Using predictive analytics to improve hatchery performance

Hatcheries have pivotal roles in poultry production systems. The data generated in this process can be better utilized with predictive analytic techniques to address common issues that affect hatchability and hatchling quality by supporting planning, preventive maintenance programs, and personnel interventions.

We will discuss the current status of data management and analysis in hatcheries, challenges and opportunities, tools available for improvement, and call attention to improve training necessary to apply advanced data analytics.

There is a significant variation in the technological level of hatcheries around the world. Despite these differences, the hatchery is one of the poultry chain segments with more control and data generation.

Hatchery data comes from monitoring egg and supply inventories, equipment operation, breeder reproductive performance, and incubation outcomes

Most frequent hatchery data includes:

• Egg flows, fertility, hatchability
• Cause of embryo mortality obtained in egg breakouts
• Hatchling output
• Quality grades
• And records of the average machine and room temperature
• Humidity, and pressures

The digitalization of hand-written records is still tricky in many hatcheries, and typing errors are still common. Data quality always requires close attention.

• The preferred method of data warehousing is Excel files with a single table or multiple tables or spreadsheets.
• Data should always be organized in continuous columns and rows.
• Unfortunately, spaces are frequently left intentionally to add notes, complicating the data analysis.

Identifying hatcheries, farms, breeder flocks, machines, and other descriptive variables should be consistent across the time to make an accurate analysis.

Some variables like egg storage time are often not well defined, and approximated ranges are entered variably.

The data analyses are most of the time conducted in Excel.

• • Visualization of the data in a graph is critical to detect patterns and unusual extreme values.
  • This is the step were typos and outliers should be seen and corrected.
  • Our experience with data collected worldwide shows that plots are included in these Excel files, but obvious typos are not always investigated or fixed.
  • Sorting and filtering the data may also help determine possible unexpected values.