<![CDATA[The era of digitalization has boosted companies to adopt innovative technology in seizing competitive superiority. However, adaptation to the change often deals with constraints, affecting operational efficiency and effectiveness. In this regard, PT Beiersdorf Indonesia faces issues regarding the impact of digital transformation of the sales process from manual to digital using Beiersdorf Sales Activity Tracking Routine (BERSATU) application, such as the difficulty of technology adaption by employees, accustomed to manual method, limitations of internet access in some areas, and technical constraints on application as well as insufficient technical support. Therefore, the research aims to analyze sentiment of user review on “BERSATU” application, using various algorithm classification and modeling topic. A total of 600 user reviews had been collected and analyzed to know positive or negative sentiments. Of 600 reviews, 60.33% was positive and 39.67% was negative. Based on the evaluated five algorithms classification, SVM and Naïve Bayes were superior with accuracy above 97% with better F1-Score. Regression Logistics had 96% accuracy, but it had low recall. Random Forest had 94% accuracy with a better F1-Score, but low recall was received. KNN has 93% accuracy, but low recall was obtained. SVM and Naïve Bayes were the recommended for further analysis. Modeling of LDA topics generated three topics with dominant keywords. Topic 1, such as “application,” “bersatu," “data,” “sales,” and “results,” covered 44.5% of the documents. Topic 2, including “application,” “good,” “visit,” and “order,” was relevant at 29.5%. Topic 3, comprising of “application,” “assistance,” “easy,” “sell,” and “store,” was relevant 26.1% of the documents.]]>
