Grocery Expiration Date Tracking: Cut Shrink in Four Tiers

Tuesday, half past seven in the morning. The owner of a 1,900-square-foot neighborhood grocery is standing in front of the dairy case with a crate of yogurt in his hands, yesterday's date on the lid. Four crates came in, two and a half sold, the rest goes in the bin today. The same scene repeats at the deli counter with sliced cheese, on the bakery shelf with yesterday's bread, in produce with the bananas. The owner calls this "shrink"; the accountant calls it "a percentage of sales".

This guide walks through grocery expiration date tracking in four tiers for anyone running a corner store, an independent grocery or a deli: from a paper log to software, from software to a markdown rule, from a markdown rule to AI. We cover what labeling law requires, how to calculate your shrink rate, and which tier pays off in a small store, with a worked example. For the wider picture of where AI fits across sectors, see our industry-by-industry map of AI use cases.

First the big picture. The UN Environment Programme's Food Waste Index puts global food waste at around a billion tonnes a year, with retail responsible for roughly one-eighth of it. Reliable category-level shrink benchmarks for independent grocers are hard to find in any market, so you will not see an "industry average" claim in this piece. We will show you how to produce your own number instead.

How do you calculate a grocery shrink rate?

Shrink rate is the cost value of product thrown away or sold off at a deep markdown in a period, divided by sales for the same period. The formula is simple; the hard part is adding up the numerator correctly. Most small stores keep no shrink log at all, and those that do keep it at the level of "we tossed two crates today". A shrink log without reason codes does not reveal the problem.

Shrink comes from four separate sources, and each needs a different fix: expired dates, damage (crushed, broken, torn packaging), weighing and cutting loss (slicing at the deli, bone share at the butcher) and unknown loss (theft, items that never scan at the register). What our grocer calls "shrink" is the sum of all four; AI helps only with the first and partly with the third. That is why the first job is adding a reason code to the shrink log.

The practical method: for one week, whoever runs each department writes down every item they throw away or mark down on a slip, like a receipt: product, quantity, cost, reason. On Friday evening, total it by the four reason codes. In the stores we see, more than 70 percent of fresh-category shrink (dairy, deli, bakery, produce) is date-driven. Yours may differ, which is why you run the count yourself.

What is the difference between "use by" and "best before"?

Most labeling regimes recognize two kinds of date. A "use by" date is a safety date, applied to foods that spoil quickly and can pose a health risk: milk, yogurt, meat, deli products. A "best before" date is a quality date: on biscuits, pasta or canned goods it means the product is at its best until then, not that it becomes unsafe afterwards. The distinction decides what you may legally sell.

In the EU and the UK the two dates are defined in law, and selling food past its "use by" date is prohibited. In the US, federal law only regulates dating on infant formula; the "best if used by" wording is an industry convention backed by FDA and USDA guidance rather than a legal cutoff, and state rules vary. In every regime, a product past its quality date can stay on the shelf if it has not spoiled; the discount grocers built on near-date stock live on exactly this distinction.

Turkey spotlight: the Turkish Food Codex labeling regulation uses the same two-date system ("son tüketim tarihi" for safety, "tavsiye edilen tüketim tarihi" for quality). Food past its safety date is classified as unsafe under Article 21 of Law 5996; selling it is banned and inspections bring administrative fines and seizure. Fine amounts are revalued every year, so check the current figure with the provincial agriculture directorate. Turkey also offers a scale reference: estimates of national food waste range from 19 to 23 million tonnes a year depending on the source.

The consequence for shrink management: tracking on safety-dated products is daily and mandatory, tracking on quality-dated products is weekly and commercial. Whatever software or AI you choose, look for a system that can manage these two groups under separate rules.

The four tiers: paper log, software, markdown rule, AI

For a small grocery, expiration date tracking progresses through four tiers, and most stores can stop at the second. Tier one is zero-cost process discipline. Tier two is a store management system with date tracking and near-date alerts. Tier three is a stepped markdown rule keyed to the date. Tier four is AI that learns sell-through speed and sets both order quantities and markdowns on its own.

Tier one, the paper log: replace first-in-first-out with first-expired-first-out (FEFO). New crates go to the back, the nearest date goes to the front. In dairy, deli and bakery you track dates, not counts. Add the weekly shrink slip described above. The cost is zero and, from what we see in stores, the impact is the largest of the four.

Tier two, software: expiration tracking and near-date alerts are now a standard feature in most grocery point-of-sale and back-office systems, from cloud POS packages to dedicated tools such as Date Check Pro, Shelflife or Upshop. The date is entered at receiving, and the system prints the list three days out. Pricing ranges from monthly subscriptions to one-off licenses; we could not verify a current price list, so get two or three quotes.

Tier three, the markdown rule: fixed steps keyed to the date: 20 percent off three days out, 40 percent one day out, 60 percent on the final day. Simple, applied by hand, a label printer is all you need.

Tier four, AI: instead of a fixed calendar, systems that calculate the markdown per product and per day based on sell-through speed, stock on hand and store traffic. More on that next.

When and how much does AI mark down near-date products?

AI-driven shrink systems set the markdown by asking "at the current rate of sale, how many units will be left when the date arrives?" The fixed calendar disappears. A slow-moving yogurt gets an early, deep markdown; a fast mover gets a late, shallow one. The goal is for the product to sell out close to full price before the date, not the 60-percent bloodbath on the final day.

The best-known example of this model is Wasteless, which computes the markdown from sales velocity, expiration date, stock and store traffic and updates electronic shelf labels several times a day. The company says it cuts food waste by up to 40 percent. That is a vendor claim with no independent verification, so read "40 percent" as a ceiling rather than a target. The same goes for the 44 to 69 percent cut in expired shrink that Date Check Pro reports: plausible, vendor-reported, unaudited. Shelf Engine and Upshop play in the same space; check POS integration before you fall for a demo.

The realistic route for a small store is to skip tier four's hardware (electronic shelf labels) and borrow its logic: a simple model that computes daily sell-through per product from the sales history already sitting in your POS, and produces a "mark these down today" list with a suggested percentage every morning. The label goes on paper. That delivers about 80 percent of the value of AI markdowns without the shelf-label investment. We covered the demand-forecasting side in detail in our small-business guide to AI demand forecasting for retail; ordering the right quantity prevents more shrink than any markdown does.

How do you track dates at the deli and on weighed goods?

The deli problem differs from packaged goods: the moment a wheel of cheese is opened, the producer's date becomes irrelevant and the store's own shelf life starts from the opening date. Deli tracking is therefore done by opening date rather than product date, and that date is what the scale prints on the label.

The sequence goes like this: an opening label (product, date opened, who opened it) for every wheel, block or tray; integration between the scale and the store system, so every slice weighed is written into the sales velocity; and a separate reason code for cutting and slicing loss. Once those three are in place, the data the AI tier needs accumulates on its own: which product sells at what speed how many days after opening. We described the same "expiry tracking" logic for medicines in our piece on AI in pharmacy inventory management; at the deli the rule is identical, the shelf life just far shorter.

The worked example: which tier pays off?

The numbers are assumptions; the method is real. Say the store's monthly sales are about $30,000. The fresh group (dairy, deli, bakery, produce) is 35 percent of sales, roughly $10,500. The week-long shrink slip showed a 4 percent shrink rate in the fresh group: about $420 a month in the bin or sold at rock-bottom markdowns. That figure is an example, not a benchmark; your slip may show 2 percent or 7.

Tier one (FEFO and reason codes) on its own cuts shrink by a quarter to a third in the stores we have watched; for our grocer that is roughly $100 to $140 a month at zero cost. Tier two (software alerts) brings missed dates close to zero; going from 4 percent to 2.5 percent is worth about $160 a month. Tier three (stepped markdowns) turns product that was going in the bin into product sold at a discount; the loss does not vanish, but part of the cost comes back. Tier four, at this store's scale, only makes sense as a simple recommendation model fed by the POS data; an electronic shelf label rollout does not pay back on 1,900 square feet today.

In short: for a store this size the first two tiers capture most of the shrink. AI enters the conversation for businesses with three or more locations, or where the fresh group exceeds half of sales. If you are on the way to becoming a chain, collecting the data with reason codes starting today is the cheapest preparation for that day.

Frequently asked questions

What should a grocery shrink rate be?

No reliable category-level average exists for independent stores, and vendor benchmarks are self-serving. Take your own weekly measurement as the baseline and aim to bring it down month over month.

Can I sell products past their "best before" date?

Where there is no sensory spoilage, most regimes allow it; labeling it as a quality date and using a separate discount shelf is the common practice. For products past a "use by" or safety date the answer is a firm no.

How many days before the date should I mark down?

Three days under a fixed rule, variable under a velocity-based one. Start earlier on slow movers; if your system keeps sales history, the number of days can be computed per product.

Does AI make sense for a single small store?

For one location, the first two tiers are enough. AI pays for itself once there is enough data to compute order quantities and markdowns per product, and once there is more than one store.

Are there camera systems that read shelves and dates?

Camera-based shelf and date reading is developing, but we could not verify a working deployment at small-store scale; entering the date at receiving remains the more reliable route.

Who should keep the shrink log?

The person running the department, at the moment the item is thrown away. Logs reconstructed from memory at closing time come out incomplete; recording on the spot, like a receipt, is the only method that works.

So what should you actually do?

• Keep a shrink slip for one week, with the four reason codes. Do not discuss software or AI before you have seen the result.
• Make FEFO the shelf rule. Nearest date to the front, new stock to the back; opening labels at the deli.
• Switch on your POS system's date alert. The feature is most likely already there; it does nothing without date entry at receiving.
• Write a stepped markdown rule and apply it with a label printer. A three-day ladder instead of a final-day bloodbath.
• Accumulate the data. When the store count grows, the AI markdown and ordering model is built on that data; a log started today lowers the cost of that day.

Our grocer's Tuesday morning with a crate of yogurt changed with a week of paper and a shelf rearrangement; the AI conversation came up when the second store opened. The order will most likely be the same in your store. Once the first week of the shrink slip is full, which tier fits you can usually be read straight off the slip. If it cannot, send us the slip and we will read it with you.