The data regarding the attributes of key stakeholder groups in the choice projects were collected in the third part of questionnaire section two. 112 respondents filled the questions10. The mean value was used to compare the stakeholders based on their attributes.
The results are shown in figure 6-6. The questions asked with regard to each attribute could be found in table 6-1 given earlier. Some notable findings are presented here.
Clients had the highest power to influence design quality decisions in the choice projects. The second position was allocated to the architects. This is not surprising as the clients often have the highest authority and architects could influence their decisions (as also found in Study One). Among the stakeholders, users and facility management team (FM) had the lowest power. Although, architects had lower power compared to the clients, they were more involved (proximity) in design quality decision-makings. The clients were in the second and contractors in the third positions. For user group and FM who had low power and proximity, the urgency was found relatively higher. This indicates that although they did not hold high power to influence decisions and were not highly involved, their interests and requirements were considered relatively higher as critical by the decision makers. Still, their urgency levels were lower than that for clients and architects.
Figure 6-6: Mean values for stakeholder attributes in choice projects
Apart from the descriptive analysis above, a correlational statistical test was used in order to assess whether there is any significant association between these stakeholder attributes. As mentioned in section 6.3.2., Spearman correlation was used. The results are shown in table 6-13 for each stakeholder group. As it can be seen, all pairs of stakeholder attributes
correlate significantly (at 0.01) with each other. Moreover, this association is achieved consistently with regard to all five stakeholder groups. The finding indicates that the possession of high power to influence design quality could be accompanied by high involvement, which itself could be related to high urgency of the stakeholder requirements. It is interesting that, the association between the stakeholder attributes was never looked into by those investigating Mitchell et al.’s (1997) salient model or Stakeholder CircleTM.
Table 6-13: Results of Spearman correlation analysis (**: significant at 0.01)
6.4.8
Discussion of the study approach
Although the identified objectives were achieved, the study had some limitations. In this section, limitations as well as the strengths of the study approach are discussed. To start with, the study successfully completed the sequential mixed methods approach defined in Chapter 3. The qualitatively derived success factors in the second study were used to develop a data collection instrument in Study Three via which the factors were quantitatively
validated (triangulation) and further evaluated. A next sequence could however be designed in which the results of this study are verified in a number of case projects, which was not feasible in this research.
The use of pre-tests for the questionnaire and the inclusion of relevant definitions and clear instructions contributed to the face and content validity of the data collection instrument. The Cronbach’s alpha with high coefficient computed for multiple question sections, also, addressed the reliability issue. The reliability was further strengthen by the participation of relevant and relatively highly experienced and senior respondents.
On the other hand, the use of non-probability sampling and approaching only two stakeholder groups in the UK could have a negative impact on the sample representativeness. The ideal was to obtain a random sample as a reliable representative of the study population, i.e., the clients and architects with experience in multi-stakeholder building projects in the UK. However, this required probabilistic methods – or probability sampling. On the other hand, non-probability sampling techniques rely on the subjective judgment of the researcher. A disadvantage of non-probability sampling is that the researcher could not make statistical inferences from the sample being studied to the population of interest. It would be very difficult to get access to and/or find a list of the population.
However, as mentioned earlier, this was necessary in order to maintain a manageable scope while approaching the most accessible, appropriate individuals. As discussed, the sample size was satisfactory for the statistical analyses in the study however the low response rate is an indication of the challenges occurred in the participant recruitment. To address this, the study utilised multiple approaches (online and paper-based) as discussed. Also, in comparison with similar studies, the sample size was still high and adequate. A reason for low response rate could be related to the questionnaire length, but one could argue that once a potential respondent, especially a professional or expert one, decides to fill the questionnaire, he or she would often do. It should also be noted that there were more
architect respondents than client ones, therefore, the findings may more closely reflects architects’ opinions. Lastly, the findings are based on subjective perception, although the unit of analysis for the second section was concerned with choice projects.
With regard to the PCA, high KMO and communality values as well as the high total variance explained (70.67%) reflected the strength of the analysis. Due to the analysis constraints, one success factor had to be dropped for the second run, but the result showed meaningful interpretable components with a simple structure. This factor omission however was mitigated by its inclusion in the regression analysis as an individual variable. The PCA approach was exploratory, therefore, a next line of enquiry could be to conduct a Confirmatory Factor Analysis using a separate sample, to verify the groupings.
For the regression analysis, although OR was the first choice, its assumption was violated. Therefore, MLR, as a common approach in similar studies, was used for which the assumptions were fulfilled. The three regression models used in the study had satisfactory R2 values, giving a good explanation of the variance by the model and effect size. In terms
of the generalisability (external validity) a number of points should be considered: the sample size was adequate; the MLR assumptions were satisfied, and the cross-validation test showed suitable results. However, still, the sample could not be claimed to be an accurate representative of the population and the applicability of the findings should be considered with caution. A larger sample size with the inclusion of a more diverse respondent groups, project sectors, and other geographical locations could improve the representativeness issue. Finally, the comparison conducted between the stakeholders in terms of their attributes was descriptive and limited to the sample. Also, the identified correlation was not conclusive, yet its recurrence for all five stakeholder groups is indicative of such relationships in the population.
6.5
Conclusion
The study resulted in a better understanding of design quality CSFs at different layers. The study validated 28 factors as critical in successful achievement of design quality in building projects, where the ‘brief’, ‘communications’, ‘leadership’, ‘knowledge & skills’ and ‘construction quality’ were found to be the highest ranked factors in the perception of the participants. It was also found that architects and clients do not significantly differ in perception regarding the importance of the factors except for stakeholder identification and POE.
The study found that in real-world projects, the CSFs interrelate with each other and could be grouped into 7 categories (6 components + F16). This provided a more parsimonious yet meaningful summarisation of the CSFs list. The study further revealed how these CSFs categories (i.e., grouping of ‘the how’ factors) impact on different criteria of design quality success ‘(the what’ to design quality achievement) in real-world projects. It was, for example, found that F16 ‘knowledge & skills’ significantly contributed to the achievement of building functionality while its build quality is influenced largely by C2 ‘overseen DQ execution with commitment’. The study successfully modelled the relationship between design quality CSFs and success criteria. Lastly, through applying the second proposed stakeholder model, it was revealed that key stakeholders differ in terms of individual attributes of power, proximity and urgency. Moreover, these attributes were found to correlate with each other.