Purpose: An evidence-based approach to improving instructional practices and student outcomes in data use. It is a systematic process of evaluating and analysing learning problems, collecting and transforming various types of data into instructional decisions, and implementing informed actions to improve instruction and student learning. Since teachers are the main actors in instructional practices, this article reports on a study aimed at predicting the influence of various teachers' characteristics on the degree of data use practices for instructional purposes. Design/methodology/approach: In this study, we conducted a survey in a developing country to gather primary data. The collected data were analysed using a supervised machine learning approach, focussing on decision tree analysis, to determine the influential factors. Findings: Our investigation identifies pedagogical knowledge, data literacy, content knowledge, knowledge of English for teaching and attitudes towards data as crucial determinants in predicting the intensity of such data use practices. Notably, pedagogical knowledge emerges as the most potent predictor, emphasising its pivotal role in shaping teachers' frequency of instructional data use practices. Surprisingly, English proficiency does not exhibit a significant influence in this predictive model. Research limitations/implications: The findings may not be generalisable to a wider context since this study relied on a relatively small teacher self-reported sample collected through surveys, and, as this study used perception data, this may or may not reflect teachers' actual knowledge and skills. Practical implications: By spotlighting the nuanced interplay between teacher individual characteristics and the practice of data use for instructional improvement, this research contributes to a nuanced understanding of the factors shaping teachers' engagement with data. Ultimately, it provides a foundation for targeted interventions and strategies aimed at fostering a culture of evidence-based practices to improve instruction and consequently student learning outcomes within educational settings. Social implications: This insight holds significant implications for policymakers, educational practitioners and providers of professional development programmes seeking to facilitate effective data use practices for instructional improvement. Originality/value: By spotlighting the nuanced interplay between teacher individual characteristics and the practice of data use for instructional purposes, this research contributes to a nuanced understanding of the factors shaping teachers' engagement with data. This study represents an empirical examination of such factors by employing a quantitative approach: a flexible decision tree analysis. This contributes to a growing body of research on factors related to teacher characteristics and much of the research in the field of data use has been done using a qualitative approach.