What a feed nutritionist taught us about our own aflatoxin pitch, and the claim we dropped because of it.
Notes from our conversations with cattle feed plants on why the mycotoxin argument we had been making was the wrong one, and what we say instead now.
We went into these conversations ready to argue that less aflatoxin means more milk. We came out having dropped the claim entirely.
That is not a comfortable thing to publish. But the pitch we walked in with does not survive a technical question, and we would rather lose it here than in someone’s meeting room. What replaced it is narrower, better evidenced, and considerably harder to knock down. This post is about the difference between the two.
The claim we walked in with
Almost every company selling quality technology into animal feed reaches for the same argument. Your maize carries aflatoxin. Aflatoxin harms the cow. Remove it and you will get more milk.
It is intuitive, it fits on one slide, and a good dairy nutritionist will take it apart in about a minute.
Controlled feeding trials that dose cattle with purified aflatoxin B1 have repeatedly failed to find an effect on milk yield, dry matter intake or feed efficiency. Not at 100 µg/kg. Not at 105. In one well-known study, not even at 471 µg/kg, which is roughly ninety times the European limit for dairy compound feed. The rumen and the liver handle the pure molecule considerably better than our slide assumed.
So the first thing we learned was to stop saying it. Aflatoxin is a real problem, but it is not the one that shows up on a milk recording sheet.
What the trials actually say
Here is the part that surprised us.
When Gallo and colleagues fed dairy cows maize that had been naturally infected in the field, intake and feed efficiency both fell. The aflatoxin B1 concentration in that diet was 17.53 µg/kg. That is a twenty-seventh of the dose that did nothing in purified form.
The same pattern turns up inside a single experiment. Applebaum’s group fed 471 µg/kg of pure AFB1 and measured no effect on milk production. They then fed 583 µg/kg of impure aflatoxin, meaning the toxin plus the other metabolites left behind by culturing Aspergillus parasiticus. Milk production fell. Same herd, same protocol, different material.
Mould never travels alone
The reason field-infected maize does damage that the pure molecule does not is that a mouldy kernel never delivers one thing. Three losses arrive together, and only one of them is a toxin.
Not one of those three is measured by an aflatoxin assay. That was the second thing we learned.
What it costs at the mixer
Feed efficiency in a dairy herd is energy-corrected milk per kilogram of dry matter consumed. Extension benchmarks put a well-run herd at or above 1.4, with the practical range running roughly 1.3 to 1.8 depending on stage of lactation.
Feeding trials on mould-contaminated maize have recorded that ratio falling from 1.46 to 1.18, a 19% loss of conversion. The cow eats the same weight of feed and returns a fifth less milk energy for it. Nothing about that shows up in a ppb reading.
Why we now say 3%, not 19%
This is where most vendor arithmetic goes wrong, including ours.
A feed plant does not run on mouldy maize. It runs on a blend, and most of that blend is fine. The 19% figure is the penalty on a batch made with tail-end lots. It is not a uniform uplift waiting to be unlocked on feed that was already clean. Multiply it across a full year of production and you produce a number that will not survive contact with a plant manager who knows his own intake data.
The honest arithmetic is a share, not a rate. If roughly one production run in six during a difficult monsoon procurement window is made with tail-end lots, and each of those loses 19% of its conversion, the annualised effect lands close to 3%. That is the number we are willing to defend. It comes entirely from cutting the bad tail out of the batch, not from making good maize better.
| Share of runs using tail-end lots | Penalty on those runs | Annualised gain if the tail is removed | When this looks like your year |
|---|---|---|---|
| 5% | 19% | 0.95% | Clean season, dry harvest, short storage |
| 10% | 19% | 1.9% | Normal year, mixed sourcing geographies |
| 16% | 19% | 3.0% | Our default assumption: unseasonal rain at harvest |
| 25% | 19% | 4.8% | Delayed harvest, wet godown, distress procurement |
Illustrative model. The only column a screening programme changes is the first one, meaning how many tail-end lots reach the mixer. The penalty column is a property of the grain, not of the technology.
We are not claiming that screening maize brings aflatoxin M1 in milk within limits. It helps, but that is a regulatory question answered by a laboratory, on a sampled consignment, against a legal threshold. We are not claiming a 19% uplift on your existing feed. And we are not claiming this replaces your lab. A vendor who tells you otherwise is describing a product that does not exist.
Two risks, two instruments
The distinction that opened this post is not a caveat. It turned out to be the whole operating logic.
Where we think RootsGoods fits
We came away with a fairly clear picture of where objective per-lot assessment slots into a feed operation, and where it should not try to do everything at once.
| Stage | What is hard today | What we can do | Why it matters |
|---|---|---|---|
| Farm gate | Fungal damage is judged by eye, differently by every buyer. | A per-lot score from a photograph, on a common scale. | Farmers get paid for clean grain instead of averaged down with everyone else. |
| Procurement centre | Moisture decides what moulds later, and nobody measures it per lot. | Internal moisture plus fungal damage before the lot is accepted. | The tail can be dried, discounted or re-routed while there is still time. |
| Godown | Clean-looking grain deteriorates silently in storage. | Shelf-life flags on the lots most likely to go. | First-out sequencing based on risk rather than arrival date. |
| Feed plant intake | Least-cost formulation assumes book values for maize. | A batch-level quality record feeding the formulation. | Energy assumptions that match the grain actually in the bin. |
What this looks like at co-operative scale
Take a large state dairy co-operative. We will call it XYZ Milk Co-operative, because the point is the structure rather than the name. Nine cattle feed plants, around 1.12 lakh tonnes of feed a month, sold to member farmers through district unions.
That structure is why the argument lands differently at a co-operative than at a private feed company. The feed and the milk sit on the same balance sheet. A conversion gain that shows up in a member’s shed comes back as procurement volume. In a private feed business it would be somebody else’s benefit entirely, which is exactly why nobody there has much reason to pay for it.
| Basis | One plant | All nine plants | Assumption |
|---|---|---|---|
| Feed produced | ~1.49 lakh MT/yr | ~13.4 lakh MT/yr | 1.12 lakh MT/month across nine plants |
| Feed value | ~₹373 Cr | ~₹3,360 Cr | At ₹25,000 per MT |
| Value of a 3% conversion gain | ~₹11 Cr | ~₹100 Cr | Feed-equivalent floor, not a milk-revenue estimate |
| Screening cost | Per lot, at the gate | Scales with lots, not tonnes | No incremental laboratory capacity required |
Stated at the conservative floor, meaning the same milk from less feed. Every row is an assumption to be replaced with real numbers during a pilot. The point of the table is the shape of the opportunity, not the last digit.
Where we go from here
None of this should be taken on faith, and we would not ask a nutritionist to take it on faith. The claim is falsifiable, which is the best thing about it.
Run it on one plant, through one monsoon procurement window. Screen every arriving lot. Split the worst-scoring lots out of the dairy stream instead of blending them in. Then compare conversion on the batches made before and after. If the tail was not costing anything, the data will say so inside a season, and it will have cost a set of photographs.
We would rather run that test than win the argument. Aflatoxin is what a regulator measures. Mould is what a cow notices. Only one of them is visible at the procurement centre, and it happens to be the one that moves the conversion ratio.
Sources
- Ogunade et al., ‘Aflatoxin in Dairy Cows: Toxicity, Occurrence in Feedstuffs and Milk and Dietary Mitigation Strategies’, Toxins 13(4):283, 2021. Source
- Ohio State University Extension, ‘Molds and Mycotoxins’, Buckeye Dairy News. Source
- DAIReXNET, ‘Mold and Mycotoxin Issues in Dairy Cattle: Effects, Prevention and Treatment’. Source
- Hutjens, M., ‘Dairy Efficiency and Dry Matter Intake’, University of Illinois. Source
- RootsGoods, ‘Digestibility by Design: How Genetics and Quality Verification Optimize Ruminant and Swine Feed Efficiency’. Source
- RootsGoods, ‘The Consignment that Costs a Crore: Aflatoxin in Maize’. Source
Data notes
Anonymisation. XYZ Milk Co-operative is an illustrative name. The capacity figures used are the publicly reported figures of a real Indian dairy co-operative of that scale, retained because the arithmetic only means something at a real operating size.
Caveat. The sensitivity and co-operative-scale tables are illustrative models, not forecasts. They apply a single stated conversion penalty across varying tail shares, and value the result at feed cost rather than milk revenue.
Publication. The 1.46 to 1.18 feed efficiency figures should be traced to their primary trial citation before external publication. The pure-versus-natural aflatoxin evidence and the 1.4 benchmark are sourced above.