Dimensiones generales del rotor

As reviewed above, after Martilla and James (1977), quite a few other studies have followed the same approach of analysing attribute importance and attribute performance with self-stated measures, which is known as traditional IPA. In these traditional IPA studies, the quadrant values are from factor importance and factor performance, as the authors find it more meaningful to research the data of importance and performance simultaneously than only to research the data of importance or performance separately.

Aaker and Day (2004) reported the importance-performance model. Traditionally, the four-quadrant matrix can identify areas needing improvement and areas of effective performance (Shieh and Wu 2009).

Deng et al. (2008b) argue that the traditional IPA has two implicit assumptions: 1.

Factor importance and factor performance are two independent variables. 2. The relationship between factor importance and factor performance is linear and symmetrical. They claim that the two assumptions are wrong, because: 1. The two variables are not independent. 2. The authors note that the relationship between them is not linear but causal (Matzler et al. 2004). This implies that the traditional IPA approach can be misleading (Bacon 2003; Matzler et al. 2004). Some researchers have tried to revise the traditional IPA.

Revised IPA by gap analysis

Apart from plotting items on a four-quadrant matrix, gap analysis is used to simultaneously consider importance and performance and identify the areas for improvement. There are two types of gap analysis in IPA.

The first type of gap is measured as performance minus importance. Platts and Gregory (1992) conducted IPA by employing the rating of importance and performance difference in manufacturing for strategy formulation. Ainin and Hisham (2008) applied the IPA to information systems in Malaysia. They indicated that the gap between the importance and performance implies the opportunities for improvement and guides the prioritization of resources and management intervention. Ford et al. (1999) developed a

―gap-based‖ approach that compared importance with performance to implicitly set improvement priorities. They regarded the gap of performance minus importance as the room to improve and applied IPA in the educational service. Two case studies were undertaken in New Zealand and the USA to develop a strategic tool for education service market improvement. Two important contributions they made to this research: 1.

They identified the problematic attributes by the importance-performance difference. If the mean performance minus mean importance is negative, the attribute shows a potential problem. The bigger the difference is, the bigger the problem is with that attribute. 2. They identified a significantly different factor structure between the two countries, although they investigated the same attributes. The results suggest that trying to develop a single model of important facts to apply in a cross-cultural context might be a mistake. Two years later, Johns (2001) noted that ―quality= performance score – expectations score‖, which is similar to the view of Ford et al. (1999).

The second type of gap is measured as focal performance minus competitor or bench marker‘s performance. The competitors‘ performance is treated as an explicit benchmark by which to judge the operation‘s performance. The performance difference (∆performance) is treated as the gap of the focal organisation to improve. This is different from traditional IPA that only considers focal performance.

Yavas and Shemwell (2001) extended the traditional IPA model by integrating competitor‘s performance. They state that a respondent‘s index score for a given attribute is equal to his evaluation of the importance of an attribute times the difference between his assessment of the competitor and the focal object performance. The X axis presents performance from low/left to high/right; the Y axis presents the relative

(mean importance rating for all the factors taken collectively) are considered as salient.

In their modified IPA matrix, if both performance and relative performance index are high, the activity is a competitive edge and should be ―keep up the good work‖; if both are low, there is a competitive disadvantage, a red alert is given and urgent actions should be taken; if performance is high and importance is low, that activity falls in the quadrant of vulnerability and competitive watch; if performance is low and importance is high, it indicates false security and opportunity alert. The entity needs to stay alert to actions.

Other researchers and authors such as Slack (1994), Johns (2001), Yeo (2003), and Lin et al. (2009) hold a similar view that traditional IPA should be integrated with gap analysis. They understand that service quality is the degree of discrepancy between customers‘ expectations (importance) and perceptions (performance) of the service. The gap between the competitor‘s performance and focal performance needs improving, if the result of competitor performance minus focal performance is positive. The bigger the gap is, the more effort is needed to improve the focal performance.

Shieh and Wu (2009) apply IPA to the retail sector to examine how services in convenience stores could be improved. They evaluate the performance by the mean value and evaluate importance by the variance-based methods. The basic idea of variance-based importance is that the larger variance a variable has in its ratings, the more important the variable is. This method is particularly useful when the importance is not directly available from the survey.

Re-dividing the quadrants of the IPA model

Slack (1994) and Slack et al. (2001) analyse the relationship between importance and performance and modify the traditional IPA to reflect managers‘ perceived relationships between ―importance‖, ―performance‖ and ―priority for improvement‖. They note that derivation of a ranked list of competitive factors is crucial for a business operations strategy and the importance-performance matrix is important for both internal and external service improvement. They segregate the importance-performance matrix into four zones instead of four quadrants, namely, the ―appropriate‖ zone, the ―improve‖

zone, the ―urgent action‖ zone and the ―excess‖ zone, which implies very different strategies. The four zones are separated by three lines: lower bound of acceptability, distinction and approximate boundary. The manager‘s views of ―better than competitors‖

are treated as the boundary line of acceptability, which distinguishes what is acceptable performance and what is unacceptable performance. The authors consider that

―importance‖ and ―performance‖ act together to determine ―priority for improvement‖, as there are some interesting indications of the links between importance, performance and perceived improvement priority. The priority which managers give to improving particular competitive criteria depends on their importance, and managers are concerned only with performance levels that are clearly below those of their competitors.

Slack (1994) argues that the 2x2 zoning does not hold for intermediate points and suggests a 9x9 zoning for the importance-performance relationship analysis. The 9x9 format (Figure 4.5) is quite different from the 2x2 format because the boundary lines are quite different, although it follows the same intuitively acceptable rationale. The boundary might be low in practice as managers would tolerate poor performance if that activity is relatively unimportant (8 or 9 on the importance scale), as shown in the area of ―appropriate‖.

Figure 4. 5 Importance-performance matrix zones (Source: Slack 1994)

The minimum boundary line of acceptability (line AB) is the competitors‘ performance.

Above it is appropriate except factors of ―excess‖ (separated by line EF between

―appropriate‖ and ―too good‖) that are over-performed with low importance. Below the boundary line are factors that need improving except the factors of ―urgent action‖

(separated by distinction line CD between ―urgent priority zone‖ and ―less urgent improvement zone‖) that are very important with very low performance. The short-term objective is to raise the performance of ―urgent action‖ up to the ―improve‖ zone.

The 9x9 tool is theoretically better than the 2x2 matrix as it distinguishes the grid more concisely. Low performance is tolerable when the importance is low, while performance needs improving even though the performance is not poor, as the benchmark is competitors‘ performance. However, the boundary line between acceptable and unacceptable is blurred and difficult to define, as managers‘ views are quite subjective and competitors‘ performance is hard to define. This tool is difficult to employ practically. In the current research, as the competitors were hard to define, it was more difficult and not possible to define an accurate value for the boundary line of AB. For lines CD and EF, it is even more difficult to get the value to form the lines. Hence this research did not employ the 9x9 formats with four zones.

The factors with high importance and a big gap of performance difference are called

―salient factors‖ by Brooks (2000). Mangan et al (2002) identify the ―salient factors‖ on port/ferry choice in RoRo freight transportation. The authors employ the Aaker and Day Model (2004), which was applied by Deng (2008) later on. Based on their work, an importance-∆performance model is developed as Figure 4.6 shows.

Low performance dif. High

―Possible overkill‖

Quadrant II

―Keep up the good work‖

Quadrant I

“Salient factors”

―Low Priority‖

Quadrant III

―Concentrate here‖

Quadrant IV

Low Importance High Figure 4. 6 Importance-performance analysis

Source: adapted from Martilla and James 1977, Mangan et al.2002, Deng et al. 2008

Each quadrant provides management information or service strategies. Variables in quadrant I (high importance and high performance) represent competition and are deemed major strengths; the service should be maintained, leveraged, and heavily promoted (Lambert & Sharma, 1990). The organisation should ―keep up the good work‖

because it shows the focal performance meets customers‘ satisfaction. Items here are identified as ―salient‖ factors. Quadrant II represents low importance but high performance, which means the resources are over allotted. The organisation can thus allocate a portion of the resources to the variables with high importance and improve the other variables, for example Quadrant IV variables (‗concentrate here‘) to achieve a

more efficient flow and allocation of the port‘s resources. Quadrant III represents low importance and low performance. Thus, the organisation should consider stopping or decreasing the resources to these variables. Quadrant IV represents high importance but low performance; these items are major weaknesses and should be top priority and targeted for immediate improvement efforts.

Revised IPA employing the three-factor theory

Matzler (2003), Deng (2008) and Deng et al. (2008) question traditional IPA by employing the 3-factor theory. The three factors refer to basic factors, performance factors and excitement factors. According to Matzler et al. (2003), the basic factors are minimum requirements that cause dissatisfaction if not fulfilled but do not lead to customer satisfaction if fulfilled or exceeded. They are basic requirement and of utmost importance. Performance factors lead to satisfaction if fulfilled or exceeded and lead to dissatisfaction if not fulfilled. They cause satisfaction or dissatisfaction depending on their performance level. They are the second most important. Excitement factors can increase customer satisfaction if delivered but do not cause dissatisfaction if not. They are the least important as they comprise augmented or enhanced services.

The authors argue that this theory has two features: 1. Importance of a basic or excitement factor is based on its performance. Basic factors are crucial when performance is low and excitement factors are crucial when performance is high (Matzler et al. 2004). 2. The relationship between factor performance and overall customer satisfaction is asymmetrical.

Various ways of positioning the grid lines

IPA is a graphic technique and the interpretation of the ―Action Grid‖ depends on the quadrant where the factor is accurately placed (Crompton and Duray1985). The correct positioning of the factors is critical to derive the marketing strategy.

The literature has reviewed different ways of positioning the grid lines. Firstly, Martilla and James (1977) suggest that the positioning of the boundary lines is a matter of judgement. They note that the value of IPA lies in determining relative rather than absolute levels of importance and performance. That is why they practically move the axes in case of the absence of low importance and low performance ratings. Secondly, Guo and Zhang (1997) treat the centre of the scale as the grid lines. For example, if the data is collected by a 5-point Likert scale, the centre is 3, so the matrix is divided by 3 on the X axis and 3 on the Y axis. Lastly, Martilla and James (1997) adopt grand mean

instead of medians as the grid lines to avoid discarding the additional information contained. Following that, Ford et al. (1999) and Huang et al. (2006) employ grand mean as the grid lines, which is the overall average of all the attributes on importance and on performance. They use the means of overall service expectation and satisfaction as the boundary lines to separate the grid into four quadrants. However, they explain that median values are theoretically preferable to means, as a true interval scale might not exist.

Although a few options exist in the literature to use median, mean and centre of scale, the grand mean is the most widely adopted way for the boundary. This explains why this research adopts grand mean as the gridline.

Various methods of measuring attribute importance

A variety of methods have been employed to measure variable importance. Basically, there are two types of importance: explicit self-stated importance and implicitly derived importance.

Explicit self-stated importance

Griffin and Hauser (1993) compare three different measures of explicit importance, namely, direct rating, constant-sum scale and anchored scale. They find no significant differences between these methods. Crompton and Duray (1985) investigate two self-stated methods: plotting by mean values and plotting by median values. Their empirical investigation show little difference between the two self-stated methods. Matzler et al.

(2003) employ three explicit methods (direct rating of importance on a five-point rating scale, a partial ranking method and the mean ranking). A comparison of these rankings shows a strong correlation between the three methods. These results suggest a low sensitivity of importance weights to the measures of explicit (self-stated) importance.

Implicitly derived statistical importance

Self-rated importance rating has some shortcomings: 1. The researchers tend to include attributes salient to the customers (Chu 2002); 2. Self-rating of importance is subject to response bias due to the influence of social norms and the importance is not predictive of satisfaction (Brooks et al. 2010). Implicit importance, which aims to incorporate the determinant attribute of performance into importance, can complement the shortcoming.

Researchers have presented different methods to generate the implicit statistical importance and develop the revised IPA, including regression analysis, partial correlation, bivariate correlation and composite ranking (Bacon, 2003; Matzler et al.,

2003; Deng et al., 2008a; Huang et al., 2006; Slack, 1994). They use the different methods to infer the priorities for improvement from the importance-performance space as well as different methods to measure the importance of the attributes. Matzler et al.

(2003) note that when some form of implicit measurement of importance is used (e.g., the variable correlation is with an external criterion like overall satisfaction), implicit importance is derived, given the current level of variable satisfaction. The implicit importance measures are derived based on performance perceptions (Van Lttersum et al.

2007). The implicitly derived importance might reduce the errors arising from the subjective judgement by customers who give self-stated importance (Deng et al., 2008b).

Crompton and Duray (1985) investigate two statistical methods to derive implicit importance: plotting by Pearson correlation coefficient and plotting by Spearman correlation coefficient. They find little difference between the two statistical methods.

Deng (2008) uses weightings from partial correlation coefficients instead of mean or median for the analysis. The value of performance is presented in percentage (%) instead of mean/median. He claims that the use of relative importance and relative performance is more suitable for strategy analysis. The partial correlation coefficients are used as implicitly derived importance weights, which are gained by correlating variable performance (satisfaction) with overall customer satisfaction (OCS), as Oliver (1997) states that implicitly derived importance relies on an actual assessment of how each variable is related to overall satisfaction.

Chudasama (2009) derives importance weights from the factor loadings of the principal component analysis. Lin et al. (2009) employ a method of ratio to produce the implicit importance. They use relative importance (RI=importance/average importance) and relative performance (RP=performance/average performance) and combine them with the traditional IPA to produce a revised matrix of importance-performance gap analysis (IPGA). The revised IPGA matrix is similar to the traditional IPA model. The RI and RP measures are represented by the Y axis and X axis respectively to form a two-dimensional matrix. These two axes divide the IPA grid into four quadrants through which the crosshairs are set at the grand means of RI and RP. In this way, each attribute can show up according to its mean rating value respectively.

As Crompton and Duray (1985) and Matzler et al. (2003) identify that implicit importance by difference methods result in similar results, employing one method to derive implicit importance can be representative of implicit importance. Van Lttersum

two measures are complementary, each providing a different perspective on the value of the criterion.

Bacon (2003) reviews different approaches used to undertake IPA and compares them across 15 databases. He finds that using direct measures of importance instead of correlation coefficients is better, as the underlying assumptions of correlations are often not met. However, Matzler et al. (2003) identify that statistical methods are more appropriate than explicit importance as they correlate more closely with actual perceptions (Crompton and Duray, 1985; Neslin, 1981). To make this research rigorous, both explicit importance and implicit importance are considered, so that subjective and objective importance are addressed and compared to avoid biases.

Revised IPA model with explicit and implicit importance

The above gives the growing evidence that both explicit importance and implicit importance are important and should be considered simultaneously to produce the factor structure of customer satisfaction.

Matzler et al. (2003) propose a model importance grid that distinguishes the factors of customer satisfaction into three categories and places them into four quadrants in the matrix. The three-factor theory suggests that customers‘ evaluation of variable importance does not adequately measure the implicit importance of variables. Based on the three-factor theory (Section 2.3.4), based on the work done by Matzler et al. (2003), Deng (2008), Deng et al. (2008), Lin et al. (2009), and based on the underpinning of IPA, a new model is put forward as Figure 4.7 shows.

Low Implicitimportance High Excitement Factors (3) High implicit importance/

In this model, the horizontal axis represents explicit importance while the vertical axis represents implicit importance. Different methods can be employed to produce the values of the explicit importance scale and implicit importance scale, as presented earlier. Items that fall in Quadrant IV (low implicit importance and high explicit importance) are basic factors. Items that fall in Quadrant I (high explicit importance and

high implicit importance) and III (low explicit importance and low explicit importance) are performance factors. Items that fall in Quadrant II (low explicit importance and high implicit importance) are excitement factors.