Companies have a handful of acceptable methods they can use to figure out what should go on a nutrition label. One way to do it is to send the food product to a lab for analysis. Any food that is fried, coated or salted needs to have their nutrition information determined by a lab because of its complexity. But, this is time consuming and expensive because the labs have to follow strict FDA procedures.
But, if a food isn’t fried, coated or salted, there’s another option.
A company can instead use a “nutritional database” to figure out what to put on their food label. This means that the company can go on the internet and pick one of the online nutrition websites, and enter the ingredients for their food item, and use those internet results to generate their nutrition label.
The FDA does not regulate nutrition labels on a proactive basis. The FDA has guidelines, to be sure, but the FDA does not (and could not possibly) verify whether every single nutrition label ever printed is correct. If there’s a complaint about a product, the FDA can investigate and issue a recall of a product, but that typically only happens if there are reports/complaints about a product.
The FDA also allows a margin of error of 20% either way. So if the item in question has either 20% more or 20% less of a nutrient/substance than is indicated on its label, that is considered acceptable. So a product labeled 100 carbs per serving could actually have 80 carbs or 120 carbs and still be considered accurately labeled.
A random audit in the 1990s discovered that 90% of food labels were within this acceptable 20% margin of error, and 10% were not.
So, the short answer to the question posed during the webinar is: A nutrition label is 90% likely to be up to 20% wrong and 10% likely to be worse than 20% wrong.