In recent years, Japanese farmers are gradually increasing their use of direct sales channels to deliver agricultural products directly to consumers to increase their profits. In addition, with the widespread use of the Internet, e-commerce is being put into practice in agriculture. However, the share of direct sales in the total volume of agricultural products shipped is still very small. At present, most of the farmers send their harvested agricultural products to Japan Agricultural Cooperative in order to focus on cultivation activities and ensure stable income. On the other hand, many substandard agricultural products are returned to the farmers due to the strict purchasing standards set by Japan Agricultural Cooperative, especially for fruits such as peaches, the color, sweetness, shape, etc. are strictly controlled by sorting machines and visual inspection. As a result, problems such as decreased income for agricultural producers and increased food loss due to waste disposal are on the rise. Selling substandard agricultural products directly to consumers through D2C platforms is considered a good approach to solve such problems. However, the variation in the condition of substandard products is quite large. Packing based on individual experience results in a significant difference in product value per package, which can lead to customer dissatisfaction. In this study, we focus on the direct sale of substandard peaches, assuming that attributes such as color and fruit spots are quantified by new sorting machines. To improve customer satisfaction, a decision-making model for packing peaches that maximizes the value of each package while minimizing the difference in product value between packages is proposed. The effectiveness of the proposed algorithm is confirmed under different distribution of quantity and quality.