Gross feed conversion efficiency represents the quantity of Animal source food (ASF) produced in relation to the total quantity of feed used. This can be calculated per kg of product or by taking into account only the protein or energy components of the feed produced and consumed.

Improving gross feed conversion efficiency has been a major focus of research and development in animal production. Various levers such as husbandry, feeding, genetic selection, nutrition and animal health have been mobilised (Garnett et al., 2015), leading to significant improvements, particularly for poultry.

Ruminants, and in particular suckler cattle systems, which are less standardised than monogastric systems (Gerber et al., 2015), have highly variable gross efficiencies (GEs) by mass. For example, Wilkinson (2011) has shown that beef production systems in the UK consume between 7.5 (GE= 13%) and 27.5 kg (GE= 4%) of feed per kg of meat produced, depending on the type of animal and feed considered.

These differences are repeated when we look at gross energy efficiency. Laisse et al. (2018) calculated gross energy efficiencies of 25% for broilers, 26% for pork and only 4% for beef. The difference between broilers and other meat products increases when the protein fraction of feed is considered. In fact, gross protein efficiency reaches 54% for broilers versus 40-42% for pork and 8% for beef. This may cultivate the simplistic idea that replacing all beef with poultry would enable feeding for a greater number of people. In the United States, this would represent 116 million extra people being fed (Shepon et al., 2016).

Gross protein feed conversion efficiency reaches 27% for laying hen systems and 19-24% for dairy cow systems (Laisse et al., 2018).

Compared with monogastric animals, which use enzymatic digestion, ruminants are able to make better use of grass and other fibre-rich feedstuffs that are inedible by humans, thanks to the specific microbial predigestion that takes place in their rumen. Steinfeld et al. (1997) first proposed a 'human-edible feed conversion efficiency', estimating that, worldwide, animals use 1.4 times more human-edible feed than they produce ASF. Wilkinson (2011) formalised the 'edible feed conversion ratio' indicator and described the performance of a variety of livestock production systems in the UK. Net feed conversion efficiency was proposed later (Ertl et al., 2015; Laisse et al., 2018) and is referenced in this paper. It represents the inverse of the indicator presented by Wilkinson (2011); the amount of ASF produced is divided by the amount of human edible feed used.

To calculate this ratio, the human edible fraction of each food is estimated as the proportion of the product that can currently be valued as food for humans (Figure 2). Since food crops are split into different fractions (for example, milled wheat grains are separated into flour, wheat bran and gluten), the amounts of protein and energy directly edible by humans represent the weighted proportions of protein and energy found in each of the fractions edible to humans (Figure 2). For many animal feeds, the energy and protein fractions consumable by humans are very similar; however, some have more significant differences. For example, maize is largely valued for its starch while the protein-rich by-products (corn gluten feed and corn gluten meal) can currently not be eaten by humans (Laisse et al., 2018).

CP: Crude protein.

In order to represent the difference in protein quality and digestibility, the Digestibility of Indispensable Amino Acids Score (DIAAS) has been proposed by the FAO (2013) as a reference method. This score represents the content of the first limiting indispensable amino acid in the protein tested versus the content of the same amino acid in a reference protein corresponding to the needs of a child aged six months to three years and based on the actual ileal digestibility of indispensable amino acids (Rutherfurd et al., 2015).

Initially, it was suggested that DIAAS values should be truncated at 100% because the higher values represent a surplus in relation to human nutritional requirements and are therefore not valued by the human body if the food is the only component of the plate (FAO & WHO, 1991). Later, it was argued that non-truncated values should be considered because, in mixed diets, a high-quality protein-based food can supplement another food that is deficient in essential amino acids (Rutherfurd et al., 2015). The FAO (2013) suggests using the amino acid requirements for a child aged six months to three years as the reference protein. Taking DIAAS into account when calculating net efficiency multiplies net efficiency values by 1.7 to 2.4 for milk production, 1.6 for egg production and 1.4 to 1.9 for meat production (Ertl et al., 2016a, 2016b; Laisse et al., 2018).

The efficiencies obtained from the different sources, according to the different variants presented above, are summarised in Figure 3. Figure 3B depicts the net efficiency of systems with a double beef herd as a function of the proportion of suckler and dairy cows. The net protein efficiencies for the different types of beef herd (Figure 3C) and dairy herd (Figure 3D) are also illustrated.

A: Protein (top), energy (bottom), gross (left), net (centre) and DIAAS (right, for protein) efficiencies of beef systems. The data comes from different sources (colours in the legend). The y-axis is on a logarithmic scale.

B: Net protein efficiency, with protein quality taken into account for Walloon beef farms (data source: DAEA) as a function of the number of dairy and suckler cows.

C: Net protein efficiency for different types and regions of suckler farms.

D: Net protein efficiency for different types of dairy farm.

In order to gain better understanding of the variability of the performances observed, a summary of the correlations between management parameters and efficiency, land use and productivity indicators (calculated separately for the different databases used) is set out in Table 3. Graphs of the variables resulting from the principal component analysis are presented in Figure 6. Gross efficiencies (protein and energy) are strongly correlated with productivity per cow for dairy farms and weight gain per LU for meat farms, and therefore, in both cases, with the proportion of maize in the ration. Gross efficiency is also negatively correlated with age at first calving and calving-to-calving interval. In fact, the lower the age at first calving and the shorter the calving-to-calving interval, the lower the unproductive, and therefore inefficient periods for the animals. Net efficiencies (protein and energy) are negatively correlated with the use of concentrates (per litre or per kg of live weight) and the proportion of maize in the ration but positively correlated with the proportion of pasture for dairy farms. These findings support the observations made by Laisse et al. (2018) regarding a less productive grass-fed dairy cattle system but with higher net efficiency than the maize-based dairy system.

Land use, where we are aiming for a low score and therefore correlations to be interpreted in the opposite way, is correlated with age at first calving and negatively correlated with milk and meat productivity. In contrast, the use of arable land is positively correlated with the use of maize and concentrates.

Net productivity is positively correlated with milk productivity and grass yield but negatively correlated with age at first calving and the use of maize and, in the case of beef farms, concentrates. Battheu-Noirfalise et al. (2023) also observed a correlation with the protein content of concentrates used in dairy systems.

For dairy farms, net productivity is negatively correlated with arable land use and correlated with net efficiency as observed by Battheu-Noirfalise et al. (2023). For suckler farms, net productivity is strongly correlated with net efficiency and both variables are negatively correlated with arable land use.

The interpretation of indicators such as net efficiency can be simple. In this case, a farming system is considered to be a net protein producer if it has an efficiency greater than one. However, this indicator provides only partial information on the system's performance. For example, having an efficiency level much higher than one does not always imply better contribution to food security than a lower value since this indicator does not take into account the level of livestock production per unit area, which can be very low. The net productivity indicator, which factors in production and surface area used, is not subject to this shortcoming. A net productivity of 300 kg of protein per hectare that does not compete with humans will, by definition, contribute more to protein production for humans than a farm producing 100 kg of protein per hectare. However, the levels achieved for this indicator also need to be put into perspective. The soil and climate context largely determines production potential, and depending on how it is implemented (as a basis for comparison), it does not necessarily supply information on how far systems can progress. By comparing it with optimal crop production, the LUR theoretically takes better account of the margins for progress than other indicators. However, the yield assumptions used to calculate the LUR remain theoretical. For the LUR, comparing systems to an optimum yield also raises questions. Is there not another optimum than the all-plant system? What about animal alternatives or even mixed crop-livestock systems that aim to optimise animal-plant synergies and maximise food production for humans on the same land?

The indicators described in this contribution aim to quantify the efficiency and production levels achieved on the basis of plant and soil resources. However, many other criteria currently need to be taken into account to meet current challenges, such as environmental impacts and/or services or economic and social performance. In particular, methane emissions are another major criticism of ruminant livestock farming. Solutions therefore need to be found to combine food production with a reduction in greenhouse gas (GHG) emissions. Ineichen et al. (2024) proposed combining net production and GHG emissions in a single indicator. For 87 Swiss farms, net protein production per kg CO₂ eq averaged 16.8 g crude protein/kg CO₂ eq and was strongly negatively correlated with the use of concentrates.

In addition, based on modelling, Mertens et al. (2023) and Kearney et al. (2022) evaluated sustainability indicators, including profitability, methane emissions and net efficiency of meat production from dairy calves, demonstrating the interest of these systems from an environmental point of view and the compromises to be found between profitability and use of arable land.

However, there is no guarantee that the various indicators will converge towards values that are favourable to the various sustainability criteria and adjustments will certainly have to be made in order to define the development paths to be implemented. Such an exercise has already been carried out with a multi-stakeholder panel using a decision support tool (Battheu-Noirfalise et al., 2024a) for dairy systems.

This summary brings together the many indicators and large datasets used to characterise current ruminant farming systems and their capacity to act as net producers of food for human consumption. The performance of the systems is analysed as a function of the type of livestock, crop rotation and management parameters. The study shows that there is room for improvement, specifically by basing livestock farming on grass and reducing the use of maize and concentrates. This summary, which focuses on France and a number of neighbouring countries as well as other international articles, demonstrates the value of ruminant livestock systems for making the most of biomass that is inedible by humans. In the future, when demand for food is set to increase and plant resources are likely to be mobilised for other purposes (e.g. energy, fibre, etc.), the key issues affecting ruminants will focus on optimising the use of food that cannot be consumed by humans and land that cannot be farmed, without overlooking the other services provided by livestock farming. This means rethinking the place and optimal practices of livestock farming in agro-ecosystems.

Alexandre Mertens: conceptualisation, data curation, visualisation, editing - original version; Caroline Battheu-Noirfalise: conceptualisation, resources, validation, editing - original version; Pauline Madrange: editing - revision and correction; Michaël Mathot: conceptualisation, editing - revision and correction; Alice Berchoux: resources; René Baumont: conceptualisation, resources, editing - revision and correction.

The authors would like to thank the partners of the AUTOPROT, Sustainbeef and ERADAL projects, the GIS Élevage Demain and Avenir Élevage and the Inosys breeding networks for sharing their data. This research is partially funded by the Fonds pour la formation à la recherche dans l'industrie et dans l'agriculture (FRIA).

This article was first translated with www.DeepL.com/Translator and the authors thank ELAN Languages for English language editing.