People's reliance on the internet for acquiring institutional rankings to make comparisons has increased significantly in this day and age. Recently, researchers have begun to correlate institutional rank lists provided by government organizations such as the Research Excellence Framework (REF, UK) with ranked lists compiled by Google Scholar (GS). In this technological era, where well established tools Google Scholar and Microsoft Academic are complementing conventional sources Scopus and Web of Science. Unlike Google Scholar still provides the maximum coverage in contrast of other sources. To address this gap why is it so? An empirical investigation has been performed in comparison to Google Scholar, Web of Science, Scopus and Microsoft Academic. The data obtained from these sources is turned into structured data and subsequently used to perform statistical analyses on citations. These analyses are then used to determine rankings based on citation and publication metrics. Most popular of all is the bespoke algorithm that produces a correlation greater than 0.78 between google scholar and REF generated rank list. This study extends the bespoke algorithm by re-implementing it in python language; integrating exceptional handling and client-server architecture to make it scalable, and finally excluding duplicate author’s profiles for refining consolidated citations before generating the rank list of the institutes using Google Scholar [1]. These methods are specialized in extracting the institutional citation but lack to capture feature dependencies effectively with other sources. After the experiment, the results compared to check whether the results are altered or the institutes ranking changed by altering the source. This study stood for the utilization of a predictive approach within the framework of correlational research. Moreover, the experimental evaluation yielded promising results when compared to the other multiple data extraction sources which implemented by the proposed taxonomy. This paper examines the existing evaluation limitations and proposes potential strategies for enhancing the research impact algorithm.