Certified Cost Professional (CCP). Cost engineers use it to prepare competitive bids, negotiate change orders, forecast cash flows, and establish production targets.

---

1. Core Principles & Sources of Learning

Learning curve theory is governed by a fundamental axiom:

Each time the cumulative production quantity of a repetitive product doubles, the time (or cost) required per unit decreases at a constant percentage rate.

+-----------------------------------------------------------------------------+
|                        SOURCES OF PRODUCTIVITY GAINS                        |
|                                                                             |
|   1. OPERATOR PROFICIENCY:  Muscle memory, dexterity, familiarity with      |
|                             tools, drawings, and physical assemblies.       |
|   2. TOOLING & JIGS:        Development of specialized fixtures, custom     |
|                             templates, and ergonomic holding rigs.          |
|   3. METHOD ENGINEERING:    Optimization of work sequence, waste elimination|
|                             (Lean principles), and reduced motion fatigue.  |
|   4. MANAGEMENT LOGISTICS:  Improved material staging, standardized parts,   |
|                             and reduced supervisory instruction time.       |
|   5. QUALITY REDUCTION:     Fewer fabrication defects, less non-destructive |
|                             testing (NDT) rejection, and reduced rework.    |
+-----------------------------------------------------------------------------+

Learning Rate (LR) & Rate of Learning (RoL):

• Learning Rate (LR or Progress Ratio PR): The percentage of labor hours (or cost) remaining after a doubling of production quantity (e.g., 80% or 0.80).
• Rate of Learning (RoL / Percentage Reduction): The percentage reduction achieved per doubling: RoL = 1 - LR (e.g., for an 80% curve, RoL = 1.0 - 0.80 = 20% reduction).
• Zero Learning (100% Curve): Completely machine-paced operations where human repetition yields 0% time reduction (LR = 1.00).

---

2. The Two Primary Mathematical Models: Wright vs. Crawford

In cost engineering, two distinct mathematical formulations exist. Distinguishing between them is essential on the CCP exam:

+-----------------------------------------------------------------------------+
|                     WRIGHT VS. CRAWFORD MATHEMATICAL MATRIX                 |
|                                                                             |
|   MODEL:              T.P. WRIGHT (1936)        J.R. CRAWFORD (1944)        |
|                       "Cumulative Average Model" "Individual Unit Model"    |
|   ------------------  ------------------------- --------------------------  |
|   Applies LR to:      CUMULATIVE AVERAGE of     INDIVIDUAL x-th UNIT        |
|                       all units produced so far direct unit time            |
|                                                                             |
|   Primary Formula:    Y_bar(x) = k * x^s        y(x) = k * x^b              |
|                                                                             |
|   Slope Parameter:    s = ln(LR) / ln(2)        b = ln(LR) / ln(2)          |
|                                                                             |
|   Total Hours (Tx):   Tx = x * Y_bar(x)         Tx = Sum_{i=1}^x y(i)       |
|                       Tx = k * x^(1 + s)        Tx ~ [k / (1+b)] * x^(1+b)  |
|                                                                             |
|   Marginal Unit x:    y(x) = Tx - T(x-1)        y(x) = k * x^b              |
|                                                                             |
|   Historical Domain:  Aerospace, defense,       Discrete manufacturing,     |
|                       long-cycle programs       modular fabrication, constr.|
+-----------------------------------------------------------------------------+

---

3. T.P. Wright's Cumulative Average Model

Under Wright's model, the cumulative average time [Y_bar(x)] across the first x units decreases by the factor LR whenever cumulative output doubles:

Y_bar(x) = k * x^s
ln[Y_bar(x)] = ln(k) + s * ln(x)

Where:

• Y_bar(x) = Cumulative average labor-hours (or cost) per unit for the first x units
• k = Labor-hours (or cost) required to produce the first unit (Unit 1)
• x = Cumulative production quantity
• s = ln(LR) / ln(2) = Learning curve slope index (s < 0)

Total Program Hours (Tx) and Individual Unit Hours [y(x)]:

Tx = x * Y_bar(x) = x * [k * x^s] = k * x^(1 + s)
y(x) = Tx - T(x-1) = k * [x^(1+s) - (x-1)^(1+s)]

Wright's Doubling Table (k = 1,000 hours, LR = 80%, s = -0.32193):

---

4. J.R. Crawford's Unit Model (Stanford Model)

Under Crawford's model, the direct labor-hours required to produce the individual x-th unit [y(x)] decreases by the factor LR whenever the unit number doubles:

y(x) = k * x^b
ln[y(x)] = ln(k) + b * ln(x)

Where:

• y(x) = Direct labor-hours (or cost) required to manufacture the specific x-th unit
• k = Labor-hours (or cost) for the first unit (Unit 1)
• b = ln(LR) / ln(2) = Learning curve slope index (b < 0)

Total Program Hours (Tx) under Crawford:

Tx = Sum_{i=1}^x y(i) = k * Sum_{i=1}^x [i^b] ~ [k / (1+b)] * x^(1+b)   (continuous approximation)
Y_bar(x) = Tx / x ~ [k / (1+b)] * x^b

Crawford's Doubling Table (k = 1,000 hours, LR = 80%, b = -0.32193):

---

5. Critical Comparison: Wright vs. Crawford for the Same Learning Rate

+-----------------------------------------------------------------------------+
|               COMPARISON OF TOTAL HOURS FOR 16 UNITS (k = 1,000 LH, LR = 80%)|
|                                                                             |
|   MODEL              UNIT 16 TIME      TOTAL 16 UNITS    CUMULATIVE AVERAGE |
|   -----------------  ----------------  ----------------  ------------------ |
|   Wright Cumulative  280.0 Hours       6,553.6 Hours     409.6 Hours/Unit   |
|   Crawford Unit      409.6 Hours       8,919.7 Hours     557.5 Hours/Unit   |
|                                                                             |
|   KEY TAKEAWAY: Wright's model produces significantly faster total program  |
|   labor reduction and lower total hours than Crawford's model for the same  |
|   nominal learning rate percentage!                                         |
+-----------------------------------------------------------------------------+

---

6. Industry Typical Learning Rates

Learning rates vary across industries depending on the ratio of manual craft labor to automated machine control:

+-----------------------------------------------------------------------------+
|                   TYPICAL INDUSTRY LEARNING RATE BENCHMARKS                 |
|                                                                             |
|   LEARNING RATE (LR)   INDUSTRY / OPERATION TYPE                            |
|   ------------------   --------------------------------------------------   |
|   70% - 75%            Aerospace, manual airframe assembly, complex piping  |
|                        (High manual labor content, ~75% labor / 25% machine)|
|                                                                             |
|   80% - 85%            Standard manufacturing, modular precast fabrication, |
|                        electromechanical assembly (50% labor / 50% machine) |
|                                                                             |
|   90% - 95%            Repetitive construction operations: multi-story floor|
|                        concrete framing, pipeline spreads, solar PV racking |
|                        (High mechanized content, ~25% labor / 75% equipment)|
|                                                                             |
|   100%                 Fully automated robotic machining / continuous       |
|                        chemical refining (Zero human learning effect)       |
+-----------------------------------------------------------------------------+

---

7. Step-by-Step Worked Numerical Case Studies

Case Study 1: Crawford Unit Model — Batch Cost Forecasting

Problem: A modular fabrication yard is manufacturing 8 identical pre-assembled pipe racks (PARs). The first unit (Unit 1) required 1,200 labor-hours. The contractor's historical learning curve for pipe rack modularization is 85% (LR = 0.85).

1. Calculate the labor hours required to fabricate the 8th unit [y(8)].
2. Calculate the labor hours required for the 5th unit [y(5)].

Solution:

• Step 1: Compute Slope Index (b): b = ln(0.85) / ln(2) = -0.1625189 / 0.6931472 = -0.234465
• Step 2: Find Hours for Unit 8 via Doubling Rules:
  • Unit 1: 1,200.0 hrs
  • Unit 2 (1st doubling): 1,200 * 0.85 = 1,020.0 hrs
  • Unit 4 (2nd doubling): 1,020 * 0.85 = 867.0 hrs
  • Unit 8 (3rd doubling): 867 * 0.85 = 736.95 hrs
• Step 3: Find Hours for Unit 5 using Crawford Equation: y(5) = k * x^b = 1,200 * 5^(-0.234465) 5^(-0.234465) = e^[-0.234465 * ln(5)] = e^[-0.234465 * 1.609438] = e^(-0.377356) = 0.685673 y(5) = 1,200 * 0.685673 = 822.81 labor-hours

---

Case Study 2: Wright Cumulative Average Model — Program Bid Estimating

Problem: A defense contractor bids on a production batch of 4 specialized radar modules. Unit 1 requires 5,000 labor-hours at a direct labor rate of $80/hour. The contract specifies an 80% Wright Cumulative Average curve.

1. What is the cumulative average hours per unit for the 4 units?
2. What is the total direct labor cost for the entire 4-unit contract?

Solution:

• Step 1: Calculate Cumulative Average for 4 Units (2 doublings):
  • Y_bar(1) = 5,000 hrs
  • Y_bar(2) = 5,000 * 0.80 = 4,000 hrs
  • Y_bar(4) = 4,000 * 0.80 = 3,200 hours/unit
• Step 2: Calculate Total Labor Hours for 4 Units (T4): T4 = x * Y_bar(4) = 4 * 3,200 hrs = 12,800 total labor-hours
• Step 3: Calculate Total Direct Labor Cost: Total Cost = 12,800 LH * $80/LH = $1,024,000

---

8. Disruptors, The Forgetting Curve & The Plateau Effect

In real-world construction and manufacturing, learning curve efficiency gains do not continue indefinitely and can be severely disrupted:

+-----------------------------------------------------------------------------+
|                   LEARNING CURVE DISRUPTORS & LIMITATIONS                   |
|                                                                             |
|   1. DESIGN MODIFICATIONS & CHANGE ORDERS:                                  |
|      - Altering drawings midway through production destroys muscle memory,  |
|        invalidates prefabricated jigs, and shifts the curve upward.         |
|                                                                             |
|   2. THE FORGETTING CURVE (Production Breaks & Gaps):                       |
|      - Shutting down production for weeks/months results in skill decay.    |
|      - Upon restart, unit hours spike backward toward initial Unit 1 levels.|
|                                                                             |
|   3. CRAFT LABOR TURNOVER:                                                  |
|      - Replacing experienced workers with new hires dilutes collective gang |
|        proficiency and resets the learning progression.                     |
|                                                                             |
|   4. THE PLATEAU EFFECT (Technological Floor):                              |
|      - After significant volume (e.g., 50-100 units), human learning gains  |
|        reach a physical plateau dictated by machine cycle limits and        |
|        material handling constraints. No further reduction occurs.          |
+-----------------------------------------------------------------------------+

[!IMPORTANT] AACE CCP Exam Alert — Learning Curve Key Principles:

• Wright's Model applies learning rate LR to the Cumulative Average [Y_bar(x) = k * x^s].
• Crawford's Model applies learning rate LR to the Individual Unit [y(x) = k * x^b].
• The exponent parameter is s = b = ln(LR) / ln(2). For an 80% curve, s = -0.322; for an 85% curve, s = -0.234; for a 90% curve, s = -0.152.
• For the same nominal LR, Wright yields fewer total hours than Crawford.