The most fundamental challenge of all is to identify rightly the Research problem. Investigation is not worthwhile, unless it carries a firm direction and measured outcome. For instance, when a manager reports about drop in sales, the tendency is to state the problem of Sales as “dip in sales”. Going a bit deeper, one can understand the problem seem to lie not simply with Sales, but customer relationship itself or the CRM. To sum up, when customers do not perceive the value of product offering, it might lead to lesser sales but is a typical CRM issue. Secondly, the handling of research problem is a challenge of its own. Problems at a larger level might at times appear too complex to solve. Nevertheless, if the same can be broken into multiple units, the solution is much easier to find. This approach will help us develop both bird’s view (macro) and worm’s (micro) view at the same time. Problems emerge, that has earlier (antecedence) appeared in different contexts and has a common thread in the form of Theoretical Framework. Therefore, setting of the research problem in the context of a sound theoretical background is critical. Once, this is done, our attention drifts towards how to measure the phenomenon we seek to diagnose. Identifying a Suitable measure is battle half won. Psychometric studies are normally quite challenging, since they do not measure directly and cannot measure directly the deeper, subtler aspects of human behaviour. For instance, perception and preference may prima facie appear common, but they are not identical. They have different measures, and choosing right one is the next challenge. Further, the measures we adopt should be reliable, repetitive and must be in alignment to the previous literature on that particular domain. Now that the measure is identified and instrument designed; applying the appropriate statistical tool is important. We often end up choosing an inappropriate technique, mostly due to our inability to clearly identify nature of data we collected through our instrument. For instance, for a dichotomous scale we cannot go with linear regression and for sales prediction-essentially a numerical value we cannot apply binary logistic regression. What we need to understand is that, choice of statistical technique is based on three important conditions
- Purpose of the study
- Type of data and its distribution and
- Number of dependent and independent variables.
Further, to compute the data, there is no dearth for software. Plenty of which is available in the form of SPSS, SAS, STATA, R and others. Though software’s does its function predictably, for instance psychometric analysis is better computed on SPSS, whereas econometric data is best done with STATA; Secondly, whether the software is open source or licensed product also to be understood, since open source product like “R” has tremendous flexibility in terms of additional and new algorithms added every day by a community of exceptional professionals, available at virtually zero cost. Finally, reporting is the critical portion we often tend to neglect. It is not what we compute but what we communicate that matters. Reporting is based on target audience, if it is on journal, following the particular style like APA is advisable, if it is common man, neither a statistician nor a researcher, story telling approach is much better than explain through spread sheet, rigour can be down played and focus on how study outcomes is relevant for target audience.