Coffee is one of the most popular drinks for many people. Out of every three people, one is a coffee drinker. Coffee is really delicious if you drink it either in the morning or at night when work is piling up. PT. Mulya Sari Mandiri is a coffee factory that produces coffee, both dry coffee beans and coffee powder. The coffee obtained by this company comes from various regions such as Sumatra and Java. The coffee beans that enter from the sending area are then selected to get the best coffee. Many techniques and methods are applied in the forecasting process. In solving a problem in research, of course the researcher must have a way or a method that will be applied in solving the problem so that the research carried out can be completed properly and in accordance with the expected results. The research method is carried out to look for something systematically by using the scientific method and applicable sources. In the process of this research aimed at the PT. Mulya Sari Mandiri, especially in the area of predicting the number of coffee export needs by providing more meaningful results. The result description is an illustration of what the artificial neural network system looks like in predicting a data number of export requirements using the Backpropagation method. An overview of the results is usually made in the form of user interface design or interface design, a good design is directly proportional to the quality of the program, when the interface design is good, the quality of the program will also be good. From the results of research conducted at PT. Mulya Sari Mandiri greatly adds knowledge and insight, by collecting data related to the prediction of the amount of need for export coffee, From the input learning rate and maximum epoch, the Backpropagation method can be carried out with training or training with convergent results, namely with or target error achieved with iteration 3160 the training process (time) is 0.00.25 seconds, with a permomance value of 0.0999 for Arabica coffee and for the robusta type with iteration 4957 the training process (time) is 0.00.07 seconds, with a permomance value of 0.2. From the data on the number of export coffees used as training, training targets and test data, the Backpropagtion method can be recognized and produces predictions that the number of Arabica coffee will experience a total decrease from 2021 data of 1481 with predicted results of 1410 tons.