The Locker Wants Kill Dates 11 Months Out. How Many Beef Halves Sell in November Is a Forecast Now, Not a Hunch.

Fourteen quarters sat in the freezer on the third of March. They were cut in October, they were paid for in nobody's name, and by March they represented about eleven thousand dollars of beef that had already cost money to raise, kill, cut, wrap and freeze. The owner booked twenty-four head of kill dates the previous spring because that is what the locker had open, and twenty-four head is what he took in. Seventeen sold. That gap is not a marketing problem. It is a forecasting problem, and until this year it was genuinely hard to solve at the size most cattle operations run at.

Anyone selling beef by the half, the quarter or the eighth knows the two facts that make this ugly. First, the processing slot is the binding constraint — since 2020 the small state-inspected and custom-exempt lockers in most cattle country book kill dates eight to twelve months ahead, and the date you did not book in April is a date you do not get in December. Second, your demand is not smooth. It piles into September through Thanksgiving when people fill freezers before winter, it gets crowded out at the locker by deer season at exactly the same moment, it takes a smaller bump in late May before grilling weather, and it goes flat in February.

What the guess actually costs when it goes the wrong way

The guess is wrong in two directions and both cost money. Guess low and you turn away buyers in October with no date to give them, and a customer who wanted a half and got told "next year" often buys from a neighbor and stays there. Guess high and you carry beef: freezer space, a chest freezer running for months, and the discount you eventually take to move quarters in March that you would have gotten full price for in November.

Demand forecasting here means using your own three years of order dates, deposits and cancellations to estimate how many halves you will actually sell in each month ahead, so you book kill dates against a number rather than against last year's feeling. That is not a new idea. What is new is that it works on the amount of history a direct-beef operation actually has.

Why 2026 changed this for an operation with 60 orders, not 60,000

Demand forecasting is now one of the highest-adoption uses of AI anywhere — around 48 percent in manufacturing, sitting alongside pricing optimisation at around 72 percent in retail and e-commerce. Those numbers come from businesses with a lot of rows. The reason it reached a cow-calf outfit selling forty head a year direct is that modern models handle messy, sparse, seasonal history that classical methods choked on: sixty orders across three years, half of them clustered in eight weeks, with two years distorted by a drought cull and one by a locker that closed.

The old spreadsheet approach needed clean, regular history to say anything, and with thin data it either refused or produced a straight line. The 2026 approach can be handed the actual mess — the order list, the deposit dates, the cancellations, the waiting list, the year you had to sell twelve cows early — and asked in plain English what November looks like and what would have to be true for it to be wrong. It costs a few dollars a month to re-run every week, which is the other half of why it is finally usable.

flowchart TD
  A["3 years of half and quarter orders"] --> E["Weekly demand estimate by month"]
  B["Kill dates already booked at the locker"] --> E
  C["Deposits taken and waiting list"] --> E
  D["Corn, hay and feeder calf prices"] --> E
  E --> F{"More buyers than booked dates?"}
  F -->|Yes| G["Call the locker for another date, hold the last four halves at full price"]
  F -->|No| H["Open the waiting list, push quarters, no discount before October"]

What it looks like on the calendar, not on a screen

The useful output is not a chart. It is four decisions with dates on them.

April: how many kill dates to book for the following twelve months, split by month. July: which animals in the feeding group get held for direct sale and which go on the truck as feeders, decided against the forecast instead of against how the pen looks. Late August: the price per hanging pound for the fall run, set once and published, plus the deposit you will require to hold a date. Mid-January: whether to discount the remainder or hold it, which is a real question because holding costs you freezer and cash while discounting trains customers to wait until January.

Dairies run the same logic with different words. If your co-op puts you on a base-excess program, the equivalent question is how much milk you will actually ship each month against your base as spring flush comes on, because over-base milk is paid at a different price and the difference is decided by your calving distribution six months earlier. Same shape of problem, same class of tool, different form.

The arithmetic: twenty-four booked dates, seventeen sold

Now the other direction. Suppose the forecast instead says book 19 and open a waiting list. If you cut the over-book to two head, you recover roughly $4,500 of that. If the forecast is wrong the other way and you turn away three buyers, the cost is three halves of margin you did not earn — call it $2,200 — plus the customers who do not come back, which is the number nobody can put on a spreadsheet and everybody in this business knows is the real one. Use your own prices; the structure of the calculation is the part worth copying.

Where the model will be wrong and you will know before it does

Forecasting cannot see the things that actually move small-scale beef demand: a local plant closing and putting three hundred people out of work, a new competitor at your farmers market, the packing house that shut its cutting room for renovations in October. You know all three of those before any model does, so treat the forecast as an argument you push back on, not an answer.

It also cannot price against your reputation. If you have been at the same farmers market for twelve years, some of your pricing power is not in any number you can feed it. A model told to optimise price will happily recommend that you charge more in November — sometimes that is right, and sometimes it costs you the customer who has bought a half every year since 2014 and would notice.

And it must not touch the animal decisions. Which heifers stay, when to wean, whether that steer is ready, whether the group is finishing on the grass you have — that is your eye and your nutritionist. The forecast tells you how many head to aim at. It does not tell you which ones.

Frequently asked questions

I only have order history in a notebook and some texts. Is that enough?

Usually yes, and this is the part that changed. Photograph the notebook pages, export the texts, add the deposit list. Three seasons of thin, messy history is now workable input where two years ago you would have been told to come back when you had a proper spreadsheet.

Should I let it set my price per hanging pound?

Let it propose a price and show its reasoning against your own costs and last year's sell-through, then you set it. Pricing is where these models are strongest and where an owner's judgment about a specific customer base is still worth more. Publish one price for the season rather than quoting per caller; it saves arguments and it makes next year's forecast cleaner.

How far ahead can it actually see?

Far enough to book kill dates, which is the only horizon that matters here — a monthly estimate eight to twelve months out with a range, not a single number. Anything claiming to tell you exactly how many halves sell the week of November 12 is selling you certainty that does not exist.

Does this help with cull cows and feeder calves too?

It helps with timing, not with beating the market. Knowing your own seasonal pattern of when you ship culls, and putting that next to your feed costs, is useful for deciding whether to hold thin cows and add weight. For actual price risk on feeders, that is what Livestock Risk Protection and the futures market are for, and a forecast is not a substitute for either.

One last practical note. Every one of those October buyers calls before they order — how much is a half, when do I pick it up, do you need a deposit, can I get the tongue and the oxtail. That is thirty or forty calls a week landing while you are sorting cattle. CallSphere builds voice and chat agents that answer the farm line, quote the price and pickup dates you published, take the deposit request and put the buyer on the list — so the demand you forecast actually gets captured instead of going to voicemail on a Saturday.