What 113 Studies Taught Me About Shelf-Life Dates
I often ask myself how anyone can know when a food will go bad before it has had the chance to.
Look at almost anything in your pantry. A jar of peanut butter. A box of crackers. A bottle of salad dressing. Somewhere on it there is a date, printed with total confidence, often a year or two away.
Nobody waited two years to find out.
That date is not a measurement. It is a prediction. I've spent the last few weeks testing how good those predictions really are, against 113 published studies, and the answer surprised me a little.
But first, the story of how we learned to make them.
It begins with a Swedish chemist who had nothing to do with food at all.
In 1889, Svante Arrhenius, building on work the Dutch chemist Jacobus van 't Hoff had published a few years earlier, proposed a simple idea about why reactions speed up when things get warm. Molecules are always jostling. Only the ones that collide with enough energy, more than a threshold he called the activation energy, actually react. Warm things up, and a much larger share of them clear the bar.
The math that came out of it is elegant. Plot the logarithm of a reaction's rate against one over its temperature, and you get a straight line. And a straight line, as every scientist knows, can be extended.
That extension is the whole trick of accelerated shelf-life testing, or ASLT. A food company that can't wait two years puts its product in warm chambers, say 35, 45 and 55 °C, and watches it fail fast: Oils going rancid, colors fading, vitamins disappearing, crunch going soft. It measures how fast each change runs at each temperature, draws the line, and slides it back down to the temperature of a pantry or a refrigerator.
Out comes a number. The number becomes a date.
Food scientists like Ted Labuza spent decades turning the method into everyday practice, and it now sits behind a huge share of the dates on our shelves. It also sits behind a good share of my own work: Clients ask me about shelf life more than almost anything else.
And the date on the package doesn't even mean what most people think. In the United States, federal law requires a date on only one food, infant formula, which must carry a "Use By" date. Everything else is voluntary. In 2017, two big grocery trade groups urged the industry to settle on "Best if Used By" for quality and "Use By" for safety, and the FDA backed "Best if Used By" in 2019. Most of the dates in your pantry are a company's best guess about when the food will taste less good, not when it becomes unsafe.
So how good is that slide down the line? It turns out nobody had checked very carefully, at least not across the whole literature.
Here's what I did. I gathered 113 published studies where researchers measured a food degrading at three or more temperatures, which gave me 409 separate series of data. For each one, I hid the coolest temperature, fit the Arrhenius line to the warmer ones, and predicted what should have happened at the temperature I'd hidden. Then I compared the prediction to what the researchers actually measured.
It's the same test a food company would want to run on its own data, done 409 times.
The good news first. About 3 in 4 predictions landed within a factor of two of the real answer. When I kept only the studies with the cleanest data, it was closer to 9 in 10. And across the whole set, there was no overall lean toward predicting too long or too short. For a 135-year-old equation, that's really quite wild.
The less comforting news comes in three parts.
First, the further you stretch the line, the more it flatters the food. When the prediction had to travel a long way from the tested temperatures, it ran about 1.4 times too long on average. That finding is exploratory, since there were fewer long stretches to look at, but it points in the direction that matters: A product that "should" last 18 months might really last closer to 13.
Second, not every kind of spoilage plays by Arrhenius's rules. Vitamin C loss and phenolic breakdown behaved beautifully, with about 90% of predictions inside the 2× window. Texture changes, anthocyanin color loss and sensory scores were the weak spots, at only 57% to 67%. My bet is that these involve several reactions at once, or physical changes like starch and fat shifting state, which don't follow one clean line. (Your crackers going stale is not a simple chemical reaction, and it apparently refuses to pretend otherwise.)
Third, when published studies "validate" their predictions, they mostly check a temperature inside the range they already tested. That's a much easier test than predicting the shelf, where the food will actually live.
If you make food, here's the practical part: Test at least one temperature close to your real storage conditions, and treat any long extrapolation as an estimate that needs a real-time check, not a date to print.
To make that easier, I built a free calculator that tells you how far your own prediction is stretching and how much to trust it: the ASLT Extrapolation Check. If you want the plain-language version of the findings, there's a whitepaper, and for folks who want every method and number, the full technical report is there too.
There's a small irony in all this that I keep coming back to.
A few years after his equation, Arrhenius turned the same kind of thinking toward the whole planet. Around 1896, if I remember the history right, he worked out by hand how much the Earth would warm if the carbon dioxide in the air doubled. The man who taught us to predict how fast a jar of jam would fade also gave us one of the first predictions of how fast our world would change.
Both are the same question, really. How long can something we care about keep?
Every food on the shelf is a slow reaction running toward stillness. Fats oxidize, colors fade, vitamins quietly come apart. A date on the package is our best attempt to put a number on that, to say this is how long the good part lasts. It's a guess made in a hot little chamber, and most of the time it's a pretty good one.
But the further we stretch our predictions from the conditions we've actually lived through, the more generous we tend to be with ourselves. That's true of crackers. It might be true of most things.
So test close to home. Check the real thing. And enjoy it while it's good.
Dr. Bryan Quoc Le is the Founder and Principal Food Consultant of Mendocino Food Consulting. He earned his Ph.D. in Food Science from the University of Wisconsin, Madison and his B.S. and M.S. in Chemistry from the University of California, Irvine. He currently lives in Mendocino, California with his wife and two dogs.
Are you working on a new food or beverage product? Interested in working with me and my team to get started?
Click on the button below to get in touch and set up a meeting today!
