How to back an expiry date with evidence, tell a quality date from a safety control, label it with the phrase that matches what it means, and run a routine that moves near-expiry stock before it becomes waste.
Do these first
Real help from real food people
Chef Diego runs a real food plant. If this page didn't get you there, tell us — a person reads every message.
After this lesson you can set an expiry date you can stand behind: back it with evidence instead of a guess, know whether your date is a quality signal or a safety control, print the phrase on the label that matches what it actually means, and run a routine that moves near-expiry stock before it turns into waste or a rejected delivery.
The date you stamp on a product is a claim you are making about it. It says that until this date, stored the way the label directs, the product will still be what you sold — safe where safety is at stake, and acceptable in taste, texture, and nutrition. The moment a customer, a retailer, or an auditor asks how you arrived at that date, you are either holding evidence or holding a guess.
is the length of that span; the expiry date is where you decide it ends. The strictest example in federal law shows the standard even where it does not force it on you. Infant formula is the one packaged food that must carry a date: under 21 CFR 107.20, the manufacturer picks a "Use by" date and has to base it on tests or information showing that until that date the formula still holds the labeled amount of each nutrient and stays of acceptable quality. That is the shape of a defensible date for anything you make — a date chosen from evidence about how the product behaves over time, not a round number that felt about right.
You cannot set a date you can defend until you know how the product goes out of spec. There are three ways it happens, and most products have one that leads.
Microbial
Pathogens or spoilage organisms grow to the point where the food is unsafe or spoiled. Two properties govern how fast: and pH. The acidified-foods rule, 21 CFR 114.3, draws its line at a water activity greater than 0.85 and a finished equilibrium pH of 4.6 or below — the range where the low-acid and acidified food rules treat microbial growth as a serious hazard. Push either property below those thresholds and growth slows or stops; sit above both and time becomes something you have to control.
Chemical and physical
The food changes on its own without any microbe. Fats oxidize and go rancid, color and vitamins fade, moisture migrates and softens a crisp product or dries a moist one, emulsions separate, and staling sets in. None of this makes the food unsafe, but all of it ends the window a customer will accept.
Sensory
Taste, smell, texture, and appearance drift past what a customer expects even while the food is safe and its chemistry is fine. This is what most "best by" dates are really about — the day the product stops being at its best, not the day it becomes a hazard.
Which of these leads for a given product is largely decided upstream, when the product and its process were developed — the formulation's water activity and pH, the fat that can go rancid, the packaging that keeps oxygen and moisture out. Setting the date is where you measure the consequence of those choices and put a number on it.
This is the distinction operators get wrong most often, in both directions. Ask two separate questions about your date and keep the answers apart.
The first is a labeling question. As a matter of federal labeling law, except for infant formula, a date is not required and the common phrases signal quality, not safety. USDA's Food Product Dating guidance is plain about it: a "Use-By" date marks the last day of peak quality and is not a safety date except on infant formula, a "Sell-By" date is for the store's own inventory and display and is not a safety date at all, and dating in general is not required by federal law outside infant formula (verified July 2026). So for a jar of shelf-stable sauce, the printed date is not a cliff the food falls off — it is your best read on when quality declines.
The second question is a food-safety question, and it can override the first. For some products your own hazard analysis makes the shelf life a real safety control — a refrigerated ready-to-eat food where Listeria can grow slowly over storage, or a reduced-oxygen-packaged product where Clostridium botulinum is the concern. When storage time and temperature are what stand between the product and a pathogen, the date is a , and 21 CFR 117.160 requires you to validate it — to obtain and evaluate scientific and technical evidence, or run studies, showing the control actually holds the hazard, with the work overseen by a . The trap runs both ways: do not treat a safety-critical date as a soft quality suggestion, and do not scare customers with a hard "unsafe after" date on a product whose date is really only about flavor.
Whichever failure mode leads, you find the date the same way: you decide in advance what "out of spec" means, then you watch the product cross that line.
A runs in one of two ways. A real-time study stores the product at its actual storage conditions and tests it at intervals — a microbial count, a peroxide value for rancidity, a trained sensory panel — until it fails the limit you set. It is slow, because it takes as long as the shelf life itself, but its result is the one you can defend without an asterisk. An pushes the product harder so it ages faster and you get a number sooner. The catch is that it is an estimate: heat can drive a change that never happens at real storage, and some failure modes do not speed up in step with temperature. Treat an accelerated result as a working number to launch on, and confirm it with real-time data as it comes in. Set your date short of the point where the product failed, with a margin, and keep the study records — they are the evidence behind the claim.
Two different dates can end up on a package, doing two different jobs. is the one a person reads. is the code you trace by — the same lot-and-date code the lot codes and date codes lesson covers, which is how a date on a shelf ties back to the batch behind it.
For the open date, match the phrase to what the date means. When the date is about quality, FDA and USDA both point industry toward the voluntary "Best if Used By" wording, because it tells a customer the product may still be fine after that day if it was stored right, which cuts needless waste. Stamping "Use by" — the peak-quality phrase — on a product where you actually mean a safety limit, or a hard-sounding date on a product that is only losing flavor, sends the wrong signal. Check your format against FDA's food labeling guidance, and remember the one hard requirement: infant formula carries a "Use by" date, and nothing else has to unless a specific rule for your product says so.
A date is only worth the routine that acts on it. The routine is a standing near-expiry review: on a set cadence you surface every lot approaching its date and decide what happens to it — ship it first, discount it, divert it to a use that fits its remaining life, or dispose of it and record why. Without that review, a defensible date just sits in a record while the stock quietly ages past it.
The review feeds directly into rotation. Shipping the soonest-expiring lot first — — only works if something is watching the dates and telling the picker which lot to pull, as the FEFO lesson covers. It also protects the deliveries themselves: many retailers reject a lot that arrives with too little shelf life left, so moving your nearest-expiry stock first is often the only way to keep enough life on the product to be accepted.
Software built for food records each lot's expiry and surfaces the nearest dates first, so the near-expiry review runs off real dates instead of a manual sweep of the cooler. Bettr Manager, the platform this site is part of, works that way — the expiry you set drives which lot the picker pulls first, and the stock list reads in expiry order, so whatever is closest to its date sits at the top. It is one way to run the routine; a well-kept spreadsheet reviewed on a schedule reaches the same place with more hand-work. For one concrete way to see what stock you are holding and its lots, the docs walkthrough on browsing your products shows the view.
A defensible date is not finished until it lands on the right units and into the record for the lot that carries it. The date has to be printed where the customer and the picker can read it, and captured in the batch record so every lot points back to the study behind it. That connection — recipe, lot, date, and the log the floor keeps — is where this course turns next: pulling all of it together into the batch records the people doing the work actually fill in.
Do this in your operation