What am I missing here? I'm sure there's plenty of science and feed testing that goes on in computing dairy rations, but... isn't this something where you could get a grip on some of these variables with some well-written software that interfaces with the herd records? ie, you have the software know what is going in, the herd records show the milk response, software (written correctly) would start to tease out the correlations and signals from feed changes.

OK, let's back up one more level. Lemme quit being the computer nerd here and ask why these nutritionists aren't paid for performance -- ie they get some fee for working with you/your herd, but then they get a bonus if they increase milk production by X%?