[Music] all right thank you to pair Andre and this year for inviting me to talk to you my talks gonna be a little bit different to what we've seen thus far a little bit more philosophical and a little bit more cynical I have I think I want to start by saying and reminding you that the opposite of variability is uniformity and uniformity is one of the previous speakers have said is really what our objective as nutritionists is we want to produce feed that meets our specification hopefully with a consistent ingredient composition we don't want to chop and change too much this allows us to achieve adequate feed intakes remember the birds suffer from mera phobia anything that changes they don't like so they stop eating it so we want to keep our feeds consistent and hopefully we can produce uniform carcasses and consistent product so that's our objective as commercial nutritionists as I've said consistent feed stimulates feed intake because you avoid feed refusal this improves performance and it makes any predictions that we may wish to make a lot more accurate so we at least know that we're going to achieve the feed-in takes the FC ours and the target masses we set out to do I think importantly that if you perform better as a broiler producer you become more sustainable and I don't this isn't to talk about sustainability but you need to remember that sustainability has different components to it and people tend to forget this there's an environmental leg we want to reduce our usage of of land to grow soybean meal for example we want to reduce our pollution from nitrogen these are the environmental components is a well-being opponent which includes both human well-being and animal well-being and somewhat cynically some Australian workers have shown that most humans are concerned concerned about their own well-being and then they worry about the animals because Jamie Oliver and all the other celebrity chefs have told them if they eat organic chicken they're going to be healthier okay so they worried about that they're not worried about the chicken and then of course all of those of us in business know that if we are not financially sustainable we don't have a business what tends to happen cynically an aside is that each of us worry about our own particular component of the sustainability debate so you'll find the welfarist s-- completely forget about the environment and the environmentalists can completely forget about the money and that's many people we four completely forget about everything else so we need a bit of a balance here but the good news is if you do your your your nutrition correctly you actually impact on all three areas from a variability point of view as a nutritionist there are four major areas that are of concern probably the simplest and most important is measuring your moisture content in your diet and then doing something about it okay secondly we need to look at protein and I should have put digest digestible amino acid levels in there and we've made really good strides with this in the last while particularly with NIR systems I think the aspect that is probably the most challenging is determining the dietary energy levels and I don't know any any of you read Gonzalo mitosis paper in applied poultry research was published towards the end of last year and in there he said you can't I'm paraphrasing him you can't believe it that it's 2018 and we still can't determine ami accurately it's in his paper it's a it's really better written than that I've just said and then we need to be aware of mineral levels and Bob has just told us about guys who add calcium to the soybean meal to make it flow better in the ships I assume Nestor is in the ships and the silos so these are things that we all need to be in very much aware of we talked about it briefly ingredient form handling and the degree of processing all of these things impact on our variability and the other speakers have done a very good job of explaining this already I think that it's important that we realize that our requirements for our Birds and our feed intakes vary dramatically between different production systems and this is both at a flock and an individual level so an open-sided house in Zimbabwe with an earthen floor you're not going to get the same kind of feed intakes that you're going to get in a controlled environment House component so these things we need to consider as nutritionists but I think the the point I'm trying to get here to here and and and Bob alluded to this is that the benefits of over supplying Neutrik nutrients is far outweighed by the penalties that are imposed by the under supply so we need to be very very careful over over supply costs us money but and the supply costs us a lot more and we see a lot of poor performance at the bays a PSS meeting in Sydney it was a fantastic paper that made me think about things in a slightly different way and unfortunately the the topic was the feeding of slow drying broilers and I immediately think birds growing to 2.6 kilos in 56 days and in fact that's not what was about at all it was about slow growing individuals in a normal flock of birds and what they found is that slow growing birds concern a consume surplice amounts of essential amino acid and fast growing birds over consume non-essential amino acids so within one flock you've got some birds of eating a adequate amount of essential amino acids and and so on and it was quite interesting that they found that given the choice when they gave a birds a choice of feed slowly growing Birds consumed diets rich in non-essential amino acids as opposed to essential amino acids who would have thought that that would have been possible so we feeding these flocks of mixed birds and within that flock the birds have got different different characteristics I think the most interesting thing in that paper was that they found that slow growing birds were able to regulate gene expression for alanine catabolism so they could actually identify slow growing birds by measuring this gene expression and at the same time the uptake of the non-essential gene that genes were down regulated so we think we feeding a flock of birds or we feeding an average bird but in every flock they're just such big differences that it makes the variability that we see in ingredients quite small okay so what do we want to do is commercial nutritionists in the perfect world in the perfect world the NIR results that we achieve in our feed mills would be perfect okay and we would be able to use a system like peony to analyze each parcel of in of incoming ingredient and by that I mean every half a truckload or something each parcel because that's how big the variability is we would be able to preserve the identity of those ingredients some way and and then possibly formulate our diets on a real-time basis on-the-fly and I know that the idea of putting an N I are feeding into the into the way hoppers above feed mills is great gaining traction so we can formulate on a real-time basis now and we would like to eliminate all wane mixing and cross contamination issues and then of course we would like to feed average birds which don't exist so this is the perfect world in reality what happens is that an NIR is only as good as the calibration used to build it put it together okay most of us do not check enough incoming material at all we just don't do it and I know companies that got in ours that hardly use them wonder why okay we don't preserve the identity of these ingredients very well we've got a a silo that's got soybean meal in it and when those trucks start coming in it just goes into the silo there's no other option where are you going to put it so that's a problem then the famous one is that what we do is we formulating using an average for our ingredients really how done can we be but that's the only way we can do it there limitations to formulation and then we formulate assuming that all our parameters are both linear and additive and we know very well that fat for example is not linear if you use a little bit of fat you have a huge extra caloric effect effect if you use too much oil you have a depression and energy it's not linear but we formulate on a linear basis this is a real problem okay then what we do we're really clever we round our formulations so that they easier to manage in the feedmill doesn't make a lot of sense most of us make weighing and mixing errors you just don't need to run your mixer long enough and you've got a mixed thing you're right ok we incur milling losses I accept that a lot of that milling losses moisture but do we ever measure whether that dust that is going out of our mill is carbohydrate or starch I bet you you don't I don't I've never done it never thought of it alright then most of my clever customers clients use this famous thing called mixer meal we just put 5% mixer merely that's the feed that's come back or out of stock or out of out of spec you know that stuff that accumulates in your feed mill I see people looking at me they don't know what mixer Milla's returns returns to the feed mill and then we put it into our feed great really clever when you're trying to reduce your variability and then of course some clown will always deliver the wrong feed to the wrong farm it only happens in Africa doesn't it you know we really do our best to mess it up which has me worried sometimes when you start looking at variability in feed and you start finding some things that you don't really like the look of and the first is that ingredient variability accounts for somewhere between 20 and 25% of total product finished variability or finished product variability this is increased by about 80 to 85% when you start worrying about things like weighing errors the batch size sampling and analytical errors and so on so it's not just about the incoming ingredients I put there that formula formulation errors could and should be added and I'm sure that Franco you're gonna deal with that more tomorrow but we forget about the errors we make as nutritionists I think that the important point is we need to remember though that variability is not additive its multiplicative it's not two plus two equals what it's not three plus three equals six or three times three equals nine okay and very small changes in one parameter can make a large difference now I'm all about this long and hard method how can I I Ella straight this geographically so I hope this works so what we want is a high degree of uniformity we need to consider our ingredients are weighing errors and our mixing and formulation for example when we have ingredient variability and I've just put crude protein there we reduce our degree of uniformity if we have a macro scale weighing error we reduce off of uniformity can you see that that this is all this this degree of uniformity is decreasing multiplicatively so you can see that you can carry on adding these things on and so your degree of uniformity declines and and and decreases dramatically so if you can eliminate mixing error you can make a big difference to your uniformity or you can have accurate macro scales also on so the opposite is also true so yes they multiplicative but if you can make small differences to the different aspects you can make a huge difference to your overall variability in your diet so it's a whole whole team effort that is required in order for us to formulate that accurately we need an accurate description of the nutrients of the ingredients and we need an idea of the variability we will encounter and I think we've seen that very very well in the in the previous talks today I want to tell you from a slightly cynical point of view that long-term historical data to me is practically worthless we have to have it to build up our collaborations for NIR but I'm not interested in the 2018 crop report because I'm actually mixing feed in 2019 I want to know what's happening now histories means to me okay so I've got a problem with that and the point is that we've seen today that both our crops and our analytical methods change so if you showing me that NRC from 1994 no disrespect one has to wonder how they did the analysis in 1994 when we know how much everything else is improved and I don't know where this is not the same paper but this is what Bob showed earlier is that soybeans are changing with time the protein content of soybeans is dropping and I'm not going to go through that and interestingly enough exactly the same thing is happening with corn and that the crude protein content of corn is dropping whereas the starch content is increasing and it's it's not done intentionally but remember that it's starts that is needed to make ethanol and corn syrup and all of these things so the plant breeders are selecting for higher yields of starch the net result is that the crude protein is dropped so what I'm trying to say in a roundabout way if your date is 10 years old it's probably wrong and if you're using Ewing which was published in 1932 I've got a sample copy of my room I draw on my bookshelf it's going to be completely wrong so so you've got to keep up to date with these things otherwise you're going to be wrong so how do we deal with the variability as commercial practitioners P Andre gave the example of champagne with 70 different varieties of of grape and actually I think that the truth is that more the more ingredients used and the less very less variability is an issue and I would agree completely that the number of ingredients that feed mills are using is reducing meat and bone meal for example has been banned in Europe and in my country but the other thing of course is that we used to be able to use small parcels ingredients because we had a small mill making small amounts of feed those small parcels of ingredients in the big mill or a nightmare you haven't got a bin for them you've got nowhere to put them so you just don't buy them so what's happened is Mills Eve used less and less ingredients okay however if you wanted to use lesser known ingredients they can be problematic because the data is not there and I I don't know how many of you get somebody sends you oh we've got some macadamia nut residue and they send you an analysis which has got crude protein fat and an energy value worked out on the airport Atwater factor for human beings and at what I would remind you was published in 1906 so it's really up to date the human nutritionists are right at the cutting edge of energy metabolism all right we do know that every time you weigh something you make a weighing error so the more ingredients you have the more weighing errors you like you to make so that's offset the fact that you're using more ingredients okay and as I've said most Mills in my opinion and I are using fewer ingredients and these few ingredients need to be and a source of absolute consistency and the honest truth is if we understand our grain and probably our soybean meal in most circumstances you you will deal fairly adequately with your with your variability the other thing you could do is you could use precision real-time analytics and I say NIR but remember the accuracy depends on the calibration and you've got to have enough samples so you need a large sample database because you have to consider those outliers and that's why say historic data does have value it does have value for for for calibration and then you need to be aware that the method that was used in drawing up the calibration is really important and you've got to remember that no NIR output is going to be better than the input in other words no NIR output is ever going to be better than the data that was used in building the calibration and this is some data that were shared by Pascal and where is from from additio and it's really quite interesting because this is the ami of soybean meal and you can see there's this many many hundreds of thousands of samples and there's a fairly sort of steady block there but the real issue is on the x-axis we have the ami from the WPS a equation and on the y-axis we have the ami from the p and e NIR system and you can see here the relationship between the two is this blue dotted line if the two values were the same the relationship would be that orange line so what it basically says if you're using the WPS a equation you're going to be under estimating the high value material and over estimating the low value material so the importance of using and remember that the PNE system is built up from real ami determination of heating boilers it's not an equation it's real data so you can see how if you use the wrong data and building your database you can be misled by an hour just for interest paradoxically of course all the data that is produced by the in our and is difficult to use meaningfully and one of the previous questions was what do we do with all this in our data and if it's historic my advice to you is to ignore it you know worry about the stuff that's happening now okay so what do most of us do we still tend to use the mean we don't make too many changes or we're pretty good at completely ignoring the data and that has its own limitations again this is the DCO data and I just it's exactly the same just data set that Bob just showed you and if you in the average area if you use an average value you're okay not too bad but what happens if you're not in the average area and you're actually below what you would think the line should be I want to tell you those unexplained broiler performance problems you have these are that's where it comes from and then of course if you don't capture the upside you lose money so there's a huge amount of data in here and then I said we're very good at ignoring things and I was put up to doing this and I think it's a really good idea this is me fiddling around with the formulation I use maize at 12% moisture and then I increase the moisture content of the maize to 14% and all I did is I substituted the one for the other you can see exactly the same formulation just substitute one for the other and I recalculated the energy so I've lost point two ver mega Joule which is 50 kilo Cal and in order to overcome that I reformulated with the correct level of moisture the normal manner and you can see that both the the soybean meal and the oil have gone up but really what's what I wanted to show you is if my expected FCR was 1.6 with my formulation being correct if I hadn't made the adjustment the formulation the FCO would have gone out to one point six five all right remember every 25 calories gives you one point of FCR roughly and you can see that my cost per kilogram of broiler has gone up by three percent when I reformulate correctly and I bring my energy back to what it should be that cost difference is more but I'm not losing money so ignore the moisture content of your corn at your peril all right it's gonna cost you money in poor performance I don't want to talk a lot about formulation issues just remember there's the risk of errors all right and yes we do have models that can deal with variability within formulation systems but they tend to be quite complex to use you laughing and the trouble is complexity leads to complication and you know what human nature is if it's complicated we won't do it so it's tend not to be a great solution we need some practical solutions in order to deal with variability in formulation systems and these systems need to be understandable and simple to use and I hope Franco that tomorrow you're gonna show us exactly that because you can have them too complicated okay so what do we want to do is feed manufacturers to deal with variability as I said we want to preserve the identity of our disparate feed ingredients and that's really difficult in the feed mill sometimes we need to weigh and mix accurately and it's unbelievable actually wanted to have a photograph I've seen people weighing a hundred grams of feed additive on an old-fashioned Avery balanced scale true story so people do the most amazing things and they wonder why they can't make feed simple idea would be to use more bins but bins cost money and often there's not space to put them into a feed mill I would like to see rounding errors minimum weighing quantities and the use of mixer mill outlawed but that's not going to happen because these things are still there they're not gonna go away and I know that there's some arguments for using real-time feed formulation and I know some companies are doing this and what they're basically doing is measuring the ingredients coming into their way hoppers and formulating on-the-fly the problem with it is it makes managing feed mill help so I just did a very simple example I'm making a hundred tons of layer meal and 300 tons of broiler grower and I did a very simple thing I changed the protein from our soybean meal from 47 percent to 46 percent and I changed my sunflower meal protein from 37 percent to 38 percent now these are 1% differences and this is perfectly understandable and what you would expect to happen the trouble is that the formulation changed completely you can see here I've suddenly gone to a maximum a sunflower meal and here I was at my minimum that's that's fine it's not a big problem we can do that and the birds will perform normally but when you start looking at your inventory of raw materials of ingredients you'll see that you use a little bit less brand but here's the real problem you use less swiming meal but you use 200% more sunflower meal than you thought you were going to use so guess what happens you run this formulation for two days and you run out of sunflower meal then what have you got to do the famous substitution formulation that costs you a fortune okay so you think that doing real time formulation is a good idea think again it makes managing your inventories a nightmare and it makes your counting even more difficult so I don't know whether this is a practical solution I know the guys sending in our machines want us to do all of this and because they sell these in line in ours and they work really well but I don't know what whether you can use that information sensibly weighing errors are probably our single major source of of variability and firstly we want to eliminate them by making sure that the formulations are implemented correctly and I I've been in this business a few years and and I can't tell you how many times I've seen formulations that being implemented correctly and somehow when it's automatic and it goes from the formulation system to the process control everybody assumes it's correct it doesn't work that way you got to check these things you got it somebody's got to check them manually batching software can mix these things up and make mistakes size and weights on scales are very very important and people tend to forget that and remember that there's inflight weight of materials good batching software takes consideration of this but poor systems don't so you can have variability there and you certainly have issues with ingredient size and particle size and density limestone's a classic example nearly every feed mill I ever visit the calcium level is higher than it should be because Pete mills do not weigh limestone very well because it's so dense and it flows so fast okay as I've said to you the more ingredients the greater is the error but there's one positive and that is the larger the batch size relatively speaking the smaller the weighing areas and as feed mills get bigger and bigger our batch sizes are getting bigger and bigger so actually by default we weigh more and more accurately I'm still amazed that there are companies who make half-ton mixers and sell them because you can just imagine how big the weighing areas on a half-ton mixer compared to a 5 ton mixer it's it's huge then as I said to you we do this all the fancy work and managing our variability of our feed and guess what we gonna do we gonna feed chickens and and there is variable there more variable than the soybean meal and lots of people will tell you that the very low feed intakes particularly of small broiler chicks is why variability of the feed is so important I want to challenge anybody here to go and find any research on this because I've tried I really tried and there's very little evidence of it I did find one work it was pub one paper that was published by McCoy in 1994 and what they did is they used the common diet and they stimulated or simulated variability but poor mixing intermediate mixing an adequate mixing and when they fed them two brothers they found there was an improvement from poor to intermediate but there was no improvement from intermediate to adequate okay and then pink and I unfortunately don't have a datin on this particular reference leave the TVs of up to 20 percent on feet may be adequate for broiler growth Wow so we worrying about variability in the chickens don't really mind then I came across this interesting MSC while I was looking at this work and what what they were actually doing was looking at meat and bone meal and they've hit a 57% material and a 50% material and when they formulated these diets taking that into consideration you can see that they've got the same FCR and the same the same FCR and the same body weight this is just a different ages they then had a very interesting thing and they took this 50% material and they mixed it with the 57 on an alternating basis and that's what the a stands for and you can see that it made no real difference to the FCR at all now they're mixing 50 and fifty seven percent material they're trying to induce variability and I couldn't do it and then when they formulated correctly with 50 percent material or what they they're sorry they didn't they formulated with 57 and they just substituted 50 percent material and you can see here we've got not even there have we got a significant slightly slightly different slight difference in FCR two points so then we're trying to miss these birds up and they didn't really succeed so what I want to say to you is this broiler that we have is amazing it really is amazing and sometimes I think we overthink these things I really do okay when one goes and looks at individual laying hens and I've just been feeling some some individual Birds three different energy levels for different lysine levels and when you go and look at the average feed intake was 111 grams and nobody's going to say that that's good or bad that's pretty average but the bird that ate the least in the flock 881 grams and the bird that ate the most 854 145 grams there's normal laying high light Browns can you believe that the feed intake would be that different interestingly enough there were two birds that didn't lay and they only ate 63 grams a day then an eat at all they were just coasting along staying alive so yeah you think variability is important that don't know whether the chicken thinks it's that important so I want to end off by saying that we don't live in a perfect world but it's essential that we measure in order to manage and and and in our and it's calibration is a case in point you really have got a system that is is effective and yes we appreciate that the system is not perfect but remember that whatever we apply only has to be better than what we were doing before and in our might not be perfect but it is way way way better than the old wet chemistry labs that I grew up with when I was a boy nutritionist it's hugely better all right the big thing about it is it's going to allow us or it should allow us to move away from the use of averages and historic data and it should allow us to make those small changes in variability that have big outcomes remember variability is multiplicative so I'm really positive about the direction we're moving in as an industry and and I want to say in our calibrations are a classic example of machine learning the more data you feed them the smarter they get and the better they get and and and companies are investing huge amount of money into these systems or to be congratulated because we can all participate in this improvement so systems such as the Odysseus system allow us to avoid materials all suppliers that are variable or unreliable they allow us to manage ingredients in such a way as to minimize variability and hopefully they will allow us to formulate our diets more accurately and as good as they are they do not make up for poor feed manufacture and they do not play any role in flop variability and I want to end off by it's it looks like it's a quote but it's me as always everything we need to do we need to strike a balance between what is possible and what is pragmatic or practical systems must be too complicated but that mustn't also not be too simple thank you very much [Applause]