<![CDATA[Malang Regency has many different types of land and coffee varieties in each sub-district. It spans from highland Arabica coffee in Poncokusumo, grown at 900 to 1,400 meters above sea level, to lowland Robusta coffee in Gedangan, grown at less than 300 meters above sea level. These differences make the green coffee beans have very different sensory profiles. However, the way people decide how to roast the beans is still mostly based on experience and cannot be easily shared. In this study, a seven-stage Soft Systems Methodology (SSM) was used to create a conceptual Decision Support System (DSS) framework that can recommend coffee roasting parameters. Data was collected through field observations and interviews with experienced roasters, farmers/farmer cooperative leaders, and staff from the Malang Regency Agriculture Office in all ten sub-districts. The SSM process resulted in several outputs, including a Rich Picture that shows six groups of actors and three main social and technical problems, a root definition using CATWOE that explains how unclassified green beans can be turned into roasting profile recommendations for each origin, a conceptual model with ten activities, and a gap analysis that found the biggest problems are the lack of formal roasting knowledge and the absence of a system to deliver recommendations to users. The proposed framework has a seven-layer DSS design that uses computer vision with edge computing to classify green bean quality and includes a knowledge base for each sub-district. It is made to work offline in rural areas with limited resources. The framework was tested using the Efficacy, Efficiency, Effectiveness, and Elegance (4Es) criteria. The results showed that SSM can be used as a structured and repeatable way to design agroindustrial DSS in tropical coffee-producing regions with many different stakeholders and limited infrastructure.]]>
