Why the Flock Itself Is Becoming the Next Thing to Measure in Broiler Production
Climate controllers report the air at a few fixed points. The birds report everything else, and for now only a person standing in the house can read them.

Climate controllers report the air at a few fixed points. The birds report everything else, and for now only a person standing in the house can read them.
The controller shows every value in range. Temperature, humidity and air exchange all read normal. And the birds say otherwise: huddled in one part of the house, gone from another, with the litter wet in patches.
I have seen this happen more than once on a broiler operation in Ukraine that raises more than a million birds a year. Each time, the person who caught the problem was an operator walking the flock. The fix was manual too. The team raised the temperature, increased air exchange and spread dry litter where it was needed.
This is a business problem before it is a technical one. Margins in broiler production are decided on feed conversion, on mortality and on carcass quality. Feed is the largest cost in the house, and a bird lost late has already eaten most of its share. A problem that waits for the next walk-through costs more with every hour it waits.
In an earlier article in this series I argued that what separates livestock operations is how fast a recorded number reaches someone who can act on it. This one goes a level down, to what gets recorded in the first place. Recent research points in one direction: the next gain will come from measuring the birds. Measuring the air around them more precisely will not be enough.
What the Controller Sees
On that operation, temperature sensors hang 40 to 50 centimeters above the birds' heads. A house has several of them, depending on its size, spaced evenly along its length so the controller gets a reading from each part of the building. Any grower who runs a controller and walks a house will recognize the arrangement. It is a sound layout. It is also a handful of points in a building full of birds.
A study from the University of São Paulo, published in Results in Engineering in December 2025, shows how much a single point can miss. The researchers placed three identical sensor units in a 135-meter commercial broiler house on day 36 of the cycle. One was fixed two meters above the litter. One was fixed at 30 centimeters. The third rode on a small robot at the same 30 centimeters and stopped at 35 points in each of the house's three sections.
In the first section the high sensor averaged 54.4% relative humidity and the low fixed sensor 57.7%. The robot, at the same height as the low sensor, averaged 73.6%. Litter moisture in that section was above 40%, against a recommended maximum of 35%. In the third section the robot and the low fixed sensor almost agreed, at 58.1% and 59.2%.
Height explained part of the gap. Location explained more of it. A fixed sensor answers for the spot where it hangs.
What the Operator Sees
Experienced operators read the flock on every walk. They look at how the birds are spread across the floor, then at the litter, then at the general condition of the birds. Of these, behavior and distribution are the most reliable.
They are also the one reading that, in most houses, no system records. It exists while someone is standing there, and it is gone when they leave.
It can be recorded, though. A group at the University of Oxford has spent more than a decade pointing ordinary cameras at commercial broiler flocks and measuring how the flock moves as a whole, without tracking any single bird. In a 2017 study in Veterinary Record, covering 50 flocks on one farm with 33,000 to 35,000 birds per house, how unevenly a flock moved at four days of age predicted how much footpad dermatitis and hock burn it would show at slaughter. Bodyweight and water consumption at the same age did not. A healthy flock moves alike. A flock heading for trouble has stragglers, and a camera can see them weeks before the plant does.
The Same Gap on the Scale
Weight has the same weakness. An operator catching birds by hand tends to pick the large, healthy, active ones. Weak birds get passed over on the reasoning that they will probably be culled anyway. The average that goes into the record then describes the best part of the flock.
This is a commercial point as much as a husbandry one. How uniform a flock is, and how close it comes to its target weight, both affect what it earns at the plant. A record built from the best birds hides exactly the spread that decides that result.
My rule is simple. A bird that should be culled gets culled. Every bird that stays in the house gets weighed. The scale is rarely the problem. The choice of which birds reach it is.
Where the New Tools Actually Stand
None of this is ready to buy, and that includes camera systems, the category my own company works in.
The São Paulo team calls its robot a proof of concept, and the paper is candid about why. The measurements cover one day in one house. The navigation model made the right decision 89% of the time in testing and 77.5% of the time among live birds. At the end of each feeder line a person turned the robot by remote control. It averaged 0.031 meters per second, which means more than 70 minutes to travel the length of the house once. The authors write that the system is "not yet ready for immediate large-scale deployment."
Cameras show the same drop between the lab and the house. A 2026 study in the journal Agriculture trained a bird-counting model on a large public video dataset and then pointed it at a ceiling camera in a Korean commercial house holding about 27,000 birds. In one frame with 281 birds it found 62. After the researchers added 311 annotated images from that same camera, it found 271, and counting accuracy reached 97.6%. The result holds for one camera in one house, and the authors say other farms and seasons still have to be tested.
Two recent reviews describe the field the same way. One, from the University of Debrecen in 2025, found the methods mainly under testing and wrote that information on how common such systems are on broiler farms "is not available in the literature." The other, from the University of Georgia in 2026, found most systems to be prototypes in research facilities, with only a few sold as commercial tools.
Growers reading trade coverage of the robot asked two blunt questions. Does it pick up dead birds? And who pays for it on a contract farm? Neither has an answer yet.
What to Do Before Buying Anything
Two steps cost nothing. Weigh every bird that comes to hand. And write down what the operator saw on each walk, in the same place every time, so the observations build into a record rather than a memory.
When a vendor does come, I would open with one question: what does this system show that the controller and the operator do not already see? A robot that carries climate sensors around the house draws a better map of the same air. That has value, but it is narrow. The larger gain is in recording what the operator reads on a walk, meaning how the flock is distributed, how it behaves and how uniform it is, and recording it through the hours when nobody is in the house.
The long-term direction is a machine that does the operator's rounds, including collecting mortality. On the published evidence that is years away.
Disclosure: I am the founder of FeatherOS, an early-stage venture developing operational monitoring technology for animal production. FeatherOS is in pilot validation and is not yet on the market. No product, platform or vendor is named or recommended in this article.
Sources
Balthazar, Silveira, Soares, Silva. RobôFrango: Proof of concept of a mobile biosensor robot for environmental monitoring in broiler houses. Results in Engineering, vol. 28, 2025.
Brassó, Komlósi, Várszegi. Modern Technologies for Improving Broiler Production and Welfare: A Review. Animals, vol. 15, 2025.
Dawkins, Roberts, Cain, Nickson, Donnelly. Early warning of footpad dermatitis and hockburn in broiler chicken flocks using optical flow, bodyweight and water consumption. Veterinary Record, vol. 180, 2017.
Dhungana, Paneru, Dahal, Chai. A Narrative Review on Internet of Things and Artificial Intelligence for Poultry Production. Animals, vol. 16, 2026.
Hwang et al. Combining Public and On-Farm Datasets for Deep Learning-Based Broiler Detection and Counting in a Commercial Broiler House. Agriculture, vol. 16, 2026.

Section Editor, Agribusiness & Industrial Operations
Dmytro Rozhko is an agribusiness executive and livestock industry expert with Ukrainian roots who now works in the United States, where he develops his own agricultural projects.