Analysis of broiler data

26-02-2018 | | |
Only with perfect understanding of data variables essential answers on future development of the company can be found. Photo: Hans Prinsen
Only with perfect understanding of data variables essential answers on future development of the company can be found. Photo: Hans Prinsen

While working with customers’ broiler operations, it became clear to Aviagen that help was needed to interpret the large quantities of broiler data collected in order to improve performance.

The amount of information that comes out of a modern broiler house seemed a bit overwhelming, as some companies had many variables to deal with and thousands of flocks per year. The challenge was how to manage an operation that was so big and with so many variables that could influence the outcome. Only with perfect understanding of data variables essential answers on future development of the company can be found. Proven data can determine where to build new farms, what breed to use, what types of housing equipment, the feed type and nutrients as well as the optimal number of birds per drinker or per feeder.

In 2005 Aviagen, in conjunction with LIDM Software Systems started developing software to undertake the statistical analysis of broiler performance data. This proprietary software is capable of measuring the effect of multiple attributes that affect performance factors, such as: adjusted weight, adjusted feed conversion ratio (FCR), Livability and condemns, breast meat yield and the overall poultry efficiency factor.

Over the past 10 years, Aviagen has helped numerous companies address the actual impact of many variables on production, therefore improving customers’ decision-making process. Customer requests and interests varied, and we have been asked to investigate how factors such as house ventilation and house orientation impact performance.

Case definition

A company had broiler results for more than 3,500 flocks over two years. The company has three different operations each targeting a different market weight. The company used contract farmers who invested in their own farms, but the farmers control only management and farm equipment, while the company manages everything else (feed, breed, processing age etc.). How could that company define a list of best-to-worst growers to give bonuses?

After explaining what the analysis can do, we brainstormed with production employees to come up with a standardised questionnaire developed to be used by their field service personnel. The questionnaire was used to collect information about farm characteristics such as equipment type, number of feeders and drinkers, ventilation parameters, floor area (to calculate stocking density), nutrition information, and many other attributes that could affect production. This process provided an understanding of the concerns and questions management had about its production.

Data collection

A survey is important to get a true picture of the contents of each poultry house, as variables tend to change over time due to equipment maintenance. This effort is usually divided between all the flock supervisors and entered real time on electronic tablets. That way, as they do their routine farm visits, they have a chance to collect all pertinent information.

The broiler results for at least two years, to get the seasonality effect in the study, was then linked to each of the house attributes in preparation for the analysis. As part of the overall data movement, we also collected other data sources to combine with the survey and broiler results, such as weather data from meteorological stations near the farm location.

Foto: 3

Data analysis

The first step of any data analysis is to check the data for errors and properly understand what each data variable is showing so that we can perform the correct analysis. We need to understand if each data variable is a class variable (e.g. drinker type name, breed name) or a continuous variable (e.g. processing age, stocking density) and if the continuous variable has a linear or polynomial relationship with the production parameters. It is then that we can design our multiple regression analysis and run the program.

In one run we can check for outliers and remove them from the analysis. In the same run we can adjust FCR and weight to a common age, screen a large number of variables and calculate the impact of all significant factors on performance. For example, feeder A may have 2 pts FCR better than feeder B. Last but not least we can find a solution for the feed intake versus live weight challenge, as it uses a methodology to uncouple the relationship between feed intake and bird weight.


By including measurements of the farm quality such as feeder type and house design in the analysis, we can measure a farmer’s management skills with the equipment. This way is much more reliable than the normal ranking based on average farm performance, which may be misleading as it does not consider that a grower may have an unfair advantage due to luck with a particular flock (flock source age, sex, breed and processing age). The idea is to reward the best growers that do not necessarily have the best equipment, enabling them to invest in farm equipment and continue to produce well. In this particular case the customer’s goal was to grant bonuses to the best growers, independent of all other attributes, the customer began seeing an improved return on investment, as the proven best growers were then able to re-invest the bonus money in their chicken facilities.

The results are fascinating as they allow us to see exactly which attributes impact production and by how much, enabling management to focus on the ones that bring immediate improvement and ensure better decisions for the future.


When analyses are run but the output does not match the real-world conditions, or, perhaps worst of all, if the conclusions would work but sit unused, the analytics exercise has failed. As we believe advanced data analysis is essentially a business matter, at the final presentation we ask that top executives be part of the meeting to ensure they have a clear understanding of how to translate the results into action. To help in the process, specialists in the areas that have been found to have the biggest impact in production will be available to the executives. For example, if the analysis shows that birds coming from hatchery “A” is negatively impacting final results in production, our hatchery specialist will visit the location and report the findings.

A normal review section starts at the beginning of the week with production managers and an executive responsible for the area present. We then work with our client to address factors identified in the analysis. A summarised presentation and improvement plan is then given to the company’s executive team to create a clear picture of the overall production.

Join 31,000+ subscribers

Subscribe to our newsletter to stay updated about all the need-to-know content in the poultry sector, three times a week.