Prediction of user behavior and users segmentation in a government-owned digital application using Markov Chain model

Kurikulum Merdeka (KM) was launched by the Ministry of Education, Culture, Research and Technology in Indonesia as an alternative curriculum during the pandemic era. It uses a digital application named Platform Merdeka Mengajar (PMM) for its implementation. As of August 2022, among the five main digital products in PMM, the most popular feature in PMM is Toolkit, which then can also be considered as the users’ entry point of PMM. In order to approach every single segment in a tailored manner to promote the desired-outcome, the team is trying to identify PMM’s user segmentation and predict future segment movement based on the current users’ behavior in the platform. This paper focuses on developing a Markov Chain Model to predict user behavior and customer segmentation levels on Toolkit based on recency and frequency values of users’ download activity. During the twelve months of the observational period, this study comprised more than 400 thousand unique users who downloaded teaching-materials. As a stochastic approach, Markov chain model predicts future behavior of 16.76% of users to remain using the product. Further, the prediction on users’ segmentation using recency and frequency framework, yielded 27.98% of users would be classified as promising users. Such results were used by the team to formulate a better intervention strategy to improve the acquisition and retention in the PMM platform, such as by content enrichment and push notifications for updated content especially to possible promising segments.