PERANCANGAN MODEL PREDIKSI LEAD-TO-WEBSITE WWW.MENANGGAMING DARI KANAL INSTAGRAM @MENANGGAMING BERDASARKAN ATRIBUT KONTEN POST INSTAGRAM

• Main Author: Reynard Affandi, Antonius
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/66152
• Institution: Institut Teknologi Bandung

Infia's core business as one of player in the media and advertising industry is to present high-traffic news/content in various fields through a number of Instagram social media accounts. Infia expands its business model from only managing internal accounts on a regular basis by adding business collaborations in managing social media accounts with other brands. Since July 2019, Infia has formed a partnership with PT.X as a leading laptop manufacturer in Indonesia to form an white-label Instagram account, @menanggaming, which is indirectly expected to attract Instagram users to use the latest PT.X gaming laptop. Due to the young age of the Instagram account, Infia has the challenge of optimizing the published content to trigger customers to visit the exclusive website for gaming with PT.X, menganggaming.com. This study aims to design a quantitative model that can predict the number of leads-to- websites of menganggaming.com based on the attributes of the content and Instagram account @menanggaming, making it easier for Infia to identify variables that affect interactions with Instagram users. The prediction model is based on 456 daily data from the @mennggaming account which is spread over 2000 posts. The lead-to-website prediction model is constructed using two different methods, namely regression and machine learning. Of the several prediction models used, multiple linear regression and random forest (regressor) methods produce the best performance and easy interpretation. The R2 values for the training and testing data set of the lead-to-website prediction model from the multiple linear regression and random forest methods were 54.69%, 42.85%, 90.31%, and 73.7%, respectively. The selected variables for the design of the final model are the number of post types in images format, the number of post types in the form of videos, the number of post types in the form of a carousel, content types in the how category, number of post impressions, and content upload time in the range 00.00 – 10.00 WIB. The research concludes with suggestions for @menanggaming Instagram content marketing strategies based on the variables that most influence the number of leads-to- websites and suggest for improving the effectiveness of @menanggaming account's business processes.