DPA implementation stages
As you already know, the Dairy Production Analytics service uses data from various automation systems already in the dairy farm. These are herd management systems, and software for managing milking parlors, feed production, rations, feed preparation, data from ERP systems on feed residues and veterinary components in the warehouse, staff work schedules, and so on. DPA not only displays aggregated data from various automation systems, but can also collate this data to find different insights related to the milk production process, to further optimize these processes and to increase their efficiency. The service needs this information, without it it simply cannot work. Therefore, the main limitation for the implementation of the service is either the absence of automation systems, or the presence of incomplete and inaccurate information in them. If we analyze inaccurate information, we, at least, will make wrong decisions, in some cases, we simply will not have enough information to make any decisions.
Therefore, the first step in introducing the service is an express audit of the farm and automation equipment. At this stage, as part of a survey of farm specialists, which is carried out by filling out a questionnaire, all the necessary characteristics of the farm, the presence or absence of automation tools, the quality of staff work with these automation tools, and so on are found out. The term of such an audit is usually two to three days, up to a week. If, according to the questionnaire, the farm meets the minimum requirements for implementing DPA, then we can proceed to the second stage.
The second stage is an audit of automation tools for the availability and quality of data. At this stage, the main source of data is the herd management system. If this is a system from the list of products compatible with the DPA service, then this data is uploaded, processed, and loaded into the DPA database. If this is a herd management system incompatible with the service, then our specialists develop a module for analyzing and loading data into DPA, usually it takes up to a month, after which we can receive data into DPA for analysis. After loading the data into the service, they are adapted, normalized and verified with primary sources, and then the service visualization tools are connected. At this stage, up to a dozen reports are already available, the analysis of the data in which allows you to make a retrospective report on the problems identified on the farm, and, sometimes, allows you to outline the first steps to eliminate them. The main task of this stage is to achieve a flow of complete and reliable information from the farm, to qualitatively enrich the data in the herd management system. For example, if the name “other, other, no data” appears as the main disease in animals, then, obviously, there is nothing to analyze in this case. Changing the business process of entering reliable primary information is the main goal of this stage. The duration of the second stage varies from a month to one and a half months, depending on the situation on the farm.
At the third stage, it is recommended to start gradually connecting additional data sources: milking parlor, feed production, rations, climate data, and so on. The connection of products compatible with the service is usually quick and does not cause any particular problems. If some new system is connected, then it takes time to develop data loading tools. At this stage, it is recommended to install various types of sensors inside the farm, a weather station outside, as well as sensors for the temperature of the silo, feed, and so on. These data reveal the picture of the impact of climate on the efficiency of the production process. This stage usually begins from the third month of service operation. The term for the implementation of the stage depends on the number of automation systems and the quality of the data in them, usually by six months from the start of the service, this stage ends.
The fourth step in the implementation of the service is the adaptation of the predictive analytics model to the given enterprise. Our team has developed several predictive models for different types of farms with tethered and loose animals, with different feeding options, with adaptation to the specifics of specific milking parlors, and so on. It is advisable to have data for at least six months, or better for a year, for applying the model. The system is trained on these data using machine learning methods (70% of the data goes for training, 30% is the verification of the forecast quality), and when the required forecasting quality is achieved, the system can make a forecast for a period from one year to two or more years. The highest quality forecast with an accuracy of more than 98% is made for a period of up to one and a half years, then the accuracy decreases. This is a rather resource-intensive stage, it is necessary in the case of working out various models of farm development, for more accurate contracting with feed suppliers and milk processors, building a logistics system, veterinary measures, forage procurement, and so on. At this stage, we have a full-fledged digital twin of the farm, in which we can not only see what will happen to the farm over time, but we can also perform virtual experiments with the herd, with feeding, and so on, in order to achieve the goals of increasing the efficiency of everything. production.
There is also a certain intermediate stage, it can be called the fifth, and the third and a half. This is the stage of prescriptive analytics. Prescriptive analytics is a technique for making decisions based on input data, bypassing a person. That is, the decision is worked out by the system, and, sometimes, it is executed by it. A simple example is the use of temperature and humidity sensors, in the event of an increase in the THI index on the farm in the summer, the system can send the necessary notifications to the staff, turn on ventilation equipment or water sprays to reduce the harmful effect of the climate inside the farm on animals. Or make adjustments to the ration if the service sees that tomorrow there will be a drop in ambient temperature, and the current feed balance will not be enough to meet production targets.
By offering a pilot implementation of the DPA service to our clients, which, in fact, includes the first two stages, we are already solving the issues of finding the negative aspects of the farm's work that affect its efficiency. The transition to subsequent stages will allow solving the entire set of tasks to increase efficiency, reduce the level of animal morbidity, reduce emissions into the environment, and increase the quality of products.
Increase the efficiency of your business - request a questionnaire and you will be already at the first step towards achieving your ultimate goal.
Contacte nuestros expertos para informaciones detalladas