Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

IMA Journal of Management Mathematics (2004) 15, 111–124

Warranty servicing strategies to improve customer


satisfaction

N. JACK†
Division of Mathematical Sciences, School of Computing and Advanced Technologies,
University of Abertay Dundee, Dundee, UK

Downloaded from http://imaman.oxfordjournals.org/ at University of New South Wales on July 6, 2015


AND

D. N. P. M URTHY‡
Division of Mechanical Engineering, The University of Queensland, Brisbane Q 4072,
Australia and Norwegian University of Science and Technology, Trondheim, Norway

Customer satisfaction with a purchased product depends on its performance under warranty
and during the remainder of its useful life. Dissatisfaction with an item is important to a
manufacturer since it can lead to the loss of potential customers through the negative word-
of-mouth effect as well as existing customers switching to a competitor. In this paper, we
define satisfaction in terms of the likelihood of a customer not switching to a different
manufacturer when a new item needs to be purchased.
Manufacturers can use specific servicing strategies to reduce warranty costs and this
topic has already been addressed in the literature without considering the effect of
customer dissatisfaction. In this paper, we propose particular strategies that will increase
customer satisfaction and we discuss methods for obtaining the optimal parameters of these
strategies.

Keywords: warranty; corrective maintenance strategies; customer retention.

1. Introduction
Modern industrial societies are characterized by rapid changes in technology that have
resulted in new products appearing on the market at an increasing pace. On the positive
side, the performance capability of products has been increasing dramatically but, on the
negative side, products have been getting more complex. As part of the input to their
purchase decisions, customers expect assurance that a product will perform satisfactorily
over its useful life and manufacturers have responded to this by bundling products with
post-sale support. Warranties and extended warranties are elements of this post-sale
support. Manufacturers have also used warranties to promote and differentiate their product
from those of competitors, with better terms implying a more reliable product.
However, offering any warranty causes a manufacturer to incur additional warranty
servicing costs. These servicing costs, which can vary from 0·5 to 7·0% of sale price
(depending on the product and manufacturer), have a significant impact on the competitive
behaviour of manufacturers, who need to devise strategies to reduce this type of cost.
† Email: n.jack@abertay.ac.uk
‡ Email: murthy@mech.uq.edu.au

IMA Journal of Management Mathematics Vol. 15 No. 2 


c Institute of Mathematics and its Applications 2004; all rights reserved.
112 N . JACK AND D . N . P. MURTHY

Customer satisfaction with a purchased product depends on its performance over


the warranty period and during the remainder of its useful life. Failures occurring soon
after purchase or soon after the warranty expires both cause high dissatisfaction. Each
dissatisfied customer can impact on future sales and this has serious implications for
manufacturers. These include loss of potential new sales due to the negative word of
mouth effect and repeat purchase sales disappearing when customers decide to switch to
a competitor. General opinion is that the cost and effort required by a manufacturer to
recover a lost customer is roughly ten times the cost needed to attract a new customer, and
so customer retention is very important.
Warranty servicing strategies can be used by manufacturers to reduce item failures

Downloaded from http://imaman.oxfordjournals.org/ at University of New South Wales on July 6, 2015


and hence customer dissatisfaction. However, existing models from the literature all focus
on minimizing servicing costs to determine optimal servicing strategies and do not look
at customer dissatisfaction and its implications. Most of these models consider one-
dimensional time based warranties, and discuss both corrective maintenance (CM) and
preventive maintenance (PM) strategies for the manufacturer during the warranty period.
In a CM strategy, the rectification action at a failure can be either a replacement or a
repair and different types of repair can be modelled. Biedenweg (1981), Nguyen & Murthy
(1986, 1989), Nguyen (1984), Jack & Van der Duyn Schouten (2000), and Jack & Murthy
(2001) all split the warranty period into distinct intervals for repair and replacement.
However, in Murthy & Nguyen (1988), the decision to replace or repair the item depends
on whether the estimated repair cost exceeds a certain threshold or not. Instead of making
the usual assumption that an item under warranty experiences only two possible states
(operating or failed), Zuo et al. (2000) consider a multi-state item and discuss a servicing
strategy where the replace–repair decision depends on the degree of deterioration and the
residual length of the warranty period. Iskandar et al. (2002) consider items sold with
a two-dimensional warranty based on time and usage. In their CM strategy model, the
rectangular warranty region is split into sub-regions for repairs and replacements.
In a PM strategy, the PM actions are carried out either to reduce the likelihood of a
failure or to prolong the life of an item. Optimal PM schedules for the manufacturer during
the warranty period have been discussed by Chun (1992), Jack & Dagpunar (1994), and
Dagpunar & Jack (1994). These authors all assume imperfect PM actions and use age
reduction concepts in the modelling. Jack & Murthy (2002), however, introduce a different
imperfect PM modelling technique where the rate of occurrence of failures is reduced by a
given amount at each PM action.
In this paper we focus on customer dissatisfaction resulting from product failures for
products sold with a one-dimensional non-renewing free replacement warranty (FRW) of
period W . We consider two separate cases: in Case A, the useful life L of the item is
approximately the same as W (L ≈ W ) and in Case B, L is much larger than W (L  W ).
We discuss CM strategies for the product manufacturer that will reduce the number of
failures, customer dissatisfaction, and hence the likelihood of the customer switching to
another manufacturer for a replacement item at time L. These strategies are characterized
by two parameters, and we consider the optimal choice of these parameters in order to
achieve the tradeoff between the benefits gained from fewer failures and the resulting
servicing costs.
The outline of the paper is as follows. Model notation is established in Section 2 and
Section 3 contains the model formulations for Case A and Case B, and the corresponding
WARRANTY SERVICING STRATEGIES 113

servicing strategies. Section 4 describes the analyses of the two cases, where we show how
to derive the optimal strategy parameters. In Section 5 we give numerical examples and in
Section 6 we state our conclusions and suggest areas for further research.

2. Notation
The following notation is used in the models for both cases:
f (x), F(x), F̄(x) density function, distribution function, and survivor function
for time to first item failure;

Downloaded from http://imaman.oxfordjournals.org/ at University of New South Wales on July 6, 2015


r (x), R(x) hazard rate and cumulative hazard function;
Cr cost of a repair;
Cf penalty cost associated with losing a customer (resulting from
loss of future and/or potential sales);
δ age reduction factor at an overhaul;
C p (δ, x) cost of an overhaul at age x with age reduction (1 − δ)x;
N (t) number of item failures in the interval [0, t);
f (i) (x), F (i) (x), F̄ (i) (x) density function, distribution function, and survivor function
for time to ith item failure (i  1);
pk (a, b) probability of k failures occurring between age a and age b.
The following specific notation is needed for Case A:
φ(k) conditional probability of a repeat purchase given that
N (W ) = k;
ΦW (unconditional) probability of a repeat purchase;
qk (m, δ) probability distribution of N (W ) under the (m, δ) overhaul
strategy;
EC W manufacturer’s expected warranty servicing cost;
JW manufacturer’s expected total cost per item sold.
Finally, in Case B, the following extra notation is used:
ψ(k1 , k2 ) conditional probability of a repeat purchase given that
N (W ) = k1 and N (L) − N (W ) = k2 ;
ΨL (unconditional) probability of a repeat purchase;
sk1 ,k2 (m, δ) joint probability distribution of N (W ) and N (L) − N (W )
under the (m/W, δ) overhaul strategy;
JL manufacturer’s expected total cost per item sold.

3. Model formulation
An item is sold to a customer with a non-renewing free replacement warranty of period W .
This warranty requires the manufacturer to rectify any failure that occurs up to a time W
from the initial purchase, at no cost to the customer. We assume that the item is repaired
at each failure during [0, W ) and the duration of each repair is always small relative to the
time between failures and so can be ignored. When the hazard rate of the item’s time to
first failure is increasing, there is a greater chance of the item failing as it ages.
114 N . JACK AND D . N . P. MURTHY

[φ ( k )]

Downloaded from http://imaman.oxfordjournals.org/ at University of New South Wales on July 6, 2015


a
0
[k ] b
F IG . 1. Probability of retaining the customer [φ(k)] versus number of failures [k] during warranty.

We assume that customer satisfaction with the item depends on the failure history over
its useful life L, and we consider the two cases: (i) L ≈ W (Case A) and (ii) L  W
(Case B).

3.1 Case A
When L ≈ W , customer satisfaction depends on N (W ), and the customer will need
to purchase a replacement item shortly after the warranty expires. We assume, for
demonstration purposes, that


1 γ if 0  k  a,
φ(k) = b−k
if a < k < b, (1)


b−a
0 if b  k < ∞,
where γ  0.
This simple form for φ(k) is shown in Fig. 1. The customer will definitely purchase a
new item from the same manufacturer unless more than a failures occur under warranty.
φ(k) then decreases with k until the customer is lost for certain when at least b failures are
experienced. The parameter γ determines the rate of decrease in φ(k) and, of course, many
other more complex forms for the function φ(k) can be considered.
We consider two CM strategies for the manufacturer over the warranty period. In
the first simple strategy, all failures under warranty are rectified by minimal repair. This
maintenance action returns the failed item to an operational state without affecting the
hazard rate. The manufacturer’s expected total cost per item sold, JW , is then the sum of
the expected warranty servicing cost EC W and the expected penalty cost of losing the
customer.
The manufacturer can also use the alternative ‘(m, δ) overhaul’ strategy to control
N (W ). Here, the first m − 1 failures under warranty are rectified by minimal repair and the
item is then overhauled at the mth failure, provided this occurs before time W , otherwise no
overhaul takes place. If an overhaul does occur then all subsequent failures under warranty
are minimally repaired.
WARRANTY SERVICING STRATEGIES 115

If an overhaul occurs when the item has age x then its effect is to reduce the item’s
age to δx for some δ, where δmin (x)  δ  1 and δmin (x) is a non-decreasing function
of x. (1 − δmin (x))x is therefore the maximum age reduction available at the overhaul.
If δmin (x) = 0, then it is possible for the manufacturer to carry out a perfect overhaul at
age x that will produce a ‘good-as-new’ item, but if δmin (x) > 0 the overhaul is always
imperfect. As δ → 1, the quality of the overhaul decreases to that of a minimal repair and
as δ → δmin (x), more worn out components are replaced by new ones so as to rejuvenate
the item. The above ‘age reduction’ concept was first introduced by Kijima et al. (1988),
who used the term ‘virtual age’ for the item’s age after the maintenance action.
We assume that C p (δ, x) = Cr + Co (1 − δ)x, so the overhaul cost consists of the fixed

Downloaded from http://imaman.oxfordjournals.org/ at University of New South Wales on July 6, 2015


cost Cr equal to the cost of a minimal repair plus a term that increases linearly with the
amount of the age reduction carried out. Note that, when δ → 1, C p (δ, x) → Cr and the
overhaul is equivalent to just a minimal repair.
Under the ‘(m, δ) overhaul’ strategy, the manufacturer’s expected total cost per item
sold, JW (m, δ), is again the sum of expected warranty servicing costs EC W (m, δ) and
expected penalty costs. The decision variables are m and δ and the optimization problem
is to minimize JW (m, δ), subject to m  1, and δmin (x)  δ < 1.
In each of the above two CM strategies, the durations of all maintenance actions are
assumed to be small relative to the times between item failures and so can be ignored in
the modelling.

3.2 Case B
The customer keeps the product for its useful life L  W , and so is responsible for
rectifying any failure that occurs during the post-warranty period (W, L]. All such failures
are minimally repaired. Customer satisfaction now depends on N (W ) and N (L) − N (W ),
and we assume that
ψ(k1 , k2 ) = φ1 (k1 )φ2 (k2 ). (2)
φ1 (k1 ) and φ2 (k2 ) have forms similar to the function φ(k) given in (1), with k replaced
by k1 and k2 , a and b replaced by a1 and b1 and a2 and b2 , and γ replaced by γ1 and
γ2 , respectively. In general, a2 > a1 and b2 > b1 implying that failures during the post-
warranty period cause less dissatisfaction than failures under warranty.
As in Case A, we again consider two servicing strategies. We first assume that the
manufacturer uses only minimal repair during the warranty period. JL is then the sum
of expected warranty servicing costs EC W and the expected penalty cost of losing the
customer at time L.
The manufacturer has to try to control both N (W ) and N (L) − N (W ), so we also
consider the ‘(m/W, δ) overhaul’ strategy which is a modification of the second servicing
strategy from Case A. Now, the first m − 1 failures during (0, W ] are again rectified by
minimal repair, and the item is then overhauled either at the mth failure if this occurs before
time W , or at time W .
Under this modified CM strategy, the manufacturer always performs an overhaul and
this overhaul is assumed to have the same age reduction effect and cost as in Case A. The
decision variables are again m and δ and the optimization problem is to minimize JL (m, δ),
subject to m  1, and δmin (x)  δ < 1.
116 N . JACK AND D . N . P. MURTHY

4. Model analysis
If each item failure during the interval [0, t) is minimally repaired, then {N (t), t  0} is
a non-homogeneous Poisson process (NHPP) with intensity function r (t) and pk (a, b) =
[R(b) − R(a)]k e−[R(b)−R(a)] /k!.

4.1 Case A
Under the minimal repair servicing strategy, the probability that the customer will make a
repeat purchase from the same manufacturer is given by

Downloaded from http://imaman.oxfordjournals.org/ at University of New South Wales on July 6, 2015


∞ a b−1 

b−k
ΦW = φ(k) pk (0, W ) = pk (0, W ) + pk (0, W ), (3)
k=0 k=0 k=a+1
b−a

and the manufacturer’s expected warranty servicing cost and expected total cost are
EC W = Cr R(W ), (4)
and
JW = EC W + C f (1 − ΦW ), (5)
respectively.
Under the (m, δ) overhaul strategy, it follows that


pk (0, W ) if 0  k < m,
qk (m, δ) = W
(6)
 pk−m (δx, W − (1 − δ)x) f (m) (x) dx if k  m,
0

and the repeat purchase probability is given by



a b−1 

b−k
ΦW (m, δ) = qk (m, δ) + qk (m, δ). (7)
k=0 k=a+1
b−a

The probability that the manufacturer overhauls the item under warranty is

m−1
F (m) (W ) = 1 − pk (0, W ), (8)
k=0

and the expected number of failures that occur during (0, W ] is




ΛW (m, δ) = kqk (m, δ)
k=1

m−1 

= kqk (m, δ) + (m + k − m)qk (m, δ)
k=1 k=m

m−1
W
=m− (m − k) pk (0, W ) + [R(W − (1 − δ)x) − R(δx)] f (m) (x) dx.
k=0 0
(9)
WARRANTY SERVICING STRATEGIES 117
W
Note that, for fixed m, ∂ Λ∂δ
W
= 0 x[r (W − (1 − δ)x) − r (δx)] f (m) (x) dx  0
and so ΛW (m, δ) is increasing in δ. Hence, the expected number of failures under
warranty increases as the quality of the overhaul decreases. Note also that, for fixed δ,
ΛW (m, δ) → R(W ) as m → ∞. This implies that, as m increases, the likelihood of an
overhaul occurring decreases and the (m, δ) overhaul strategy eventually reduces to the
minimal repair servicing strategy.
To obtain the manufacturer’s expected warranty servicing cost we need to consider the
two cases N (W ) < m and N (W )  m. When N (W ) = k < m, all failures are minimally
repaired and the servicing cost is kCr . However, when N (W ) = k  m and the mth failure
occurs at age x resulting in an overhaul, the servicing cost is (k − 1)Cr + C p (δ, x). The

Downloaded from http://imaman.oxfordjournals.org/ at University of New South Wales on July 6, 2015


expected servicing cost is then given by


m−1
EC W (m, δ) = kCr pk (0, W )
k=0
∞

W 
+ (k − 1)Cr + C p (δ, x) pk−m (δx, W − (1 − δ)x) f (m) (x) dx
k=m 0

m−1
= Cr kpk (0, W ) +
k=0

 
W 
∞ 
(k − 1)Cr + C p (δ, x) pk−m (δx, W − (1 − δ)x) f (m) (x) dx
0 k=m

W 
= Cr ΛW (m, δ) + C p (δ, x) − Cr f (m) (x) dx. (10)
0

Note that the first term in the right-hand side of (10) represents the expected servicing
cost when the cost of each repair is Cr , and the second term represents the additional
expected servicing cost if the mth failure occurs under warranty and an overhaul takes
place. For fixed m, the first term is increasing in δ whilst the second term is decreasing in
δ. Note also that, for fixed δ, EC W (m, δ) → Cr R(W ) as m → ∞, and this represents the
expected cost of the minimal repair servicing strategy.
The manufacturer’s expected total cost function is

JW (m, δ) = EC W (m, δ) + C f [1 − ΦW (m, δ)] , (11)

and the optimal (m, δ) overhaul strategy is found by minimizing this function. The
optimal values of m and δ are obtained using a two-stage process. In the first stage,
we fix δ and find m ∗ (δ) by minimizing JW (m, δ). In the second stage, δ ∗ is found by
minimizing JW (m ∗ (δ), δ), and then m ∗ = m ∗ (δ ∗ ). Because of the structure of JW (m, δ),
it is impossible to derive any general analytical results and so the optimization procedure
has to carried out numerically.
118 N . JACK AND D . N . P. MURTHY

4.2 Case B
Under the minimal repair servicing strategy, the probability that the customer will make a
repeat purchase from the same manufacturer at time L is

∞ 

ΨL = ψ(k1 , k2 ) pk1 (0, W ) pk2 (W, L)
k1 =0 k2 =0

a1 
a2 
a1 2 −1
b 
b2 − k 2
= pk1 (0, W ) pk2 (W, L) + pk1 (0, W ) pk2 (W, L)
k1 =0 k2 =0 k1 =0 k2 =a2 +1
b2 − a 2

Downloaded from http://imaman.oxfordjournals.org/ at University of New South Wales on July 6, 2015


1 −1
b  
a2
b1 − k 1
+ pk1 (0, W ) pk2 (W, L)
k1 =a1 +1
b1 − a1 k2 =0
1 −1
b  2 −1
b 
b1 − k 1 b2 − k 2
+ pk1 (0, W ) pk2 (W, L) (12)
k1 =a1 +1
b1 − a1 k2 =a2 +1
b2 − a 2

and the manufacturer’s expected warranty servicing cost and expected total cost are
EC W = Cr R(W ) (13)
and
JL = EC W + C f [1 − Ψ L ], (14)
respectively.
Under the (m/W, δ) overhaul strategy, it follows that


pk1 (0, W ) pk2 (δW, L − (1 − δ)W ) if 0  k1 < m, k2  0,


W
sk1 ,k2 (m, δ) = pk1 −m (δx, W − (1 − δ)x) pk2 (W − (1 − δ)x, L − (1 − δ)x) f (m) (x) dx



0
if k1  m, k2  0,
(15)
and the repeat purchase probability is now given by

a1 
a2 
a1 2 −1
b 
b2 − k 2
Ψ L (m, δ) = sk1 ,k2 (m, δ) + sk1 ,k2 (m, δ)
k1 =0 k2 =0 k1 =0 k2 =a2 +1
b2 − a2
1 −1
b  
a2
b1 − k 1
+ sk1 ,k2 (m, δ)
k1 =a1 +1
b1 − a1 k2 =0
1 −1
b  b2 −1

b1 − k 1 b2 − k 2
+ sk1 ,k2 (m, δ). (16)
k1 =a1 +1
b1 − a1 k =a +1
b2 − a2
2 2

The expected number of failures that occur during (0, L] is


Γ L (m, δ) = ΛW (m, δ) + [R(L − (1 − δ)W ) − R(δW )] F̄m (W )

W
+ [R(L − (1 − δ)x) − R(W − (1 − δ)x)] f (m) (x) dx. (17)
0
WARRANTY SERVICING STRATEGIES 119

The manufacturer maintains the item only during the warranty period and always
performs an overhaul. As in Case A, we obtain the manufacturer’s expected warranty
servicing cost by considering the two cases N (W ) < m and N (W )  m. When
N (W ) = k1 < m, all failures during the warranty period are minimally repaired and
the overhaul occurs at time W , so the servicing cost is k1 Cr + C p (δ, W ). However, when
N (W ) = k1  m and the mth failure occurs at age x resulting in an overhaul, the servicing
cost is (k1 − 1)Cr + C p (δ, x). The expected servicing cost is then given by


m−1 
ECm/W (m, δ) = k1 Cr + C p (δ, W ) pk1 (0, W )

Downloaded from http://imaman.oxfordjournals.org/ at University of New South Wales on July 6, 2015


k1 =0
∞
W 

+ (k1 −1)Cr +C p (δ, x) pk1 −m (δx, W −(1−δ)x) f (m) (x) dx
k1 =m 0

= EC W (m, δ) + C p (δ, W ) F̄ (m) (W ). (18)

The second term in the right-hand side of (18) represents the expected extra servicing cost
when the overhaul occurs at time W .
The manufacturer’s expected total cost function is

JL (m, δ) = ECm/W (m, δ) + C f [1 − Ψ L (m, δ)], (19)

and the optimal values of m and δ are obtained using the same two-stage process described
earlier for Case A. Once again, it is impossible to derive any general analytical results.

5. Numerical examples
We consider a warranty period of length W = 2 years, and we assume that the time to
x β
first failure of the item (in years) is Weibull distributed with F(x) = 1 − e−( a ) , so
 x β
R(x) = a . The cost of a minimal repair Cr = 50 and we also let Co = 250, so the
cost of an overhaul at age x with age reduction (1 − δ)x is C p (δ, x) = 50 + 250(1 − δ)x.
Finally, we assume that it is possible for the manufacturer to do a perfect overhaul at any
age, so δmin (x) = 0, ∀x ∈ (0, W ).
In the optimization procedures for Cases A and B, all function evaluations were carried
out on an Excel spreadsheet.

5.1 Case A

 of the item L ≈ 2 years and we assume that a = 0, b = 3, and γ = 0·5,


The useful life
 1
 if k = 0,

0·8165 if k = 1,
so φ(k) = Thus, customer dissatisfaction with the item appears

 0·5774 if k = 2,

1 if k  3.
when there is at least one failure during the two-year warranty period and the customer
switches to another manufacturer for certain when the item fails more than two times.
120 N . JACK AND D . N . P. MURTHY

TABLE 1 Optimal values as δ varies


δ m ∗ (δ) JW (m ∗ (δ), δ)
0 1 519·64
0·05 1 518·27
0·08 1 518·07
0·1 1 518·19
0·2 1 521·76
0·4 1 541·78
0·6 1 572·26
0·8 1 603·52

Downloaded from http://imaman.oxfordjournals.org/ at University of New South Wales on July 6, 2015


E XAMPLE 1 Suppose the penalty cost of losing a customer C f = 1000 and the Weibull
lifetime parameters α and β are 1·5 and 3, respectively, giving a mean time to first item
failure of 1·34 years. With these parameter values for the lifetime, the likelihood of at least
one failure occurring under warranty is greater than 90%, so it is possible that the item
may suffer from a design fault or it might have been subjected to poor production quality
control. However, it is only when a number of the items are sold and warranty claims begin
to occur that the problem of unreliability emerges. The manufacturer then needs to have
servicing strategies in place until the problem can be resolved. If the manufacturer uses the
minimal repair CM strategy throughout the warranty period then we find that there is only
a 42·59% probability that the customer will make a repeat purchase, the expected number
of failures under warranty is R(W ) = 2·37, and the expected total cost per item sold is
JW = 692·65.
The optimal (m, δ) overhaul strategy is obtained by minimizing the function JW (m, δ)
given in (11) using the two-stage procedure described at the end of Section 4.1. For values
of δ varying between 0 and 1, we find that m ∗ (δ) = 1 and the expected costs JW (m ∗ (δ), δ)
are increasing in δ. Thus, the (1, 0) strategy is optimal and the manufacturer should carry
out a perfect overhaul at the first item failure during the two-year warranty period. The
expected cost of this strategy is JW (1, 0) = 568·78, the expected number of failures under
warranty is ΛW (1, 0) = 1·13, and the probability of the customer making a repeat purchase
is ΦW (1, 0) = 0·7702.
Comparing this optimal overhaul strategy with the minimal repair CM strategy, we can
see that the expected number of failures under warranty is reduced by over 50%, expected
total costs per item sold are reduced by 17·9%, and the probability that the customer makes
a repeat purchase increases significantly from 0·4259 to 0·7702.
E XAMPLE 2 If the Weibull lifetime scale parameter α is reduced to 1·25 and the penalty
cost of losing a customer C f falls to 500, we find that an imperfect overhaul is optimal.
Table 1 shows the first stage of the optimization procedure for the (m, δ) overhaul strategy.
For values of δ varying from 0 to 0·8, the optimal values of m and the corresponding
expected total costs are given.
In the second stage of the optimization procedure, the smallest expected cost in the
third column of Table 1 indicates that the (1, 0·08) overhaul strategy is optimal. The
WARRANTY SERVICING STRATEGIES 121

manufacturer should therefore overhaul the item at its first failure during the warranty
period with this overhaul reducing the item’s age by 92%. The reduction in α results in
more frequent failures under warranty but the optimal strategy is still to overhaul the item
at its first failure. However, the 50% decrease in the penalty cost C f allows the optimal
overhaul action to be imperfect.

5.2 Case B

Downloaded from http://imaman.oxfordjournals.org/ at University of New South Wales on July 6, 2015


We again consider a warranty period of length W = 2 years but now the useful life L of
the item is assumed to be 4 years. α and C f are restored to their original values of 1·5 and
1000, respectively, but the Weibull lifetime shape parameter β = 2, in order to reflect the
larger value of L. The costs of a minimal repair and an overhaul are also the same as in
Case A.
We let a1 = 0, b1 = 3, a2 = 1, b2 = 4, and γ1 = γ2 = 0·5, so
 

 1 if k1 = 0, 
 1 if 0  k2  1,
 
0·8165 if k1 = 1, 0·8165 if k2 = 2,
φ1 (k1 ) = and φ2 (k2 ) =

 0·5774 if k1 = 2, 
 0·5774 if k2 = 3,
 
0 if k1  3, 0 if k2  4.

Thus, it is certain that the customer will make a repeat purchase from the same
manufacturer if no failures occur during the two-year warranty period and at most one
failure occurs during the remainder of the item’s useful life. It is also certain that the
customer will switch to another manufacturer if the item fails either more than two times
under warranty or more than three times during the post-warranty period.
If the manufacturer uses the minimal repair CM strategy throughout the warranty
period and every failure during the post-warranty period is also minimally repaired, the
probability that the customer will make a repeat purchase at the end of the item’s useful
life is Ψ L = 0·0893, and the expected total cost per item sold is JL = 999·56.
The optimal (m/W, δ) overhaul strategy is obtained by minimizing the function
JL (m, δ) given in (19). For values of δ varying between 0 and 1, we again find that
m ∗ (δ) = 1 and the expected costs JL (m ∗ (δ), δ) are increasing in δ. Thus, the (1/2, 0)
strategy is optimal and the manufacturer should again carry out a perfect overhaul either
at the first item failure during the two-year warranty period or, if no failures occur in this
interval, then the overhaul should take place when the warranty finishes. The expected
cost of this strategy is JL (1, 0) = 980·70, and the probability of the customer making
a repeat purchase is Ψ L (1, 0) = 0·4012. Compared to minimal repair, there is only a
1·9% reduction in expected costs per item sold but the repeat purchase probability has
significantly increased.
It is also interesting to compare the (m/W, δ) strategy with the original (m, δ) strategy
used in Case A. This will enable us to determine if it is beneficial for the manufacturer
to carry out an overhaul at the end of the two-year warranty when the critical number of
failures m has not been reached.
Note that, if the (m, δ) overhaul strategy is used for Case B, then (15) has to be modified
122 N . JACK AND D . N . P. MURTHY

to


pk (0, W ) pk2 (W, L) if 0  k1 < m, k2  0,

1
 W
sk1 ,k2 (m, δ) = pk1 −m (δx, W − (1 − δ)x) pk2 (W − (1 − δ)x, L − (1 − δ)x) f (m) (x) dx



 0
if k1  m, k2  0,

and the C p (δ, W ) F̄ (m) (W ) term has to be removed from (18). The usual optimization
procedure identifies the (1, 0) strategy to be optimal with JL (1, 0) = 992·83 and
Ψ L (1, 0) = 0·2961. Thus, for the given parameter values, we can see that the (1/2, 0)

Downloaded from http://imaman.oxfordjournals.org/ at University of New South Wales on July 6, 2015


strategy is the best option for the manufacturer. Although the (1, 0) strategy has lower
expected servicing costs, the smaller repeat purchase probability for this strategy produces
much higher expected penalty costs.

6. Conclusions and further research


In this paper, we have carried out some initial work on warranty servicing and customer
satisfaction. We have proposed a method of modelling customer satisfaction that is based
on the number of times an item fails during its useful life L, and we have assumed simple
forms for the probability of retaining a customer.
In addition to minimal repair, we have discussed two other servicing strategies for
a manufacturer that reduce item failures and hence control customer satisfaction. The
particular strategy used depends on the length of the warranty period W in comparison
to L, and we have considered the two cases L ≈ W (Case A) and L  W (Case B),
respectively. In each servicing strategy the manufacturer overhauls the item if a certain
number of failures occur during the warranty period. The decision variables in the two
models are the failure number at which the overhaul takes place and the quality of the
overhaul, which is measured in terms of a reduction in the item’s age.
Lasser et al. (1998) investigate how the actual timing of item failures affects customer
dissatisfaction. They report that customers react more negatively when failures occur early
in the warranty period rather than later. They also find a similar behaviour for failures
during the post-warranty period. Failures just after the warranty expires cause higher
dissatisfaction than those that occur later, although both types of event are not covered
by the warranty.
These ideas about the timing of item failures suggest further models that can
be developed to extend our initial research. For Case A, customer satisfaction
will depend on Z , the time of the first failure in (0, W ]. The probability of
retainingthe customer, conditional on Z = z, could then be given by the function
0  if 0 < z  W1 ,
ψ(z) = 1−e−γ z where the interval (0, W1 ] represents the early
1−e−γ W
if W1 < z  W,
part of the warranty period. The manufacturer could then reduce customer dissatisfaction
by increasing the time to the first item failure. This could be achieved, for example, by
scheduling preventive maintenance (PM) at a specific time T ∈ (0, W ) if and only if no
failure had occurred up that time, and where the effect of the PM action would be to reduce
the item’s age to δT , δmin (T )  δ < 1.
WARRANTY SERVICING STRATEGIES 123

For Case B, customer satisfaction will depend on Z 1 , the time of the first failure in
(0, W ], and on Z 2 , the time of the first failure in the post-warranty period (W, L]. The
probability of retaining the customer, conditional on Z 1 = z 1 and Z 2 = z 2 could then be
given by the function ψ(z 1 , z 2 ) = ϕ1 (z 1 )ϕ2 (z 2 ), where
 
0  if 0 < z 1  W1 , 0  if W < z 2  W2 ,
ϕ1 (z 1 ) = 1−e−γ1 z1 and ϕ2 (z 2 ) = 1−e−γ2 (z2 −W2 )
1−e−γ1 W
if W1 < z 1  W, −γ2 (L−W2 ) if W2 < z  L.
1−e
Reducing customer dissatisfaction now involves the manufacturer trying to increase the
time to the first item failure during (0, W ] and during (W, L]. This could be achieved by

Downloaded from http://imaman.oxfordjournals.org/ at University of New South Wales on July 6, 2015


scheduling two PM actions at times T1 , T2 ∈ (0, W ) if and only if no failure had occurred
up time T1 , and where the age reductions (1 − δ1 )T1 and (1 − δ2 )T2 are carried out.
The forms for the probability of retaining a customer that are used in the examples in
this paper and also those mentioned above all involve parameters that need to be estimated
from customer behaviour. Collaborative research with the manufacturer’s marketing
department would provide this information and then a proper sensitivity analysis of the
model parameters could then be carried out.
In addition to servicing strategies, there is also a need to investigate other strategies for
the manufacturer that involve giving customer refunds or extensions to the basic warranty
period, depending on the item’s failure history. The modelling of such strategies will
involve factoring the effect of any refunds or warranty extensions into the probability of
retaining the customer.

R EFERENCES
B IEDENWEG , F. M. (1981) Warranty Analysis: Consumer Value vs Manufacturers Cost. Ph.D.
Thesis, Stanford University, USA.
C HUN , Y. H. (1992) Optimal number of periodic preventive maintenance operations under warranty.
Reliab. Engng Syst. Safety, 37, 223–225.
DAGPUNAR , J. S. & JACK , N. (1994) Preventive maintenance strategy for equipment under
warranty. Microelectron. Reliab., 34, 1089–1093.
I SKANDAR , B. P., M URTHY , D. N. P. & JACK , N. (2002) A new repair-replacement strategy for
items sold with a two-dimensional warranty. Comput. Oper. Res., in print.
JACK , N. & DAGPUNAR , J. S. (1994) An optimal imperfect maintenance policy over a warranty
period. Microelectron. Reliab., 34, 529–534.
JACK , N. & M URTHY , D. N. P. (2001) A servicing strategy for items sold under warranty. J. Oper.
Res. Soc., 52, 1284–1288.
JACK , N. & M URTHY , D. N. P. (2002) A new preventive maintenance strategy for items sold under
warranty. IMA J. Manage. Math., 13, 121–129.
JACK , N. & VAN DER D UYN S CHOUTEN , F. A. (2000) Optimal repair–replace strategies for a
warranted product. Int. J. Prod. Econ., 67, 95–100.
K IJIMA , M., M ORIMURA , H. & S UZUKI , Y. (1988) Periodical replacement problem without
assuming minimal repair. Eur. J. Oper. Res., 37, 194–203.
L ASSAR , W. M., F OLKES , V. S., G REWAL , D. & C OSTLEY, C. (1998) Consumer affective
reactions to product problems when timing of warranty expiration varies. J. Business Res., 42,
265–270.
M URTHY , D. N. P. & N GUYEN , D. G. (1988) An optimal repair cost limit policy for servicing
warranty. Math. Comput. Modell., 11, 595–599.
124 N . JACK AND D . N . P. MURTHY

N GUYEN , D. G. (1984) Studies in Warranty Policies and Product Reliability. Ph.D. Thesis, The
University of Queensland, Australia.
N GUYEN , D. G. & M URTHY , D. N. P. (1986) An optimal policy for servicing warranty. J. Oper.
Res. Soc., 37, 1081–1088.
N GUYEN , D. G. & M URTHY , D. N. P. (1989) Optimal replace-repair strategy for servicing items
sold with warranty. Eur. J. Oper. Res., 39, 206–212.
Z UO , M. J., L IU , B. & M URTHY , D. N. P. (2000) Replacement-repair policy for multi-state
deteriorating products under warranty. Eur. J. Oper. Res., 123, 519–530.

Downloaded from http://imaman.oxfordjournals.org/ at University of New South Wales on July 6, 2015

You might also like