Learning Curves

LEARNING CURVES IN SERVICES Usw:; Somwa. fCfl lEAIINING Cl.fl\lES

AND MANUFACTURING Sil.YEO l'MBlEMS
APPLYING THE LEARNING CURVE INTEllf,tl ANO S1\IOENl CO-ROM ExERCISES
Arithmetic Approach 01SOOSSKIN 0 UESTIOHS
Logarithmic Approach AcnVE MOOEL EXEROSE
Leaming,Cutve Coefficient Approach PlioBlfMS
STRATEGIC IMPLICATIONS INTEllf,tl HOMl'n~AI( PROBlEMS
LEARNING OBJECTIVES
OF LEARNING CURVES
CAsE s,uo,: SMrs NEoo,,.,~• wm, IBM \Vhe11 you complne this module you
LIMITATIONS OF LEARNING CURVES
Bl!llllXiRAl'KV should be able. lo
SU\OMRY
INTEllf,tl RES-ES
Kr< TERM
IDENTIFY OR DEFINE:
What a learning curve is
Example of leaming cur\'es
111e doubJing concept

DESCRIBE OR EXPLAIN:
How to compute learning curve effects
Why learning curves are important
The strategic implicmions of learning
curves
772 M ODU L E E, L EA R N I NG CURVES

Me.dkal procedures sud, as hean :mrgeryfollow « leami11g cun~. Rtse.ard, indfrx1trs tl1<1t 1l1c death rote from hr<1t1
transplants drc,ps at" 79% !taming cun·e. « !taming rote 1101 unlike. thm in many illdustrial st:ttings, It <1p11e-ars
11,at ,u ductors a11d mnlical It.ams imprt1t•e w11l1 e>.pcrle11ce, :w do yo11r odds as" p(llient, Ifthe de<1th rotr ls l1a/1,"('d
tl't'')' th1-r-c u11onti()11s. practic< m<1y l11drtd make pe,fect,

Most organizations learn and improve over cime. As firms and employees perfonn a task over and
over. they learn how to perform more efficiently. Tllis means that task times and cosL~ decrease.
Leaming curves Learning curves are based on the premise that people and organizations become beuer at their
The p,emise lhot people tasks as the tasks are repealed. A learning cur"e gr.tph (illustrntcd in Figure E.l) displays l;tbor-
a nd organizations get hours per unit versus the number of units produ<.'Cd. From il we sec that the time needed to produce
better ot theil toslcs os a unit decreases, mma11y following a negative exponential cur\'e, as the person or company produces
the tosl<s ore repeoted; more units. ln other words, it rakes less time to complete each additional unit a firm produces.
sometimes called
However, we also sec in Figure E. 1 that the time savings in completing each subsc<1uent unil
experience cuNes.
det:rettses. These are the major attributes of ll1e learning curve.
Leaming curves were first applied 10 industry in a repon by T. P. Wright of Cunis-Wright Corp.
in 1936.I Wright described how direct labor costs of making a partkular airplane decreased with
teaming, a theory since confimted by other airc-raft manufacturers. Regardless of the tjme needed to
produce ll1e first plane, learning curves are found to apply to various categories of afr frames (e.g.,

FIGURE E.1 ■

The Leornlng.Curve
Effect Stotes Thot Time
per Repetition
Decrea ses as the
Numbe r o f Repe titions
Increases

0 Numbor of repetitions (VOiume)

1T. P. Wright, ·•1;.ie1ors Affecting lhc Cos1of Airplanes:· Joomol <>/ 1/1e Mrorw111ia1I Sci('n<('S (February 1936).
 L t:ARN I NC CURVt:S IN Se1w 1Ct:S AND MANUl'l"ACTURINO 773

jeLfighters versus passenger planes versus bombers). Leaming curves have since been upplied not
only to labor but also to a wide varieLy of other costs. including material and purchased components.
The power or the learning curve is so significant thal it plays a m:tjor role in many str.ttegic deci-
s ions rcktted to employment levels, costs, capacity, and pricing,
T he learning curve is based on a doubling of production: T hat is. when production doubles, the
decrease i n time pel' unit affects the rnte of lhe teaming curve. So, if the lea.ming curve is an 80%
rate, the second unit takes 80% of the time of the first unit. the fourtl1 unit takes 80% of the time of
the second unit, the eighth unit takes 80% of the time of lhe fourth unit. and so fo11h. This pl'inciple
is shown as

T x L" = Time required for the nth unit ( E-1)

where T = unit cost or unit time of the first unit

l = le:lrning curve rale
,, = number of times Tis doubled

lf the first unit of a particular product took IO labor-hours, and if a 70% lea.ming curve is present, the
hours the fourth uniL will take require doubling twi<.-c-from I to 2 to 4. T herefore, the formula is

Hours required for unit 4 = 10 x (.7)2 = 4.9 hours

LEARNING CURVES IN SERVICES AND MANUFACTURING

Try testing the leaming- Different org:miz;ttions-indoed, differe nt products-have different learning curves. The rate or
curve effect on some lea.ming varies depending on the <1uality of management and the potential of the pro<.-css and prod-
activity you may be uct. Any cha11ge in process, product, or pusonnef disrupts the Leaming curve. Therefore, c:tution
performing. For example, should be exercised in assuming ll1at a learning c.urve is oonLinuing and pemrnnent.
if you need to assemble As you can see in Table E.1 , indusny learning curves val'y widely. 111e lower the numbel' (s.ay
four bookshelves, time
70% compared to 90%), the steeper the slope and the faster ll1e drop in costs. By tradition, learning
your work on each and
note the rate of curves are defined in terms of the c,>mplemems of their impl'ovemem rates. For example. a 70%
improvement. learning cun·e implies a 30% decrease in Lime each time the number of repeti tions is doubled. A
90% cur\'e means there is a corresponding I0% mLe of impro"emenl.
Stable, standardiz.cd products and processes tend to have <.'OSIS th:ll decline more sleeply than
others. Between 1920 and 1955. for insL:mce. Lhe steel industry was able to reduce labor-hours per
unit to 79% each time cumulative production doubled.
Leaming curves have application in services as well as industry. As was noLcd in the caption for
the opening photogmph, I-year death races of he.art transplant patiems at Temple U ni"ersity
Hospital follow a 79% le.arning c.u n•e. The l'esults of that hospital"s 3-yeal' study of 62 paliems

TABLE E.1 ■
L •EARNJ:-.G-
Examples of Leaming- CURVR
Curve Effects Cm,1uLA·m•i:; SLOPK
IMPROVING PARAMF.TF.R pARAMF.TRR (%)
I. Model-TFol'd production Price Units produced 86
2. Aircraft IL,;scmbl)' Dirct.i labor hours per unit
4
Units produced 80
3. Equipmet11 maintenance Aveiage lime to replace a Number of replacements 76
at GE group of pans
4. Stoel produc-tion Production \1.:orkcr labor-hours Units produced 79
per uni1 produced
S. Integrated circuits Average price per unit Units produced 72"
6. Hand-held calculator Average factory .selling prk-c Units produced 74
7. Disk memory drh·es Avei.-ige price per bil Number of bi1s 76
8. He-art transplants 1-year dealh rates Transplanls complelcd 79
oConitont dOIIOfS.
sovrcos: JOm8'$ A. O J ~ -IJ$ing lh& Leorl't"'Q Cul"Y$ OS O ~ r ' l l TOOi. ~ IEE£ $p61C!n..m(Jol'l8 1980): 45,
C 1980 IEEE: OM OOvid B. Smith Or.d Jon l , LOlsson, -Tho lmp¢CI or L&Orl'ing on Cost: TM C<:z$0 Of H80rl
Tre71$pl0ntotlon, • Hospital 0/'ld HeOlth SeMcesAdm(nl'Sftoflon(sprlng 1989): 85-97.
774 MODULE E, LEARNING CURVES

receiving transplants found that every throe opcrmions resulted in a halving or the I-year death rnte.
As more hospita1s face pressure from both insurance companies and the government to enter fixed-
price negotiations for their services. their ability to learn from experience b<:comes incre;lsingly crit-
ical. ln addition to having applications in both services ;md industry. learning curves arc useful for a
Failure to consider the variety of purposes. These include:
effects of learning can lead
1. lntermLI: labor forecasting, scheduling, establishing costs and budgctS.
to overestimates of labor
needs and underestimates 2. ExternaJ: supply-chain negotiations (see the SMT c-ase study at the end of this module).
of material needs. 3. Strategic: evalu;Hion of company and industry perfonnance, inc1uding costs and pricing.

APPLYING THE LEARNING CURVE

A mathematical rcl;ttionship enables us to express the time required to produce a cert;lin unit. This
re lationship is a function of how many units have been produced before the unit in question and how
long it took to produce them. Although this procedure determines how long it takes to produce a
gh·en unit, lhe consequences of th is anaJysis are more far reaching. Costs drop and efficiency goes
Trade j ournals publish up for individual firms and the industry. Therefore, severe problems i n scheduJing occur if opera-
industrywide data on tions are not adjusted for implications of the learning curve. For instance, if learning-curve improve-
specific operations' ment is not considel'ed when scheduling. l11e result may be labor and pl'oduc.tive facilities being idle
learning rates. a portion of the lime. Funhennore. firms may l'efuse additional work because they d o not considel'
the improvement in the ir own efficiency th;H results from lea.ming. From a s upply-chain perspective,
our interest is in negotiating what our suppliers' costs s hould be for further production of units based
on the size of an order. The foregoing arc only a few or the rnmifications or the effect of learning
curves.
With this in mind. let us look at three ways to approach the m;Hhematics of learning curves: arith-
metic analysis, logarithmic a.naJysis. and learning-curve coefficients.

Arithmetic Approach
The arithmetic ;tpproach is the simplest approach to learning-curve problems. As we noted at the
beginning of this module, each time that production doubles, labor per unit declines by a const;mt
factor, known as the learning rate. So. if we know that the learning rate is 80% and that the first unit
produced took I00 hours, the hours required to produce the second, fourth, eighth, and s ixteenth
units are as foJlows:

NTH U~1T PRODUCRD HOL'RS fOR Nm Usrr

I 100.0
2 80.0 = (.8 x IO0)
4 64,0 = (,8 X 80)
8 5l.2=(.8 x 64)
16 41.0 = (,8 X 51.2)

As long ;ts we wish to find the hours required to produce N units and N is one of the doubled values,
then this approach works. Arithmetic analysis docs not tell us how many hours will be needed to
produce other units. For this flexibil ity. we must turn to the logarithmic approach.
TABLE E.2 ■

Leomlng Curve Logarithmic Approach

Voluesor b Tlle logaJ"il11mic approach alJows us to determine Jabor for any unit, TN. by the fonnula

L EAKNIN(.;
(E,2)
R,n,(%1 b
70 - .5 15 w here 1:v = time for l11e N th uni t
75 -.4 15 7 1 = hours to produce l11e first u nit
80 - .322
b = ( log of l11e learning rate)/(log 2) = slope of the ]earning cuJ"ve
85 - .234
90 - .152
Some of th e values for bare presented in Table E.2. Example E 1 shows how this fonnula works.
 APPLYING THE L EARN I NC CU RVI'.: 775

Example El The learning rote for a particular operation is 80%. and the first unit of production took 100 hours. The
Using logs to compute hours required to produce the third unit may be computed IL'> follows:
looming curves
TN e Tj(Nh)
Excel OM T3 = (100 hours)(3•)
•
Data File = (100)(3~•-• 1"'•2)
ModEExEl.xla
= (IOO)(rm) = 70.2 labo<-hours

The logarithmic approach allows us to determine the hours required for a11y unit produced, but there
is a simpler method.

Learning-Curve Coefficient Approach

The learning-curve coefficient technique is embodied in Table E.3 and the following cqu:Hion:

( E-3)

where TN= number of ktbor-hours re<1uired to produce the Nth unit

7 1 = number of ktbor-hours re<1uircd to produce the first unit
C = learning-curve coefficient found in Table E.3

The leaming-cun•e coefficient, C. depends on both the teaming rate (70%, 75%. 80%, and so on)
and the unit of interest.

TABLE E.3 ■ learning-Curve Coefficients, Where Coefficient, C = N (tog o11eornlng rOle/log 2)

70% 75% 80% 85% 90%

UNIT
Nuri.mF.R Usrr TOTAi. U~1T TOTAL UNIT T<rrA(, UNIT TOTAL UNlT TOTAL
(/V)TIME n:,.u: TIMt 1))1.K TIME TIMt TIME TIME Tm~ Trn.t TL\L~

I 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000
2 .700 1.700 .750 1,750 .800 1.800 .850 1.850 .900 1.900
3 .568 2.268 .634 2.384 .702 2.502 .773 2.623 .846 2.746
4 .4')() 2.758 .562 2.946 .640 3.142 .723 3.345 .810 3.SS6
5 .437 3.195 .513 3.459 .596 3.738 .686 4.03) .783 4.339
6 .398 3.593 .475 3.934 .562 4.299 .651 4.688 .762 S.101
7 .367 3.960 .446 4.380 .534 4.834 .634 S.322 .744 S.845
8 .343 4.303 .422 4.802 .5 12 5.346 .614 5.936 .729 6.574
9 .323 4.626 .402 S.204 .493 S.839 .591 6.533 .716 7,2')()
10 .306 4.932 .385 5.589 .477 6.3 15 .583 7.1 16 .705 7,994
II .291 5.223 .370 5.958 .462 6.777 .570 7,686 .695 8.689
12 .278 S.SOI .357 6.315 .449 7.227 .SS8 8.244 .685 9.374
13 .267 5.769 .345 6.(,<)0 .438 7.665 .548 8.792 .677 10.052
14 .257 6.026 .334 6.994 .428 8.0!12 .539 9.33) .670 10.721
IS .248 6.274 .325 7.319 .4 18 8.S I I .530 9.861 .663 11.384
16 .240 6.514 .316 7.635 .4 10 8.920 .522 10.383 .656 12.040
17 .233 6.747 .309 7,944 .402 9.322 .515 10.898 .650 12.690
18 .226 6.973 .301 8.245 .394 9.716 .508 11.405 .644 13.334
19 .220 7.192 .295 8.540 .388 )0,)04 .501 11.907 .639 13.974
20 .214 7.407 .288 8.828 .381 l0.485 .495 12.402 .634 14.608
25 .19 1 8.404 .263 10.19 1 .355 12.309 .470 14.801 .613 17.713
30 .174 9.305 .244 11 ,446 .335 )4.020 .450 17.0!ll .596 20.727
35 . )60 10,133 .229 12.618 .318 15.643 .434 19.294 .583 23.666
40 .ISO 10.902 .2 16 13.723 .305 17.193 .421 21.425 .571 26.543
45 .141 11.625 .206 14.773 .294 18.684 .410 23.500 .561 29.366
50 ,)34 12.307 .197 15.776 .284 20.122 .400 25.5 13 .552 32.142
776 MODULE E, LEARNING CURVES

Example E2 uses the preceding equation and Table E.3 to calculate learning-curve effects .

Example E2 ll took a Korean .shipyard 125.000 labor-hours to produce the fir.st of .several lugboat.s thal you cxpecl to
Uslng learning-c urve purcha.-.e for your .shipping company. Great Lakes. Inc. Boats 2 and 3 have been produced by the KorearL,;
coefficients with a learning factor of 85%. Al S40 per hour. whal should you, as purchasing agent, expec1 to pay for 1he
fourth uni1?
Excel OM Fir.;t. search Table E.3 for the fourth unit and a learning factor of 85%. The learning-cur..-e coc.ffident.
Data File C, is .723. To produce 1be fourth unit. then. takes
• ModEExE2.xla
TN = T1C
T, = (125,000 hours)(.723)
• Active Model E. 1 = 90.375 hour.;
To find the co.st mulliply by $40:
Examples E2 and E3
are further illustrated 90,375 hours x S40 per hour= 53,615,000
in Active Model e.1on
the CD-ROM and In the
Exe<cise on page 780. 'fhble E.3 also shows cumulati\ie mlues. These ;dlow us lo compute the total number of hours needed
to complete a specified number of uniL~. Aga.in. the computation is straightforwal'd. Just multiply the
table value Limes the time required for the first unit. Example E3 i llustrates this concept.

Example E3 Example E2 computed the time 10 complete the founh tugboat that Grem Lakes plans to buy. How long will
Using cumulative all four OO{l1s requfre'?
coefficients Looking this time ttl the "total time" column in Table E.3. we find that the cumultttive c«-ffident is
3.345. Thus, the time required is
TN=T1C
1~ = (1'25.000)(3.34S) = 41 8.125 hours in total for all 4 boat,;
For an illustrnlion of how E.xccl Or-.•1 <.-an be used to .solve Examples E2 nnd E3. see J>rogram E.. I al the end
of this module.

Using Table E.3 requires that we know how long it t;lkes to complete the firsl unit Yet. whal hap-
pens if our most recenl or most relktblc infonnation available pertains lo some other unit? T he
answer is that we must use these data to rind a revised estimate for the firsl unit and then apply the
table to that number. Example E4 iUu.strntes this concept.

Example E4 Great Lakes, Inc., believes that unusual circumsrnnoes in producing the first boa1 (see Example E.2) imply
Revising Jeorning -curve thttl the time estimate of 125.000 hours is not as valid n b1L,;e a.,; the time required to produce the third boat.
esHmotes Boat number 3 was completed in 100.000 hours.
To soh,e for the revised estimate for boo.I number I, we return to Table EJ, with a uni1 value of N = 3
and a leaming-cun,-e coefficienl of C = .773 in the 85% column. To find the revised eslimnte. we divide the
actual lime for boat number 3. 100.000 hours. by C= .773

JOO.OOO
- - - • 1->9. 366 houn.
.773
So, 129.366 bouts is 1be new (revised) estimaie for OO{l1 I.

STRATEGIC IMPLICATIONS OF LEARNING CURVES

So far. we have shown how operations managers can forec-.ast labor-hour requirements for a product.
We have ;dso shown how purchasing agenlS can determine a supplier's cost, knowledge that can help
in price negotiations. Another important application of learning curves concerns strategic planning.
An example of a company cost line and industry price line ;tre so labeled in Figure E.2. T hese
learning curves arc straight bccau.se both scales are log scales. When the rale of change is constant, a
log-log graph yields a straight line. If ;m organization believes its cost line to be the ...company cost"
l ine, and the i ndustry price is i ndic-.ated by the dashed horizontal line, then the company must have
cosLc;; at ll1e poims below the doned line (for example. point a or b) or else operate at a loss (poim c).
 SUMMARY 777
FIGURE E.2 ■

Industry l earning Curve

for Price Compared w ith
Company l earning
Curve for Cost
Note: Both the vertical and
norJzontoJ oxes of thiS figure
ore log scoles. This is known
OS O 10{}·/0g gtCJph.
Loss

Gross pt'Ofit
margin
r---
~

Accumulated volume (log scale}

Applications of the Lower costs are not automatic: they must be managed down. When a firm ·s strategy is to pursue
learning cuive: a cun·e steeper lhan the industry average (the company cost line in Figure E.2). it does lhis by
1. Internal ~ determine
labor standards and 1. Following an aggressive pricing policy.
rates o f material supply 2. Focusing on continuing cost reduction and productivity improvement.
required. 3. Building on shared experience.
2. External - determine 4. Keeping capacity growing ahead of demand.
purchase costs.
3. Strategic -• determine Costs may drop as a finn pursues the learning curve, bot \'Olume must increase for the learning curve
volume-cost changes. to exist. Moreo\fer. managers must undersland competitors before embarking on a lea.ming-curve
strategy. Weak competitors aJ'e undel'capitalized, stuck with high costs. or do not understand the
logic of learning curves. However. Slrong and dangerous competitors control theil' costs. have solid
financiaJ positions for lhe large investments needed, and have a track record of using an aggressive
learning-curve strategy. Taking on such a competitor in a price war may help only the consumer.

LIMITATIONS OF LEARNING CURVES

Before using learning curves, some c-.autions are in order:
• Because learning curves differ from company to company. as well as industry to industry.
estimates for each organization should be developed rather than applying someoneelse·s.
• Learning c.ul'ves are often based on the time necessary to complete the early units: thel'efol'e,
those times must be accurate. As current infonnation becomes available., reevaluation is
appropl'iate.
• A ny changes in personnel, design, or procedure can be ex.peeled to alter the learning curve,
causing the cur"e to spike up for a short lime. even if it is going to drop in lhe long run.
• While workers and pr0<.-ess may improve, lhe same learning curves do nol always ;tpply lo
indirect labor and material.
• The culture of the workplnce, as well as resource availability and changes in lhe process. may
alter the learning curve. For instance., as a project nears its end. worker interest and e.ffon
may drop, cuna.iling progress down the c.ul've.

SUMMARY The learning curve is a powerful tool for the openllions manager. This tool can ;ts.sist operations
m;magers in dctennining future cosl st:md;trds for ilcms produced as well as purch;tscd. In addition.
lhe learning curve can provide understanding about company and induslry perform;tnce. We saw
lhrcc approaches to leRrning curves: ari1hmelic analysis. log:trithmic analysis. and learning-curve
coefficients found in tables. Software can also help anaJyze learning curves.
778 MODU L E E, L EA R N I NG C URVES

KEY J ERM Leaming c urves /p. n 2)

USING SOFTWARE FOR LEARNING CURVES

E.xcel. Excel OM. and POM for Windows may all be used in ana.lyzing learning curves. You ~ use the ideas
in the following section on Excel OM to build your own Excel s preacl.,heet if you wish.

~ U5ing Excel OM
Program E. I shows how Excel Q{\•1 develops a spre acl.,hcet for leaming-c-urve calc.ulations. The input dal.a
come from Examples Eland E3. ln cell 8 7. weemer the unit number for 1he base unic (which does not have to
be I) . and in 08 we enter the time for this unit.

, ................----. These are used for computations. Do not touch these

1
----=-----~ae<._-J
!)Ea iOt ~ t-t F-wmat I.ooll Qau. ~

Great Lakes Service

cell$,tnce11 8 11, the tim e for thefir$tUnit ,$COmpvted,
allowing us to use initial units other than unit 1. In ceU
812, the power to be raised to is computed, maki.ng
the form ulM in the rest of c:ofumn 8 much $1mp'8r.
2
3 L.amngCurv.s
•5
Oetemining bmes
~""'=-::•.,,..:::.::.,~;;.,.:.:;,~.~.:1:...,:"1,,_,c,:,.~g,::n::.::;..,:.:-,
r.....,e=:~... l"1""'
;;:-:;-.
;;::::-,
...

AClt nttd '° be 1 Enter the nme Chis base unt took and enter
6 Data le,r111111 euve coeffit:tent, Do ~ ltle number$ in h
red Shaded na
1 lh1. number Of bew t.nl 1
8 lime tor base unit 125000
9 i,:Lea..,,..
=:::·•...,•::U,a
lW,:.:C::0::;
eli=• i:ent
:;"-''--0= 85::,
10
11 Tlme for l'ltSI: U'llt
12 b
13
14 Results
"'"" ....
1'®) ---------.
15 1.hl
16 l.ki!. 1
Tlmt
125000
time
~

125000
,::
'"""
«IXI)
17 I.hi 2
18 t.,il 3
108250 231250
96614 35 327884 35 "'°' 0 ,__ _+---~----<
90312. 418176.85 2 3

).11.L.____..:::--,---,--,-..----'L_--,_...:::=========~-J
'""
- SUM(816:B16)
PROGRAM E.1 ■ Excel OM's Learning-Curve Module, Using Data from Examples
E2 and E3

1llffl1
~ U5ing POM for Window•

POM for Windows· Leaming C urve module computes lhe length of time lha l future uniL'> will take. gi"en lhe
time required for 1he base unit and lhe learning rate (expressed as a number between O Md I), As an option, if
the tin}es required for the firs1 and Nth units are already known, the learning ra1t can be computed. See
Appendix IV for further de tails.

Salved Problem E.1

Oigicomp produces a new 1elephone system with built-in TV (b) How long will the first JI sys1ems take in total?
screens. Its lea.ming rate is 80%.
(c) As a pure-hosing agent. you expect to buy units 12 through 15
(a) lf lhe first one took 56 ho'Jrs, how long w ill ii take Digicomp of the new phone system. What would be your expected cost
10 make lhe eleventh S)'Stem? for lhe uni1s if Oigicomp charges $30 for each labor-hour?
 D 1scuss 10N Q ut:STI ONS 779

SOLUTION I from Table E ..3-80% unil time

(a) T~=r,c I
T 11 = (56 hours)(.462) = 25.9 hours

(b) Total time ror 1he tirs.t I J units = (56 hours.)(6.777) = 379.5 hours.

from Table E.l-80% 101al lime _J

(c) lb find the time for uniL<i 12 lhmugh IS. we lake the total cumulative l ime for units I to IS and s ubtract
1be total time for units l to I J, which w.-is computed in l){lrt (b). Total lime for the firSl 15 unjis = (56
hours)(S.5 l I) = 476.6 hours. So, 1he time for uniIs l2 through 15 is 476.6 - 379.5 = 97. 1 hours. (fhi s
figure could also be confirmed by computing the time.,; for units 12. 13. 14. and 15 separately using the
unit•time column M d 1ben adding them.) Expec1ed ~ • ror units 12 th.rough 15 = (97 .I hours.) ($30 per
hour) = $2,9 I3.

Solved Problem E.2

1( the first time you performed a job took 60 minutes. bow long will
the eighth job 1ake if you a.re on an 80% lea.ming curve?

SOUITION
Three doublings from I lo 2 to 4 to 8 implies .83. There fore . we ha\'e

60x (.8) 3 = 60 x .5 12 = 30.72 minuIes

or. using Table E.3. we have C = .512. The refore:

60 x .S 12 = 30.72 minutes

INTERNET AND STUDENT CD-ROM EXERCISES

Visit our Compa11io11 Web site or use your st11de111 CD-ROM to help with material i11 this 11wd11/e.

I:Z I On Our Companion Web site, www.prenhaU.com/heizer 0 On Your Student C D-ROM

• Self.Study Quizzes • PowerPoint Leclun:-
• Practice Problems • Prac1ice Problems
• l nteme t Exe rcise..,; • Active Mode E.xe rcises
• Jmernet Homework Problems • Excel OM
• E.xcel OM E.xample Daw Files
• POr-.•1 for Windows

@ DISCUSSION QUESTIONS
I. What are some of 1he limitations 10 the use or le.-imfog CUC\'es? S. W hy i.sn'l the lenrning c urve conce pt a<i applica ble in a high
4 4

2. Ide nt ify three applications of the learning curve. volume assembly line .-is it is in most O(her hum.-in actiYities?
3. What are the .-ipproa.ches 10 solving lea.mingmCurve problems? 6. W hal are lhe elemenL<i lhat can disrup4 the learning curve?
4. Refer 10 Example E2: Wllm are the imp1icaIfons for Great Lakes, 7. Explain lhe concept or lhe "doubling" effecl in learning <."t.u ves.
Inc-. . ir the engineering department wnnls to change the engine in 8. What technjques can a fil'm use to move 10 a steeper le.a.ming
Ibe third and subsequem tugboolS thaI1he fiml purchases? curve?
780 MODULE E, LEARNING CURVES

• ACTIVE MODEL EXERCISE

This Aclive Model, found on your CD-ROM, .-illows you to evalu.-i1e impommL elemems in 1be learnfog curve
modd described in Examples E2 and E3. You may change any inpul parameter in a green colored cell.

b}ff,t (di. i-- ir- , . _

l 111&<1 .,
i - . . .,_, tjllp

~ <t.!lil P -tt a
-.,,,..
0 "' " • Cb. ,:, '""' • Ill .I
Unit Curve, Cumula.tlve Curve and Costs ,
RttN o," I ow,1,on, I LHmlng Curve
,,. ,, - -
'. ..
r.., iofhunll I 4 , .. 5
u.nn_n CUM! co•lk:•
eos. ..,,l'lollf
I 085 .,J J.!l i 1l0
5
•
~ 100
•
b
., .)
S•ltchdUl'III fllll'nbtrl • .!IJ .!I i ' r
Unllli- 911,313
!
.. :~
60 '' I
UnllCNI
ai,,mt..irwn-
0.l!Wl.alw CHI
31:12

,-,
A1S..1n
lO
''
'
ACTIVE MODEL E.2 ■
0
9 17 ,, ,. ,. •'
Great l akes, Inc.
l ea rning Curve Analysis
' 5
"
UfltNumbt1

of Boats, Using Exa mple I Unit lifflll - Cunu.lM time I

E2 a nd E3 Data

Questions
1. If the learning is not as good as expected and rises to 90%, how much will the 4th boat COS!?
2. What should the learning coefficient be 10 keep the total cost of the first 4 boots below $16,000,000?
3. How many boals need lo be produced befot'C' the cost of an individual boat i.s below $4.000.000?
4. How mMy boa.cs need to be produced before the Merage cost of each boot is below $4,000,000?

PROBLEMS*
• P. E.I Amand Heinl. an lRS auditor. took 45 minute.,; 10 process her first lax return. T he lRS uses an 85% karning

"" •)
curve. How long will the
second return take?
b) fourth return take?
c) e ighth return take?

• P. E.2 Selon Hall Tl'ucking Co.just hired Sa.II)• Kissel 10 \'erify daily invoices and a«ouncs payable. She took 9 hours
"" and 23 minutes 10 complete her rn.sk on the lirst day. Prior employees in 1his j ob have 1ended to follow a 90%
learning cun.'C. How long will lhe task take at the end o(
a) the second day?
b) the fourth day?
c) the eighth day?
d) the sixteenth day?

• P. E.3 If Professor Tacy Quinn takes JS minute.,; to grnde the first exam. and fol low.s an 80% learning curve. how long
"" •)
will it take her
to grade the 25ch ex.-im?
b) to grade the first 10 exam.,;?

•Note P me.ins the problem m.,y be soh-cd with POM for Wiodows; '-" means the problem may be solved with Eiu.:el
OM: :md ~ mc:ins 1he problem may be soh'Cd wi1h PO.M for Window~ a1)dfot Excd OM.
 PROBLEMS 781

• P. E.4 If it took S63 minutes to com plete a hospital's first cornea transplant. and the hospital uses a 90% learning rate.
.a;
whm is 1he cumulath·e tin-.e 10 complete
a) the first 3 trans plants?
b) the first 6 trans plants?
c) the first 8 tra.nsp1ants?
d) the first 16 transplant-.?

:~ ES Beth Zion Hospital has rcce.ived initial certific.ation from the st.ate of California to become n center for liver
transplants. The hospirnl, however, muS1 comple1e i1s firs1 18 uanspla.ms under great scrutiny a.nd al no cost to
the pa.tien1s. The ,·er)' first transplant, jus1 comple1ed, •~quired 30 hours. On the basis of research at the hospi•
tal. Be th Zion estimates thnt it will have an 80% learning c-urvc. Estimate the time it will take to complete
a) the fiJ1h liver transplant.
b) .all of the first S trans plants.
c) the eighteenth trnn.,;plant.
d) au 18 transplatus,

:~ E.6 Refer to Problem E.5. Be1h Zion Hospital has just been informed that onJy the Ill-st IO transplants mus1 be per•
formed nt the hospital's expense. The cost per hour of surgery is estimated to be SS.000. Again. the learning
rme is 80% and the first surgery took 30 hours.
a) How long will the tenth surgery take?
b) How much will the tenth surgery cost?
c) How moch will all IO coS11he hospital?

• P. E.7 Mance,•ille Air has j us1 produced the lirst uni1 of a. large industrial comp1~ssor 1hat incorporated new 1echool•
.a;
ogy in the control circuit,; nod a new internal venting system. The first unit took I 12 hours of labor to manu 4

facture. The company knows from past experience thal thi.s labor content will decrease .significant!)' a.,; more
units are produced. In reviewing pa.s1 production darn, it appears that 1he company has experienced a 90%
learning curve when producing s imilar designs. The company is interested in e.,;timating the total time to com 4

plete the next 7 units. Your job as the production cost estimator i.s to prepnre the estimate.

• P. E.8 Candice Conon, a studem at San Diego State Universil)', bought si.x bookcases for her dorm room. Each
.a;
required unpacking of par1s and assembly, which included son-.e nailing a.nd bolling. Candice completed the
first bookcase in 5 hours and the SC(."lOod in 4 hours.
a) What is her learning rate?
b) Assuming the same rate continues. how long will the third bookcase rnke?
c) The fourth. fifth. nod s ixth c.a.-.es?
d) All six cases'?

Professor Mary Beth Ma.ITS 100k 6 hours to prepare the lirs1 leccure in a new course. Tradi1ionally. she has
:~ E.9
expcriencod n 90% lenming fnctor. How much time should it take her to prepare the fifteenth lcc--tutc?

• P. E.10 Tbe firS1 vending machine tha1 M. D'Allessandro, Inc.• assembled took 80 labor-hours. ES1imme how long the
.a;
fourth machine will require for ea<.~h of the following le-aming rate.,;:
a) 95%
b) 87%
c) 72%

• P. E.11 Kara-Smith Systems is inS1alling networks for Advant3,ge Lnsuronce. The firs1 inS1allation took 46 klbor•hours
.a;
to c-o mple te. Estimnte how long the fourth nod the eighth ins tallations will lake for ench of the following learn
4

ing rotes:
a) 92%
b) 84%
c) 77%

: ~ E.12 BaJtimore Assessmen1 Center screens and trafos employees for a. computer assembly firm in Towson,
Maryland. The progress of all trainees is tracked and those not showing the proper progress nre moved to l~s
demanding programs. By the tenth repetition trainees must be able to complete 1he assembly task in I hour or
less. Torri Olson-Alves has just spen15 hours on the fourth unit and 4 hours completing her eighth unit, while
another trainee. J u.lie Burgmeier. took 4 hour.. on the third and 3 hours on the sixth unit. Should you cnc-ourage
either or both of the trainees to continue? Wlly?

: ~ E.IJ Tbe beuer studencs m Bal1imore Assessment Center (see Problem E. 12) have an 80% lea.ming curve and can do
a task in 20 minutes after j ust six times. You would like to weed out the weak s tudenL<i sooner and decide to
evalume them after 1he third unit. How long should 1he third onit take?
782 MODU LE E, LEARN I NG CURVES

:~ E.14 Collette Siever. the purc.hasing agent for Northeast Airlines. is interested in determining what she can expect to
pay for airplane number 4 ir the third plane took 20,000 hours 10 produce, What would Siever expect to IXIY for
plnne number S'? Number 6? Use an 85% learning curve and a $40-per-hoor labor charge.

:~ E.15 Using the darn from Problem E.14, how long will ii take to complete the twelfth plane? 'The llfteemh plane?
How long will it take to complete planes 12 through 15 inclusive? At $40 per hour, whm can Oa\•is, as pur•
chasing agent. expect to pay for a ll 4 planes'?

:~ E.16 D)'namic RAM Corp. produces semiconductors and has a learning cuNe of .7. The price per bit is 100 miHi•
cents when the ,·olume is .7 x 10 12 bits. Whal is the e.xpe,cted price a1 I .4 x 1012 bits? What is the expected price
at89.6 x I0 1?bits'?

=~ E.17 Central Power owns 25 small power gene.roting plants. It has contracted with Genco Ser\•ices to overhaul the
power turbines or each of the plants. The number or hours that Genco billed Central to complete the third 1ur 4

bine was 460. Centro.I pays Genco $60 per hour for iL'> services. As the maintenan<."C manager for Central. )'OU
are trying to estimate the cost of overhauling the fourth tu.rt:,ine. How much woold you expect to pay for the
ovethaul or number 5 and number 6? All the tutbincs an- s imi lar and an 80% learning c-urvc is appropriate.

=~ E.18 h takes 28.7 18 hour.. to produce the eighth loc-o moti"e at a large Freoc-h manufacturing firm. If the learning
foctor is 80%. how long does it take to produce the tenth locomotive?

:~ E.19 Eric Kras..o;ow's firm is aboul to bid on n new radar system. Although the produc1 lLt;CS new technology. Krassow
believes thm a lea.ming rate of 75% is app<oprime. The lirst unit is expected to take 700 hours, and the contract
is for 40 units..
a) What i.s the total amounl of hours to build the 40 units'?
b) What is the avernge time to build each of the 40 units?
c) Assume that a worker \\'Olt.s 2.080 hours per year. How many wol'kers s..hould be assigned to this contract to
complete it in a )'ear?

:~ E.20 As the estimator for Pecer Ancona Enterprises, your job is to prepare an estimate for a potential customer service
contract. The contract is for the ser\'ice of diesel locomoti"e cylinder heads. 1ne shop has done some of these in
the pa.o;t on a sporadic bao;is. ibe time required to service each C)•linder head hlL,; been exoctly 4 hours. and similar
work has been accomplished a1 an 85% learning CUt'\'e. The customer wanes you to quote in batches of 12 and 20.
a) Prepare the quote.
b) After preparing the quote. yoo find a labor licket for this cus tomer for five locomotive cylinder heads. From the
notations on the labor ticket. you conclude that the fifth unit took 2.5 hours. What do you conclude about the
learning cun.'C and your quote?

=~ E.2 1 Sara B redbcnner and Blake DeYoung are teammates al a diS(.'OUnl score: their new job i.s assembling s wing sets
for e:us1omers, Assembly of a swing set has a learning rate of 90%, They forgot to time their effon on the fir~
swing set. bot spent 4 hours on lhe second set. They hove six more sets to do. Determine approximately how
much time will be (was) required for
a) the first unit.
b) the eighth unit.
c) all eight units.
E.22 Kelly-L:uubing. Inc .• a builder or govemmem-comracted small ships. has a steady work force of 10 very
skilled craftspeople. T hese workers can s upply 2.SOO labor hours each per )'Cat. Kelly Lambing is about to
4 4

undenake a new cotu.ract. building a new ~yle of boat. Tile llrst boa1 is expected to take 6,000 hours to com•
plete. Tile finn tbfoks that 90% is the expected lea.ming rate.
a) What i.s the firm's ..capacity'" to make these boats---that is. how many unit'> c-a n the firm make in I year?
b) If the opera.lions manager can increase the learning rate to 85% instead or 90%, how many units can the li.Jm make?

=~ E.23 Fargo Production has contrac.ted with John.o;on Services to overhaul the 25 robots nt its plant. All the robot,; are
similar and an 80% le-arning curve is appropriate. The number of hours thal Johnson billed Fargo to complete the
third robot o,·erhaul was 460. Fargo pays $60 per hour for its s.e,vices. Fargo wants to estimate the following:
a) How many hours will it lake to overhaul the 13th robot'?
b) The fi fteenth robot?
c) How long will it take to complete robots JO thl'ough 15 inclusive?
d) A.,; lhe person who manages the costs for overhauling all equipment whal is )'Our estimate of the cost of the
e ntire c-ontnK-1 for overhauling nll 25 robots'?

E.24 You are considering building a plane for training pilots. You believe there is a mrutet for 50 of these planes.
which will have a top speed of 400 kn and an e mpty weight of I0.000 lb. You will need one lest plane. Use the
NASA Web site (www.jsc.nasa.go-.'lbu2Jairframe.html) to detemiine the total cost Md engineering co~ of
buildfog all 50 planes.
 CASE STUDY 783
E.25 Using the accompanying log-log graph. answer the following questions:
a) Whal are the implications ror mana,gemem if it has rorecas1 its cost on the opcimum line?
b) What could be causing the fl uctuations nbo\·e the optimum line?
c) Ir management for«-a.-.t the tenth unit on the optimum line. what was that for«-a.-.t in hours'?
d) If mMagement buih the tenth unit as indicated by the actual Jfoe, how many hours did it take?

500
400
3))

---- '
OptimJm ~ -V--Actual
...__
-
"
• ----
' \_

~ "
- ~/
"' ------
10 1 10 .,. 3))@
100
T<Xal unils procllood

INTERNET HOMEWORK PROBLEMS

See our C-0mpanion Web sile at www.prcnhall.com/hcixcr for these additional homework problems: E.26
through E.33.

CASE STUDY
SMT's Negotiation with IBM tools. jigs, and fixtures~ 5% for quality control; and 9% for pur 0

chasing burden. Then. using an 85% lea.rnfog cur,•e, he backed up

SMT and one other. much larger company were asked by LBM to bid his costs to get an estimate for the first unit He next checked the
on 80 more units of a particular compu1er product The RFQ d,ua on hours and m~terials for the 25. 15, and 38 units already
(reque.o;t for quole) asked that the O\'ernll bid be broken down to made and found that his estimate for the firsl unit wa.-. within 4% of
show the hourly tate. the par1s and materials component in the price. actual cost. His check. howe\'er. had indic.aled a 90% lc-arning•
and any charges ror subcon1racted services, SMT quoted $1.62 mil• curve effect on hours per unit.
lion and supplied the cost breakdown a.-. requested. The .second com- In the negotiation.-.. SMT was represented by one of the two
pany submined onJy one to1al ligure. $5 mmion, with no cost break• owners of the business. lwo e ngineers. and one cost estimator. The
down. The decision was made to neg()(fale with SMT. sessions opened with a discussion of lea.ming cur,,es. The IBM cost
The IBM negotiating team included two purchasing managers estimator demonstrated that SMT had in fac-t been operating on a
and two cost engineers. One cos1 engineer had developed manufac- 90% learning cur.'C. But he argued. it should be possible lo move to
turing cosl e stimates for every component. working from e ngineer• a.n 85% curve. given the longer runs. reduced setup time, ~nd
ing drawings and cost-data books that he had built up from previ- increased continuity of workers on the job thal would be possible
ous experience and 1ha1 comafoed 1fo,e foctors, bo1h setup and nm with an order for 80 uniL-.. T he owner agreed with thi..s analysis and
times. for a large variety of operalion.s. He estimated materials was willing to reduce his price by 4%.
costs by working both from data .supplied by the lBM c.orporate However. as each operation in the manufac-t uring process wa.-.
purchasi ng staff and from purchasi ng journals. He visited SMT discussed. it became dear thnl some IBM cost e.o;timates were loo
fadlitie.s to see the tooling available so that he would know whal low because cenain crating and shipping expenses had been over•
processes were being used. He assumed that Lhere would be perrect looked. The..-.e over.sights were minor. howe\·er. and in the following
conditions and trained opera1ors, and he developed cost estimates discus.o;ions. the two parties arrived nt a t.-ommon under.standing of
for the 158th unil (previous orders were for 25. JS. and 38 units). specifications and reached agreements on the costs of ea.ch mruiufoc•
He added 5% for .scrap-and-flow loss: 2% for the lL<;e of temporary luring operation.
(co111i,111ed)
784 MODU LE E, LEARN I NG CURVES

AL tbjs point. SMT representatives expressed great concern undersrn.nd why SMT had qooted such a low figure. He w::inted to be
about the possibility of inflation in material costs.. The lBM negotia• sure thm Stv1T was using the correct manufacturing process. ln any
tors \'Oluntcered to include a form of price e.o;calation in the contract. case. if Sr-.•1T estimators had made a mistake. it lillould be noted. It
as previously agreed among themselves. IBM represe11rn.1h'es sug• was IBM's policy to seek ::i foir price both for itself and for its sup,,
gested tha1 if o,·erall material costs changed by more than 10%, the pliers. IBM procurement mMagers belie\·ed that if ::i vendor w::is los•
price could be adjusted accordingly. However. if one party took the ing money on a job. there would be a tendency to cut corners. In
initiath·e to have the pl'ice revised, the other could require an analysis addition. the IBM nego1ia1or fell thai by poin1fog out the error. he
or a/J parts Md materials i,woices in an-i..,ing at the new price. gene.rmed some goodwill that woold heJp in future sessions.
Another con<."'-Crn of the SMT representatives wns tho.I a large
amount of ove.njme and subcomracting would be required to meet
IBM's specified delivery scheduJe. IBM negotiators thought that a D i scussio n Q u est i ons
relaxation in the de-livery sc-hedule might be possible if a price COO· L What are the advantages and disadvantages to IBM and SMT
cession couJd be obtained. h1 response. the SMT team offered a 5% from this approach':'
discouni, and this was a.«epted, As a result of these negotiations, the 2. How does SMT's proposed le.aming rate compare with that of
SMT price was reduced almost '20% below its original bid price. other c.ompanies?
In a subsequent meeting c::iJJed to negotiate 1he prices of certain 3. What are the limitations of the learning cun·e in this case.'?
pipes 10 be used in the system. it becrune apparent to M IBM co~
estimator that SMT representatives had seriously undercstimoted Sot11'f'e: Adapted from E.. Raymond Corey. P1V<11rrmenr Ma11agrm<'m: Srru1egy.
their costs. He pointed out this ::ipp::irent error because he couJd not Org,ml:arl,m, t)11d lkcbltJY1 MulJ11J: (New Voit: Vsn No!arnnd Re.inhold).

• BIBLIOGRAPHY
Abernathy. W. J.• and K. Wayne. ·'Limits of the Leaming Curve." McDonald. A., ::ind L. Schrattenholzer. "Learning Cur\•es and
H(lnttrd Business Rc,•i tw 52 (Sepcember--Ootober 1974): 109-119. Technology Assessment."' /tuematilmal Joumal o/Teclmology
Bailey. C. D. nnd B. N. McIntyre. "Using Parameter Predic1ion M(m(lgemem 23 (2002): 718.
Models to Forec-ast Po.st-intenuption Leam ing." /IE Transactions Smi1b . J •• Uamh,g Cur,•e for Cos, Co,11r0/. Norcross, Georgi::i:
35 (December2003): 1077. l ndustrial Engineering nnd Manage-me nt Press. lnstilute of
Camm. J. ··A Note on Le.a.ming Cur\•e P:iramciers." Deci.vfrm Sciences Industrial Eng.ineers. ( 1998).
(.summer 1985): 325-327. S munl. T. L . and C. A. Walto; ... Improving Operations Planning with
Hall, G., and S. Howell, "The Experience C\IJ\'e from the Economist's Leaming Cur\•e..-;." Journal 1,f Operatilms Management 21
Pcrspecti\'e." Strategic Ma,utgemettt ltmrnal (July-September (January 2003): 93.
1985): 197-210. Weston. M. Leaming Cun1es. New York: Crown Publi.shing (2000}.
Lapre. Michael A .. Amit Sh::inkar Mukhe1jee. ::ind Luk N. Van Zangwill. W. I.. and P. 8. Kantor. "'Tbward a Theory of Continuous
Wassenhove . ..Behind lh e Leaming C urve: Linking Learning Improvement and the Learning Curve." Managemem Sciettte 44,
Activities to Wa.-.te-Reduction.'" Managemelll Scie11ce 46. no. S no. 7 (July 1998): 910-920.
(May 2000): 597-611.

INTERNET RESOURCES
Bailey, CharJes (Unh·ersi1y of Central Florida): Production technology. Tampa. Florida:
www.bus.ucr.edu/bailey www.protech-ie.com/soflware.hlm
NASA:
www.jsc.nasa.go,•Jbu2/1earn.html