r
it
I!!Ib
p
$
11
F
ii
it
4
i1}
it
I:
' r
'.
r,
II
It
i
i p
I;
i f
i v p tu i C~r~,. rtiIw. P t~~,: $ t.tr.tPb r:t.tit } R t'I tck~v it r I, I.. oI q I It ~ lU.??'t 1 4 h~ V
Oasis or Mirage: Does Customer Delight Really Yield Disproportionate
Gains in Customer Retention and Loyalty?
Douglas C. Friedman
University of Michigan
2
ABSTRACT
The idea that businesses need to delight their customers to win their loyalty rather
than merely satisfying them has become a popular topic recently in the business press.
However, with the exception of the motion picture and possibly fastfood restaurant
industries, customer delight does not produce disproportionately large gains in customer
retention. And except for the automobile, motion picture and tobacco industries,
customer delight does not lead to disproportionately large gains in customer loyalty.
Instead, in 30 of 35 industries examined, customer satisfaction yielded diminishing
marginal returns insofar as customer retention and loyalty were concerned.
I
r_ Zt^ vhA^ iSk^t&A.. MfMHf&nkJ3 dtlf as^^ I^w^ ua 'h]RIPk' tnIt* Mmf
3
INTRODUCTION
Practitioner journals and the business press have been exhorting companies in
recent years to delight their customers, not merely satisfy them, warning that failing to
delight customers will result in having customers defect to other firms (e.g., Chandler,
1988; Schlossberg, 1990; Schumann, 1994). Jones and Sasser (1995) argue that in
intensely competitive markets such as hard and soft durables, business equipment,
financial services, and retailing. "In markets like these, there is a tremendous difference
between the loyalty of merely satisfied and completely satisfied customers" (p. 89). They
argue, in effect, that the Law of Diminishing Marginal Returns does not apply to
customer satisfaction insofar as customer retention (which they call loyalty) is concerned.
Following Hirschman's exitvoice theory, we would expect the immediate
consequences of increased customer satisfaction to be a decrease in customer complaints
and an increase in customer loyalty (Fornell and Wernerfelt, 1987 and 1988). Satisfied
customers are more likely to repurchase a brand or company than dissatisfied customers
and to have a more favorable attitude towards the brand that goes beyond simply
repurchase intentions (Fornell, Ryan, and Westbrook, 1990; Yi, 1990 at 104105).
Satisfied customers are also more willing to tolerate price increases (Anderson, 1996).
Loyal customers help improve a company's financial performance since the cost to retain
a customer is generally lower than the cost to acquire a new one and in part by
contributing to brand equity (Aaker, 1992; Bhote, 1995). Loyalty thus serves as a proxy
for profitability (Reichheld and Sasser, 1990). The influence of customer satisfaction on
loyalty varies from industry to industry (Fornell, 1992). Loyalty has generally been
4
declining regardless of whether purchasers are heavy, medium or light users, but this does
not diminish its importance (East and Hammond, 1996).
If delighting customers really is the key to customer loyalty or customer retention,
then we should expect to see a disproportionate increase in customer loyalty or
repurchase likelihood once satisfaction crosses some threshold. Jones and Sasser (1995)
define delighted or "completely satisfied" (and thus loyal) customers as those answering a
5 on a question of satisfaction. Allowing for rounding, that should work out to
approximately a 90 or higher on a 100 point scale. The delight phenomenon should
display itself by having a positive interaction component in a piecewise linear regression.
This would indicate that customer loyalty was increasing at a much faster rate above the
delight point than below it. Alternatively, in the case of an industry whose satisfactionloyalty relationship were described by a cubic function, the cubic term would be positive
so that if customer satisfaction were above the inflection point, loyalty would increase at
an increasing rate. Delight could also have a disproportionate impact on loyalty if the
industry's equation were described by a quadratic function with a positive quadratic term,
which would yield a constantly increasing slope or by an exponential function with a
positive coefficient on the exponential term, which would likewise yield a constantly
increasing slope.
METHODOLOGY
To test the argument that delighted customers are disproportionately loyal, we
used data from the American Customer Satisfaction Index, which surveys annually over
44,000 consumers about their experiences with more than 200 firms with sales exceeding
$2.7 trillion (Fornell et al., 1996). Those firms compete in 40 industries in seven sectors.
1......., L I di dt. JUZ ia.ta ik.
F~iE"T    ""tal " MIMM3,11UNIMR19 I. rwztll Ilk, ENZIMUR a M, a; MMMMIM~ =M,"UMUIL UjwazAdt EL?
E4i.01
1~4 ~~PaWft8Pa~P~pr~rrt~~~l~s 2~~ ~uB~:~~i~4~~Cs1 ~~Yr tW~B TlreF~JI~(O~!lj87?~I~B~91 ClllqPS~'IWU S~iSP~t~BHPZ~tF~~e~l~P IIR1W ~ ~R!~ZY~V"~~~~WP~~ll ~iBS~P Ilsl~FT~['I'E.~f m~l~)~ ~ PY:sF1 F~~Jgl~g~~pYl'Sp ~ ~ tlff I qmt 7Ip.10M1:r,:r1, aP ~~ q Ir.";t~~~,,M W~~~~~~~~~~~~~~~~~~~~~~~., M, I _T ", I M. ~~~~~~~~rtII~~ _ Li LII_ _ — ~~ ~I ~C~ ^I~~~lL ~ ~ICs~~L~ILIL~~~~
5
For this study, we exclude the government sector and its five entrants because there has
been no suggestion that customer delight is relevant to government services. The six
studied sectors are manufacturing nondurables; manufacturing durables; transportation,
communications, and utilities; retail; finance and insurance; and services. The 35 specific
industries in this study are listed in Tables 1 and 2. Sample sizes ranged from 230 for the
U.S. Postal Service to 4,556 for the automobile industry.
To measure customer satisfaction and loyalty, we use latent variables. To
measure customer retention, we use the response to a single question that asks about
likelihood of repurchase. The customer satisfaction (LVSAT) latent variable is obtained
by adding weighted averages of answers to questions on overall satisfaction, expectancy
disconfirmation (performance that exceeds or falls short of expectations), and
performance versus the customer's ideal product or service in the category. The customer
loyalty latent variable is similarly obtained by adding weighted averages of answers to
questions on likelihood of repurchase, price tolerance (increase) given repurchase, and
price tolerance (decrease) to induce repurchase. Using the latent variables produces an
average 22 percent improvement in precision compared to use of responses from a single
question (National Quality Research Center, 1995; Ryan, Buzas, and Ramaswamy, 1995).
Since the business press articles on customer delight have focused on customer retention
(while often calling the phenomenon being studied customer loyalty), we examine the
relationship between customer delight and the singleitem measure of repurchase
likelihood (customer retention) as well as the relationship between customer delight and
brand loyalty. With a few notable exceptions, which are addressed in the Discussion
section below, the results were fairly comparable.
6
Repurchase likelihood and loyalty were regressed separately on customer
satisfaction using the equation that best fit the data for an individual industry. Cubic,
quadratic, linear and piecewise linear equations were considered. Logarithmic,
arctangent, and exponential transformations were considered as alternatives, but these
consistently yielded much lower R2 values than the cubic, quadratic and linear equations.
FINDINGS
Table 1 lists the type of equation that describes the connection between customer
satisfaction and repurchase likelihood for each of the industries examined and the amount
of variance explained by the equation (R2). Table 2 contains similar information for the
connection between customer satisfaction and loyalty.
The cubic equations have the form:
REPURCH = Po + PILVSAT + P2LVSAT2 + J3LVSAT3 or
LOYALTY = Po + PILVSAT + 02LVSAT2 + P3LVSAT3 or
where REPURCH equals the response on a 10point scale to a question asking
how likely the respondent is to repurchase the product from the company in question,
LOYALTY equals the value of the latent construct Customer Loyalty on a 100point
scale and LVSAT equals the value of the latent construct Customer Satisfaction, which is
reported as the company's ACSI score in the published ACSI. For negative cubic
equations, 3 has a negative coefficient, which means that the function will be concave
past the inflection point. A concave function signifies diminishing returns to scale, so a
one point gain in ACSI here will provide less of an increase in loyalty than will a gain of
one point in the area to the left of the inflection point. Some of the equations have a
negative Pi term, which gives them a local minimum in the data range, usually around: U'' IIE1,91 il; M P
I
 ~fqI,21 _"T 4W 3 I "M"r f P f IS2=2J 7 j
7
LVSAT = 4. Since companies are unlikely to be concerned about moving people from 0
satisfaction to only a 4 on a 100point scale, this dip has little practical significance. In
all cases except electric utilities, the function's maximum was outside the range of data;
for utilities, where loyalty was the dependent variable, the maximum came at LVSAT =
99.6. The positive cubic function for the tobacco and motion picture industries indicates
that there is an increasing return to scale past the inflection point.
The quadratic equations have the form:
REPURCH = po + 3ILVSAT + P2LVSAT2 or
LOYALTY = po + PILVSAT + P2LVSAT2
where the variables have the same meaning as above. In the.negative quadratic
equations, P2 is negative, leading to concave functions with constantly diminishing
returns throughout the data range. All of the functions had maxima outside the data
range. For the automobile industry, where loyalty is concerned, the positive quadratic
term means the equation is convex throughout the data range  loyalty increases at an
increasing rate throughout the data range. (The minimum occurs where LVSAT is less
than zero, which is not a meaningful number given the way the variable was constructed.)
For the linear piecewise regressions, the equations take the form:
REPURCH = po + PILVSAT + 32DLT + P3DLT*LVSAT or
LOYALTY = po + 3PLVSAT + J32DLT + p3DLT*LVSAT
where DLT is a dummy variable that takes a value of 1 if LVSAT is above a point
at which the slope changes sharply and 0 if it is below that point and DLT*LVSAT
represents the interaction between DLT and LVSAT. Below the critical point, DLT will
equal zero, so the p2 and P3 terms will drop out and the equation will have the form:
8
REPURCH = Po + PiLVSAT or
LOYALTY = Po + J3ILVSAT
Above the critical point, the equation will be:
REPURCH = (po + p2) + (pI + p3)LVSAT or
LOYALTY = (Po + 02) + (PI + P3)LVSAT
If P3 is negative, there will be diminishing returns to scale for satisfaction above
the transition point. If 03 is positive, then satisfaction above a certain level will result in
disproportionately large increases in repurchase likelihood or loyalty  the delight
phenomenon at work. Only the fastfood restaurant/pizza industry displayed this form,
though the coefficient for p3 narrowly missed statistically significance for both
repurchase likelihood and loyalty. If p3 equals zero, then the equation will be linear and
an increase in customer satisfaction will result in a constant increase in customer
retention or customer loyalty.
A logarithmic function, which fit the broadcast TV industry nearly as well as a
negative cubic function where repurchase likelihood was concerned, would have the
form:
REPURCH = po + PlIn(LVSAT) or
LOYALTY= Po + [lln(LVSAT)
and would exhibit constantly declining returns to scale. An arctangent function would
substitute arctan(LVSAT) for the natural log in the equations above and would likewise
exhibit constantly declining returns to scale. The arctangent functions were generally
~F~;,,~ —,r,,,i;l;r~,HIr i — `.i IMINE;l~; l~~
I
~~~~~~~~~~~~~~ II N W141~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~..~
s' , —,.L ~ ~ ~ ~.p 2
9
TABLE 1
REPURCHASE LIKELIHOOD
i
i
L
Industry
IR2 Function Tvne
11
I Inflection Point1
      V — gr 4 ~, 1  1 W,,,
Food processing
Beveragesbeer
Beveragessoft drinks
Tobaccocigarettes
Apparel
Athletic shoes
Publishingnewspapers
Personal care products
Gasservice stations
Personal computers/ printers
Household appliances
Consumer electronics  TV & VCR
Automobiles
Parcel delivery/ express mail
US Postal service
Airlinesscheduled
Telecommunicationslong distance
Telecommunicationslocal
BroadcastingTV
Utilitieselectric service
Department stores
Discount stores
Supermarkets
Restaurantsfastfoodpizzacarry out
Banks
Life Insurance
Personal Property Insurance
Hotels
Hospitals
Motion Pictures.306 Negative Quadratic.462 Negative Cubic.320 Piecewise Negative.135 Negative Quadratic3.435 Negative Cubic.622 Piecewise Negative4.157 Negative Cubic.328 Negative Cubic.236 Negative Cubic.358 Linear.477 Linear.501 Negative Cubic
j.391 Negative Cubic
3.341 Negative Cubic.211 Linear5.338 Negative Quadratic
1.623 Negative Cubic
i.516 i Negative Cubic.289 j Negative Cubic6.519 Negative Cubic
j.462 Negative Cubic
.436 I Negative Quadratic
1.331 Negative Quadratic.328 Linear7
j.557 I Negative Cubic
I.440 Piecewise Negative8
.474 Negative Cubic.438 1 Negative Cubic.370 Negative Quadratic.101 Positive Cubic
_ _ _. _ —::::........:~::.._: ~ ~..........
II
I
I
I
I
I
I
49
872
58
84
57
51
49
67
86
54
44
45
64
52
58
51
83
45
55
46
..~ ~ ~ ~ ~ ~ ~ ~~~~~~~,,,. h ,,. 9 f _ _
Notes;
1. for cubic functions, rounded to the nearest whole number; for the piecewise function it represents the point
where the slope changes.
2. Setting the cutoff point at 86 or 88 also yielded statistically significant values.
3. The coefficient on the cubic term was positive, but not statistically significant (p =.119), so there may be
some customer delight effect here.
4. The coefficient for the interaction  the change in slope  narrowly missed statistical significance (p =.057). If the standard is enforced strictly, then the function would be linear with an R2 of.619.
5. The coefficient for the quadratic term narrow missed statistical significance (p =.071). Relaxing the
standard yields a negative quadratic with an R2 of.223.
6. A logarithmic function explained the variation almost as well, reaching an R2 of.280.
7. The coefficient for the quadratic term narrow missed statistical significance (p =.096). Relaxing the
standard yields a positive quadratic with an R2 of.329. In a piecewise regression using LVSAT = 85 as the
cutoff, yielded an almostsignificant positive interaction coefficient (p =.08) with an R2 of.326. Fitting an
exponential function yielded an R2 of.295.
8. A negative cubic function with an inflection point at 43 fit the data equally well, with the coefficient on the
cubic term narrowly missing significance (p =.066).
T22.........
10
poor in their explanatory value and were tested only to rule out alternative possibilities.
An exponential function would take the form
REPURCH = poe P *LVSAT or
LOYALTY = poe P*LVSAT
and would exhibit monotonically increasing returns to scale if pi is greater than
zero. An exponential function fit the fastfood restaurant industry data almost as well as a
linear function.
The amount of variance in customer retention explained by customer satisfaction
varies from. 101 for the motionpicture industry to.623 for the longdistance
telecommunications industry. One likely reason for the low explanatory value in the
motion picture industry is that people in most cases choose movies by the genre or the
stars in the movie, not generally by the studio that produces them. In recognition of this,
the ACSI in the future will record loyalty only at the industry level, not at the company
level. Other industries where the R2 were particularly low have their own idiosyncrasies
leading to a limited link between satisfaction and loyalty. For example, with local
newspapers and the U.S. Postal Service enjoying virtual monopolies, available
alternatives are limited; "loyalty" may simply reflect lack of choice. Drivers likely
choose service stations on the basis of factors other than satisfaction, such as
convenience, location, price, even habit. Habit, in the literal sense, may account for
cigarette smokers brand purchase decisions and thus for the low R2 and generally high
levels of loyalty observed for the tobacco industry.
"U' '' ' ' ": —I I "' i ' ~.............
I,I I9 m, aII,t iI
I
Iw —w p ' f I Q # Lf lWJ I ~ T 
11
TABLE 2
CUSTOMER LOYALTY
i
i
i
I
i
I
I
Industry
II
i
4
Function Type
I Inflection Point'.........
Food processing
Beveragesbeer
Beveragessoft drinks
Tobaccocigarettes
Apparel
Athletic shoes
Publishingnewspapers
Personal care products
Gasservice stations
Personal computers/ printers
Household appliances
Consumer electronics  TV & VCR
Automobiles
Parcel delivery/ express mail
US Postal service
Airlinesscheduled
Telecommunicationslong distance
Telecommunicationslocal
BroadcastingTV
Utilitieselectric service
Department stores
Discount stores
Supermarkets
Restaurantsfastfoodpizzacarry out
Banks
Life Insurance
Personal Property Insurance
Hotels
Hospitals
Motion Pictures
4
i
I
I
L
I.296.429.298.166.423.582.168.300.228.316.431.457.358.278.229.327.597.463.289.481.439.406.331.322.499.451.403.354.100
I
1
L
Negative Quadratic
Negative Cubic2
Piecewise Negative
Positive Cubic
Negative Cubic
Negative Quadratic
Negative Cubic
Negative Cubic
i Negative Quadratic
Negative Cubic
Linear
Negative Cubic
Positive Quadratic
Negative Cubic
Negative Quadratic
Negative Quadratic
Negative Quadratic2
Negative Cubic
Negative Cubic
Negative Cubic
Negative Cubic
Negative Quadratic
Negative Quadratic
Linear
Negative Cubic
Negative Cubic
Negative Cubic
Negative Cubic
Negative Quadratic2
Positive Cubic
I
i
1
I
tI
t
i
i
I
I!
i
i
I
I
I
t
i
I
I
I
J
I
49
87
77
57
57
52
61
63
55
46
64
51
54
51
51
41
53
46

,.

Notes:
1. for cubic functions, rounded to the nearest whole number; for the piecewise function it represents the
point where the slope changes
2. The cubic term has a p value of.05 1, just above the.05 level. If this standard is adhered to strictly, the
function becomes negative quadratic and the R2 falls to.425. The longdistance telecommunications (p =.063) and hospitals (p =.077) cubic coefficients also narrowly missed significance. If the broader p <.10
level were used, the functions on both would be negative cubic with inflection points of 38 for longdistance
and 43 for hospitals.
12
Of the four industries whose functions are recorded as linear above, the household
appliance and personal computer industries both had nonsignificant (p >.2) negative
piecewise components with the lowest p values for transition points at 82 for household
appliances and 84 for personal computers. The Postal Service might be described by a
negative quadratic, since that term narrowly missed statistical significance (p =.071).
The fastfood/pizza industry had a nonsignificant (p=.096) positive piecewise component
when satisfaction was greater than 85 and was also fit moderately well by an exponential
function, so that industry may have increasing returns to scale at least at some point.
The R2 levels ranged from.100 for the motion picture industry to.597 for the
longdistance telecommunications industry. The explanatory value was somewhat less
for loyalty than for customer retention, suggesting that customer satisfaction is a better
indicator of repurchase likelihood than of willingness to be influenced by potential price
changes, the other items in the latent loyalty measure.
Most of the equations were of the same form (e.g. negative cubic) for both
repurchase likelihood and customer loyalty, not surprising since repurchase likelihood is
an important component of loyalty. The tobacco industry appears to undergo a dramatic
change, but there was a nonsignificant (p =.117) positive cubic term for the retention
function, so the equations may be the same or it may be that the delighted customers are
willing to absorb large price increases even though customer retention is subject to
diminishing marginal returns. The reverse may be the case with athletic shoes, where
customers become constantly more likely to repurchase until they are quite satisfied (85),.
but are more influenced by price changes. The auto industry shows increasing returns to
scale for repurchase until satisfaction reaches 86; the combination of the functions for this? ~ ~ ~ ~ ~ ~ ,ei
i
i
I
4w.~,r~j~e~: IE~~ptrrlrws ~ c~~T;Ft,g9nV1~ Th Rf6At~*WU~l~WWWrl N!sf~IBB~iflel~J 18~fn~ FPr~~~sl~~~ 1~f~ ~? i~~!fl~y Ii~~?~ A~~~f;~:
13
industry may show some tolerance for price rises among satisfied customers, which may
partly explain why satisfaction with automobiles has been rising by many measures over
the past several years even as sticker prices have continued to climb. It is not entirely
clear that the functions for the life insurance industry and the Postal Service were
different, since mild relaxation of the p =.05 standard for statistical significance would
result in those having the same functions for retention and loyalty. If life insurance does
have different forms of equations, the picture is similar to that for athletic shoes, with
satisfaction leading to constant gains in willingness to repurchase (up to LVSAT = 83),
but not to tolerate price increases.
DISCUSSION
The phenomenon of customer delight can be said to exist with certainty in only
one industry: the motionpicture industry. That is the only industry showing increasing
returns to scale for both customer retention and customer loyalty. (The second derivative
is positive and increasing above the inflection point, demonstrating a slope that is
increasing at an increasing rate.) However, since customer satisfaction is a poor indicator
of either repurchase or loyalty, delight doesn't seem to mean that much.
Three other industries show at least some signs of having customer delight
operating: fastfood/pizza, tobacco, and automobiles. This presence of increasing returns
to scale likely accounts for the great emphasis both those industries lay on customer
satisfaction. The automobile industry's positive quadratic function means that no matter
how much it increases satisfaction, loyalty will increase at a faster rate; retention
increases at an increasing rate until customers become very satisfied and then retention
gains slow. The tobacco industry's positive cubic function indicates that the greatest
14
increases in loyalty per unit increase in satisfaction come when LVSAT is greater than 77.
Companies in that industry will thus experience their greatest gains in loyalty by
increasing the satisfaction of those who are already highly satisfied. The effect of
satisfaction on retention is somewhat more equivocal; there may be a delight
phenomenon at work or tobacco may face diminishing marginal gains in retention. The
tobacco and automobile industries' satisfactionloyalty functions are shown in Figure 1.
Figure 1
90
j 50.......  Tobacco
J?40 4..............Auto Industry
30
20
10
C  CvJ @to c1 0 V)
Custom er Satisfaction
The fastfood restaurantpizza industry shows signs of providing customer delight,
though not at a level that is considered statistically significant. For both customer
retention and loyalty, it is marked by a linear satisfaction function. However, a piecewise
regression shows a positive coefficient P3, though the term is not statistically significant
(p =.096 for retention and.118 for loyalty). A positive P3 term indicates that retention
and loyalty are rising more quickly above the point where LVSAT equals 85 than below
it. However, since the term is not statistically significant, we cannot conclude that this
 7. iia k Mktfkim aM
I
I
i
i lKwil lip,me r 1mm i t W Y m F w s4 n
15
industry experiences the benefits of delighting its customers. Still, the linear function
suggests at least constant returns to scale in increasing satisfaction. Since an exponential
function fits the data reasonably well, this provides some more evidence for delight
leading to disproportionate gains in retention. A graph of the predicted retention levels
using the linear, piecewise linear, and exponential functions appears in Figure 2.
Figure 2
FastFood Restaurants and Pizza
'.33[ ' ' I.
WL~ ri5 j1 E ~.:.: .......  Piecew ise!;5?: [~ i 3,::~,l E x p o n eta
0.
~~~~~~~~. *~~~~~~~~~~~~~~~~~~~~~~~~''': ~:;~ z '] {f $  ~
t~~~~~~~~~~~~~~~~t4
2  n.; F
o 3 ",qtlvvl* 3B Erl:trr
66~
x U~~~~~~~~~~~^.Bri"tnear...... IV:S:.
t ~ ~ ~ ~ ~ ~ ~ ~ ~ Ip~~~~~~~~~iecewise
ar, ~~  ~~~~~~~~~~~~~~~~~~Exponlential
Li 4 — z SEf
3 
 C C CO t u) CD oe S a 0 sf0 0) 0
Customer Satisfaction
The household appliance industry also has a linear function with a nonsignificant
piecewise term, though in this case the sign of the term is negative, though the term is
significant only at a point slightly above the.15 level (p =.153). Thus, this industry
experiences either a linear link between satisfaction and loyalty, or possibly a link that is
linear until satisfaction reaches 82 out of 100 and then the gains from additional
satisfaction decline slightly.
16
That is the picture faced by the soft drink industry, which has constant rates of
increase in retention and loyalty until satisfaction hits 87, after which the rates taper off.
The personal computer and printer industry shows constant increases in retention
from climbing satisfaction. If there is any change in slope, it is negative after satisfaction
reaches 84, but the term is not significant (p =.244). This improvement in retention does
not translate into a similar picture for loyalty, so price sensitivity in this industry is
presumably relatively strong.
For all the other industries studied, either customer satisfaction shows constantly
diminishing returns (negative quadratic) or it shows increasing returns up until a point
and then diminishing returns (negative cubic). All of the points at which diminishing
returns kick in on the cubic equations occur at relatively moderate levels of satisfaction,
so while companies may experience notable benefits from turning their moderately
dissatisfied customers into moderately satisfied customers, they will reap fewer benefits
from turning their moderately satisfied customers into very satisfied or even delighted
customers. The broadcasting industry provides an extreme case of this. While it shows
an unusual dip with an increase in satisfaction at low levels, up to about 28, the more
interesting part of the graph is over at the right, as shown in Figure 2. After satisfaction
climbs above 64, the industry starts to exhibit declining marginal gains in loyalty from
additional satisfaction until the gain where customer satisfaction reaches 100 is
practically zero  because loyalty is also virtually 100 percent.
The logical conclusion to be drawn from these results is that in most industries,
what businesses need to do is not delight customers, but rather avoid dissatisfying them.
In most industries, companies will do better at improving customer retention and loyalty
i . ~~~~~~~~~~~~~~~~~~~~~~~a~l.~~i~tdltl~~
I
I
, q  IRP WPM
  .. Wml, 711 " M. =a PIMINFRI!TT!T!!   i"
17
by turning moderately dissatisfied customers (LVSAT in the 40s or 50s) into moderately
satisfied customers (LVSAT in the 60s or 70s) than by turning moderately satisfied
customers into delighted customers (LVSAT in the 80s or 90s).
Figure 3
Broadcast TV
100
90
80
70 
60
50
0 40
30
20
10
0 IH'ill inhiiin;: it Hr 5ir;:r 1 t11111 ti iiiiir itit r:2n iu iii ii;t u iiutiiintt
N  '.CO C to, COO) 0 L r N CO) 0) LO
 C O C') C) a n It) CD c O a N s ) * a0
Customer Satisfaction
More typical are the graphs displayed by the beer industry, using either quadratic
or cubic, equations. (The cubic term was significant at the p=.051 level, so both the cubic
and quadratic functions are displayed.)
Figure 4
Beer Industry (Quadratic)
100
80
~60
3 40
20 
0
C us; LO to e  oai fto
r sXCustomer Satisfaction Ns )
Customer Satisfaction
18
Figure 5
Beer Industry (Cubic)
60
50
> 40
30
a 30
20
10
0 
r.C~O'J 0 1, r. CO) O T 0.r — N cO O) U' t"
to co co X A _ W
'*  ' C CO C ' U' ) (O D N N t C 0C m)
Customer Satisfaction
If customer delight so rarely produces the promised benefits, why did Jones and
Sasser (1995) argue that it is of major importance for firmns to delight their customers or
risk defection if they fail to do so?
One reason Jones and Sasser got the results they did is the fortuitous choice of the
automobile industry. In their article, they provide a graph of Satisfaction vs. Loyalty for
the five industries they studied. The airline and hospital industries show graphs that
appear to be negative quadratics, the same as we found. Only the personal computer and
automobile industries actually appear to show a notable effect of customer delight on
loyalty. Jones and Sasser looked at the business computer market whereas the ACSI
looks at computers for both business and personal use. It may be that customer delight
provides benefits in the business user segment but not in the home or home office
segments. Delight may function in businesstobusiness markets but not consumer
markets (Saunders, Scherer and Brown, 1995). Finally, Jones and Sasser found customer
delight operating in the automobile industry, with the same sort of positive quadratic
I'i~R~s~s~"s~YU~~~3X~t~ ~. _i, ~ ~i'7, —
  * * I, I...a
MM 11 R I  l 
I
4
I
I
1I
A
M, "ikqq.. 1?14 1'"Fm'wNwffl^" I
19
graph that we did for loyalty. Our results were different from theirs for customer
retention. The difference may be due to the recency of new car purchase. The ACSI
includes people who purchased new cars at least six months but not more than three years
previously, whereas the data Jones and Sasser used included only those who had
purchased cars within the previous year. Jones and Sasser note that 90 days after
purchasing a car, 60 to 80 percent of customers say they intend to repurchase the same
brand whereas three to four years later only 35 to 40 percent actually do so. Perhaps
customer delight's effect on retention is ephemeral. The auto industry is at most an
exception to the rule of diminishing returns; it does not prove a rule of customer delight
as Jones and Sasser suggest.
The case of the local telephone industry is also curious. Using data from an
unspecified Bell operating company, Jones and Sasser found that, as they had expected,
customer delight did not influence repurchase likelihood since even the most dissatisfied
customers have had no choice but to deal with the local phone company. However, the
ACSI findings demonstrate that the local telecommunications industry follows the same
general pattern as other industries, and willingness to repurchase is indeed influenced by
satisfaction. The days of, in the words of comedienne Lily Tomlin's Emestine the
Operator, "We're the telephone company. We don't care. We don't have to" are fast
disappearing thanks to cellular phones and pending deregulation. Without more specific
information about the nature of the data they used, it is not possible to explain the
difference in the findings.
In much of the business press, the term customer loyalty is used when customer
retention is intended. However, as Ryan et al. (1995) point out, a composite measure
20
provides a better indicator of loyalty than a singleitem measure. The singleitem intent
to repurchase measure is useful only so long as prices of competing goods remain
constant. Including price tolerance measures show how secure customers' loyalty really
is.
CONCLUSION
The muchballyhooed idea that delighting customers is the only way to ensure
loyalty turns out to be unfounded for all but a very limited group of industries. However,
this analysis did not consider whether individual companies might experience loyalty
benefits from customer delight even if their industries do not. Owing to the confidential
nature of the data, it would not be possible to publish a list of what companies might
experience such delight. The data also were collected for major companies. It may be
that for small companies operating in niche markets, customer delight is indeed a factor
in determining loyalty. The same may be true for industries not covered by the ACSI,
such as wholesale industries or fullservice restaurants. In addition, this study does not
look at what is causing the satisfaction; perhaps continually introducing that surprise and
delight attributes (i.e., unexpected favorable features) as included in the Kano Model
would lead to delight that might produce greater loyalty, but the cost of such continual
innovation is likely to be high (Kano et al., 1984). Further, just as the surprise and delight
attributes eventually become expected, the same may happen with innovation  continual
innovation becomes expected, as in the computer industry, and may fail to provide the
same delight benefits as before.
An area that this study suggests should be investigated is that of optimal
satisfaction point. Given that satisfying consumers costs firms money and that those. c j eh^.tl+kikJl4 t 4k 414fi n
p7;u4;;7w.L; ili, J11Ji 11St, 4Bi~ ~,~~ri0~fi!h~iA ~ ~~
.1
I
11
q
i 
I
I
3
1a, 2
w Ia iMr^prWInr. MM w
21
costs tend to increase more rapidly at higher levels of satisfaction, while at the same time
the gains in loyalty from increased satisfaction are diminishing, it should be possible to
calculate an optimal point that companies should strive for. That is, delighting the
customer may not be the most profitable strategy for companies to pursue.
22
REFERENCES
Aaker, David A. (1992), "The Value of Brand Equity," Journal of Business Strategy, 13
(July), 2732.
Anderson, Eugene W. (1996), "Customer Satisfaction and Price Tolerance," Marketing
Letters, 7 (3), 265274.
Bhote, Keki R. (1995), "'It's Customer Loyalty, Stupid': Nurturing and Measuring What
Really Matters," National Productivity Review, 14 (3), 3959.
Bloemer, Jose M. M. and Hans D.P. Kasper (1995), "The Complex Relationship Between
Consumer Satisfaction and Brand Loyalty," Journal of Economic Psychology, 16
(July), 311329.
Chandler, Colby H. (1988), "Quality: Beyond Customer Satisfaction  To Customer
Delight, Quality Progress 21 (October), 2023
East, Robert and Kathy Hammond (1996), "The Erosion of RepeatPurchase Loyalty,"
Marketing Letters, 7 (2), 163171.
Fornell, Claes (1992), "A National Customer Satisfaction Barometer: The Swedish
Experience," Journal of Marketing, (January), 121.
Fomell, Claes, Michael D. Johnson, Eugene W. Anderson, Jaesung Cha, and Barbara
Everitt Bryant (1996), "The American Customer Satisfaction Index: Nature,
Purpose, and Findings," Journal ofMarketing, (October), 718.
Fornell, Claes and Birger Wernerfelt (1987), "Defensive Marketing Strategy by Customer
Complaint Management: A Theoretical Analysis," Journal of Marketing
Research, 24 (November), 337346.
Fomell, Claes and Birger Wemerfelt (1988), "A Model for Customer Complaint
Management," Marketing Science, 7 (Summer), 271286.
Fornell, Claes, Michael J. Ryan, and Robert A. Westbrook (1990), "Customer
Satisfaction: The Key to Customer Retention," MOBIUS, 9 (3), 1418.
Hirschman, Albert 0. (1970), Exit, Voice, and Loyalty  Responses to Decline in Firms,
Organizations and States. Cambridge, MA: Harvard University Press.
Jones, Thomas 0. and W. Earl Sasser Jr. (1995), "Why Satisfied Customers Defect,"
Harvard Business Review, 73 (November/December) 8899.
Kano, Noriaki, Nobuhiku Seraku, Fumio Takahashi, and Shinichi Tsuji (1984),
"Attractive Quality and MustBe Quality," presented at the 12th Annual Meeting
(1982) of the Japan Society of Quality Control, B1B21.
~ I.. I  I I I
I
r  II
fv.rm[,.vL, Pfmmt Fnnm r "..1 " mg, BMW, " 414 TS, 7I,  pw irm...., I. l..,
WTIM ",RN PmraM,  ff
23
Kasper, Hans (1988), "On Problem Perception, Dissatisfaction and Brand Loyalty,"
Journal of Economic Psychology, 9 (September), 387397.
National Quality Research Center (1995), American Customer Satisfaction Index
Methodology Report, University of Michigan Business School, Milwaukee, WI:
American Society for Quality Control.
Reichheld, Frederick F. and W. Earl Sasser (1990), "Zero Defections: Quality Comes to
Services," Harvard Business Review, 68 (September/October), 105111.
Ryan, Michael J., Thomas Buzas, and Venkatram Ramaswamy (1995), "Making CSM a
Power Tool," Marketing Research, 7 (Summer), 1116.
Saunders, Paula M., Robert F. Scherer, and Herbert E. Brown (1995), "Delighting
Customers by Managing Expectations for Service Quality: An Example from the
Optical Industry," Journal of Applied Business Research, 11 (Spring), 101109.
Schlossberg, Howard (1990), "Satisfying Customers Is a Minimum; You Really Have to
Delight Them," Marketing News, 24 (May 28), 1011.
Schumann, Paul A. (1994), Innovate Straight Path to Quality, Customer Delight, and
Competitive Advantage. New York: McGrawHill.
Yi, Youjae (1990), "A Critical Review of Customer Satisfaction," in Review of Marketing
1990, Valerie Zeithaml, ed. Chicago: American Marketing Association, 68123.