The Use of Repeatable Components in Hybrid Models to Enhance Software Project Management Success

The management of software project development requires a dynamic and reactive environment to meet shorter time-to-market demands to address competition efficently in the software industry. This scenario requires the use of effective and robust methodologieswhere opportunities are not lost due to delays and failures in timely software project deliveries.The Agile Manifesto in 2001 which introduced 4 values and 12 principles was designed to develop and manage software projects in a more suitable and effective way to improvethe success rates of software projects. But, increase in overall success rates are still not significant with failure rates remaining plauteaued at about 30% over the last 10 years. Hybrids methodologies seem to have worked better as agile hybrid management methodshave shown more promise when compared to pure agile methods with an overall success rate increase of 16%. There is evidence too that by combining agile methodologies with traditional methodologies, there would be a further increase in success rates. Whilst many hybrid methodologies have been suggested and researched, the gaps in the literature review reveal there is a lack of hybrid models that have been empirically developed and studied as second order components. To build a robust hybrid model, it is important to gather the relevant information and careful consuideration must be given to the design of the questionnaireto fit second order components and models must incorporate and provide for the use repeatable ways to test models once the data is collected.This paper presents a review of the current gaps in hybrid methodologies and proposes a questionnaire design that supports the research methodology and empirical study to be undertaken with second order components (Constructs).Further it looks at the design approach in questionnaires which incorporates the use of repeatable constructs and the measures used and emphasizes this as an important ingredient for developing and testing hybrid models in research studies.


Introduction
Project management as a discipline that is not categorized as an exact science and the primary constructs in a theory are required to be well defined (Gregor, 2006)  A recent CHAOS report in 2018 (Figure1) provides evidence of a success percentage hovering between a range of 27% -31% with slight improvements over the years. The report reviwed data from over 50,000 projects software projects which ranged from tiny enhancements to massive systems re-engineering and implementations. The definition used for failure in the CHAOS report was done on failed projects which typically include projects completed but with a very poor quality both in the product and the processes used (Magne Jørgensen, 2014). Studies in this area would require an initial and independent look at the traditional and agile software project management methodology characteristics and an eventual review of the combined characteristics through models that support the various concepts and theories.
The study of characteristics can be viewed as the study of indicators in the development of constructs and models and the use of repeated indicators are primarily useful in the hybrid model development to assist in striking a balance between the two methodologies due to similarities and measures of success between the two techniques (Papadakis & Tsironis, 2018) Careful consideration is necessary for the design of questionnaires (Roopa S, 2012) which should include the various characteristics as constructs and indicators which use measures and provide for repeatable constructs to study the proposed model in terms of its validity and fit.

Materials and Methods
A review of the gaps in current traditional, agile and hybrid project success management models reveal that the hybrid architecture that requires a reactive behavior and a dynamic environment is not evident (Table 1). To build efficient hybrid constructs, the gaps in traditional and agile constructs must be combined as gaps in hybrid construct to exist as repeatable gaps when the overall model is tested. Variables should be introduced (also termed as indicators in some research papers) with measures used to formulated the type of questions that will be required in the questionnaires. This fulfils the basic architecture in hybrid models and design of the questionnaire.

Challenged Failed Successful
A summary of the gaps are provided in Table 1. As hybrid models have a 16% increase in success rate than pure models (Carvalho et al., 2012), a new hybrid project management model was developed and proposed to close some of the gaps instead of developing a new pure traditional or a new pure agile model.  An instrument in the form of a questionnaire was designed with the set of indicators that were used as measures in the model. The design comprised a set of 4 components (constructs) with ordinal (Likert-scale) questions for each component which ensured repeatable indicators could be used for analysis (see figure 3). This is summarized in Table 2.
The conceptual representation of hybrid components (constructs) is provided in Figure 2. Items in 1 st order construct will be represented as repeated items in the 2 nd order construct (Wilson & Henseler, 2007).

Figure 2.Conceptual representation of a Hybrid Construct with Repeatable Indicators
The agile hybrid proposed model was also developed from existing traditional and agile characteristics as combined for the hybrid construct to address the gaps. The proposed model and the measurement indicators are provided in Figure 3.Due to a large number of indicators the abbreviation '01 to nn' is used, e.g. T01....Tnn for Traditional Model indicators.

Figure 3. Repeatable components (items) for hybrid models
Data from 296 respondents for the various components (used as constructs) in the model were collected and a model fit was obtained using the SMART-PLStool. The hybrid component required the 1 st order and 2 nd order analysis to evaluate the model fit.

Findings
The analysis for a model-fit using SMART-PLS is provided in Table 4

Conclusion
The questionnaire design and the model fit analysis provides a good basis for researches that use hybrid model to have repeatable indicators and use 1 st and 2 nd order construct analysis to determine the model fit. As the data used in the analysis is a good sized sample for a specific interest target group, it is reasonable to suggest that a larger sample would increase the project successmodel fit percentage with more refined questions and further improve the model-fit analysis and values.