We are pleased to announce the publication of our new open-access paper in Frontiers in Veterinary Science:
“Optimising the selection of welfare indicators in farm animals” by Jon Day, Mohamed Ben Haddou, Rita Kylling, Guro Vasdal and Heleen van de Weerd. Access the full text here
Context
Today’s farm-animal welfare assessment frameworks are challenged by the sheer number of potential indicators (hazards → welfare consequences → indicators) and the practical constraints of using many indicators in real-world settings.
Our work addresses this by developing algorithmic methods so that users (researchers, auditors, policy makers, industry) can select an optimal subset of welfare indicators that balance:
- the breadth of welfare coverage (how many hazards/consequences are captured)
- the impact of those welfare consequences
- the ease of mitigation of the hazard
- the feasibility of measuring the indicator on-farm.
What we did
We built a structured database of 382 unique welfare indicators across seven farm-animal species (dairy cattle, dairy calves, beef cattle, pigs, broiler chickens, laying hens, sheep).
We assigned metadata to each indicator: how many unique welfare consequences it links to (Coverage); Impact of those consequences; Ease of hazard mitigation; Ease of use of the indicator.
We developed and contrasted two algorithms:
- A “greedy” algorithm which selects top indicators by simple ranking.
- An enhanced algorithm using branch-and-bound and backtracking (via OR‑Tools + SCIP) to identify globally optimal combinations under constraints
We tested with case-studies: for broiler chickens, and for growing/finishing pigs, showing how the enhanced algorithm gives better trade-offs and avoids the selection of redundant indicators.
What have concluded?
You don’t necessarily need a large number of indicators to capture the relevant welfare hazards/consequences; smarter selection can achieve equivalent coverage with fewer indicators.
The enhanced method avoids the “plateaus” seen when simply adding the top indicators one by one (greedy approach) and thereby increases efficiency and clarity of indicator selection.
The method is data-agnostic and flexible — you can tailor the weighting of priorities depending on your context (industry audit, certification scheme, policy monitoring, research).
Implications
If you are working in animal welfare monitoring—whether on-farm assessments, certification schemes, supply-chain audits, or policy evaluation—this approach offers a way to:
streamline your indicator set without compromising scientific robustness
make your monitoring more feasible and cost-effective
align your measurement programme with your priorities (e.g., focus on high-impact welfare consequences, or ease of implementation)
enhance transparency and credibility in welfare reporting.
Next steps
We are now progressing to explore stakeholder engagement and weighting processes (how one decides which criteria matter most), and further scenario-testing to ensure robustness across more species, systems and data sets.
We invite you to read the full paper and we welcome your comments, queries or collaboration ideas.