Skip to content

Improving Food Animal Welfare Begins with Quality Data

Share this article

By Daniel Foy & James Reynolds DVM, MPVM

Food animal welfare as a science is relatively novel and new, only spanning a few decades. For example, the University of British Columbia’s animal welfare program recently celebrated reaching 25 years in 2022, and the Dairy Cattle Welfare Council only reaches seven years in 2023. The methodology and tools (Including digital tools) commonly deployed and used for measuring on-farm food animal welfare have limitations, lack reliability, and are often subjective in nature. We find ourselves at an impasse on how to move forward in improving food animal welfare and in measuring or monitoring animal welfare. There are different societal ethics and what the consumer and industry, in general, expect from our food supply with respect to animal welfare, that has changed and grown over the last couple of decades.

The aim of food animal welfare audits is to certify that animals are managed and housed to certain welfare standards, and currently, they are used to certify that for the sale of resulting products that the consumers see on the shelf. Food animal welfare audits have been developed to ensure this; they have to date been built around the Five Freedoms of animal welfare: 1) freedom from hunger and thirst; 2) freedom from discomfort; 3) freedom from pain, injury, and disease; 4) freedom to express normal and natural behavior (e.g. accommodating for a chicken’s instinct to roost); 5) and freedom from fear and distress. The Five Freedoms approach is one that seeks to minimize bad animal welfare (lameness, injury, hunger, etc.). Animal welfare assessments and audits are transitioning to the Five Domains model, which relates to positive animal welfare experiences for animals. Information and data collected on-farm need to relate in real time to measurements associated with animal welfare that are useful to the farm and to audits.

How those audits and their associated data are used depends on who uses the audit, but there are three main buckets: 1) folks using audits as a means to mitigate risk in their value chain, 2) some folks use audits as a means to differentiate a brand, 3) and some use audits to actually improve food animal welfare. But not everybody does all three. Some do one, some two, some try to do three, and then some do none.

Within today’s on-farm welfare audits, the main measurements used are outcome (animal and facility observations) and protocol measurements. But are these the things we want to use so that the animals “tell us”, whether they are comfortable and can express normal behaviors? Are they happy? Are they healthy today?  We currently do a poor job of collecting this information and data accurately and consistently, and we have pitfalls and biases in current measurement methods (lameness, body condition score, cleanliness scores, etc.), as they can be subjective and manipulated by culling and other management practices. Often too much faith is put in the reliability of the results of an audit. There’s opportunity there because there’s got to be a way for us to have data available that helps reflect how we’re doing over a period of time, not just today or the day of the audit., There is also an opportunity in gaining efficiency for the auditor so we can spend more time using the information to improve welfare rather than just measuring it as a snapshot in time.

The limiting factor with audits is that they are typically based on a once-a-year visit by a trained auditor, but what happens in terms of welfare for the other 364 days in the year, can we measure reliably when we’re not looking, when we’re not there? There is inherent distrust in the data from audits that are mostly self-reported information, and unknown validity of the data derived from the current generation of AgriTech technology at the farm, with data being siloed in disparate locations, with unknown governance and editing, with many black box algorithms and SaaS bottleneck models, all being omnipresent.

The ongoing question is, with the current generation of AgriTech tools, are they collecting the right data points for monitoring and improving food animal welfare and the five freedoms, and are they offering the right models for the products they are supplying to the farm-level management team while enabling farm service suppliers to offer greater decision support to the farm? Improving the numbers (lameness scores, hygiene scores, mastitis rate, body condition, etc.) may not be the point. Improving lameness scores does not necessarily improve animal welfare. We want to keep cows from getting lame rather than just fixing lame cows. It’s no longer about “big data”; it’s about the right and accurate data for this welfare need and might not actually take many data metrics to meet the intended objectives.

Animal welfare is about the affective states or emotional states or what the animal feels and how they’re feeling, including animal welfare, animal health, and animal husbandry and behavior. So, while the concept is simple, the application requires changes. We may need to rethink the premise of whether it is possible to create a single algorithm that accurately fits all the situations and all the different farming systems that are out there with their regional variation and styles. We need to focus on the data rather than the widget, hardware, or ML/AI at this point. But, AgriTech has not currently delivered those models, tools, and strategies to the farm to account for these differences. The first-generation AgriTech companies provide tools designed to supply task-orientated proprietary data in cookie-cutter approaches that do not target animal welfare nor the different varieties, styles, and regions of food animal farms.

Unlike the Silicon Valley model of learning from failure, we in food animal agriculture can’t do that with food animals and their welfare. A more collective approach, between the farm and its service suppliers, in solving problems and challenges is needed. Applying ML and AI to quality, governed and owned data from the farm that aids in decision support for animal welfare management is important. Identifying the different data points of value that can be used for those decision-support tools on the farm for greater individual animal welfare is equally important. At the same time, being able to differentiate between a farm data issue and an AgriTech tool data issue is critical. The liability burden needs to be identified, as we can’t have a mistake with food animal welfare because animals may be harmed.  

On-farm digital literacy, and the farm itself, is a big part of the next generation of AgriTech.  Improving data collection, awareness, and quality is crucial to improving animal welfare. We now need to empower farm teams to become “citizen data scientists” focused on specific objectives and continuous improvement processes. As we naturally transition to the next generation of technology in food animal agriculture, it will be essential to collect consistent quality data on the individual animal and their affective and emotional states to better understand how animals are feeling. That gives confidence to the producer, processor, retailer, and ultimately the consumer in statements pertaining to best practices in relation to animal welfare.

The exciting parts of animal welfare are that we have better knowledge and understanding of animal welfare and developing technologies that will provide more objective and timely data to support on-farm decisions related to animal welfare while improving the reliability of animal welfare audits—having a person going to the farm once a year has limitations and confounding of the data (cows leaving the herd, culling, deaths). With different types of data capture and assured quality, we can start using the different data points together and apply appropriate AI/ ML applications for value and enhanced decision support at the farm and throughout the supply chain. Increased focus on promoting animal welfare establishes and maintains a trusted relationship between food animal agriculture and the consumer with respect to where their products come from.

To learn more about this topic from well-recognized industry professionals, please listen to Dr. Jim Reynolds, Dr. Jennifer Walker, and Daniel Foy discuss Data & Food Animal Welfare on the January 2023 episode of the LiveStack Podcast – here.

Reference Items: