Bufetico Join the waitlist
GrowingMethod

One bad year and you sold. My model said Amazon was shrinking. A farmer with full bins has the same problem.

Measured first to last, Amazon's cash said the company was shrinking 21.6% a year; the median of the same series said +19.7%. Craig Myhre, a Wisconsin farmer for nearly 40 years, went into the winter with his 2025 crop unsold in the bins, and judged on that one year, you'd get him just as wrong.
4 September 2026 · · 6 min read

You looked at one ugly year, decided the business was shrinking, and let it go. Then it kept growing without you. Judge a farmer on one season and you’d make the same mistake. Craig Myhre went into the winter of 2025 with his grain bins full, and he rented extra storage at a farm nearby too. He’s grown corn and soybeans in western Wisconsin for nearly 40 years. In a normal fall he’d have sold all of his beans and a good part of the corn by then, he told Wisconsin Public Radio. This time the crop sat in storage while he waited for a better price.

The problem was the price. All summer, soybeans sold for less than it cost him to grow them. So he didn’t lock in a price ahead of harvest for part of the crop, the way farmers do with a forward contract. By harvest time, the best he could see was to hold the crop until spring. And a harvest that stays in the bin doesn’t pay any bills.

Normally the money comes in during the fall. It pays off the year’s expenses and buys the seed and fertilizer for the next one. This time it didn’t come. In December 2025 he was working with his lender to pay off his 2025 expenses and to set up an operating loan for the year ahead. He didn’t expect the new year to be any easier for farmers. “If you’re borrowing money to survive, you can only do that so long,” he said.

Now imagine you’re the lender, because my model has the same job. On your desk you’ve got nearly 40 years of a man growing corn and beans, and one fall when the cash didn’t come in. Judge him on that fall alone and all you’ve measured is one price in one year. Look at the decades instead, and 2025 is one odd year at the end of a long run. It’s the same farm, but the two views point in opposite directions. So it all depends on which years you look at.

Lenders and insurers already judge a farm the second way, on its run of years. Federal crop insurance bases a farm’s coverage on its own record, the average of its last four to ten years. And the Risk Management Agency lets a farmer drop a year when the whole county came in at least half below its ten-year average, because that kind of year is usually about the weather. Lenders also put a farm’s history of yields together with the prices they expect for the year ahead. MPR News described it that way from Minnesota that same December. The history only tells you what a normal year looks like. Whether the year ahead will be a normal one, you have to figure out some other way.

My model has to make the same call for every company on my list, and the first way I did it got Amazon wrong. Take nine numbers, in billions, from 2017 to 2025: the cash Amazon’s operations brought in each year, minus what it paid for property and equipment. They run 6.4, 17.3, 21.7, 25.9, then two years below zero, minus 14.7 and minus 16.9, then 32.2, 32.9 and 7.7. Amazon’s own free cash flow figure is a little different from mine. Amazon subtracts what it got back from selling equipment and from incentives, and I don’t.

The simplest way to turn that into a growth rate is the one I started with. You take a first number and a last one, and you work out the yearly rate that gets you from one to the other. Say you start at the 25.9 of 2020 and end at the 7.7 of 2025. The rate is minus 21.6% a year. Feed that into a model that projects cash into the future, and it does exactly what you told it. It projects a company that shrinks by a fifth every year, and it values the shares as if that were true. The free tools I’ve tried measure growth this way, and I understand why. It’s one line of arithmetic.

That’s like judging Myhre on one fall. Minus 21.6% really is the rate that gets you from 25.9 to 7.7. Start a few years earlier and you get a different rate, so the first year you pick matters as much as the last. Two numbers decide everything, each of them a single year, and the years in between get thrown away. That only works if the final year is a normal one. Amazon’s 2025 wasn’t, and Myhre’s wasn’t either, by his own account. In 2021 and 2022, Amazon’s cash went below zero while it built out warehouses and data centers. Two years later it was back above 32. A first-to-last measure can’t see any of that. It takes the small number at the end as the place the company is headed.

So now I measure it another way. You work out the change from each year to the next, line those changes up from smallest to biggest, and take the one in the middle. That’s the median. You skip the pairs that start or end below zero, because a percentage change from a negative number doesn’t mean anything. Do that, and Amazon comes back at plus 19.7%.

So the same nine numbers give you two answers. First-to-last tells you where the company ended up compared with where it started. The median tells you what a normal year looks like for this company. For Myhre, a normal year is the one he described to WPR: the beans sold by the fall, and a good part of the corn too. If you’re trying to picture the years ahead, that’s the one you want.

I need a measure that holds up across thirty thousand companies, because I can’t check them one at a time. An average doesn’t hold up, because one freak year drags it. Let’s say a company grew 12%, 14%, 11%, and then 400% because it bought a competitor. The average of those four is a number that never happened and won’t happen again. A median doesn’t care how big the 400% is, only that it sits at one end of the list.

Where this breaks

The median is slow. Take a company that really is in decline, with three straight years of falling cash and no building program behind it. It still shows a decent number for a while, because the median waits for most of the years to turn before it moves. With Amazon, the median also rests on very little. Once you set aside the three pairs that touch a negative year, it sits on five changes, and one of them is a drop of three quarters. So a normal year here is one reading out of five. That’s why the growth number can’t be the only thing I look at. I read the shape of the series too, and a run of falling cash should knock a company out no matter what its median says. And a projection starts from two numbers: the rate, and the year it grows from. The median fixed the rate. I’ll get to the year it grows from in another post.

The same slowness shows up on a farm. WPR notes that soybean prices had been down since the end of 2023. That makes Myhre’s 2025 the second low year in a row, with a good crop sitting in the bins. Two bad years can be the start of a pattern, and the median wouldn’t have moved yet.

The beans in Myhre’s bins

Myhre’s 2025 money went into the bins. It’s a good crop he was holding for a better price, and in mid-December he was still waiting.

Amazon’s 2025 cash didn’t vanish either. That year it spent $131,819 million buying property and equipment, against $82,999 million in 2024. Amazon’s annual report, the 10-K, says that money went mainly into technology infrastructure and into more room in its fulfillment network, the warehouses that ship its orders. Most of the technology part was for AWS, its cloud business. The 7.7 at the end of my series is what was left. And its operations brought in $139,514 million, more than in any year before.

Key factsThe numbers and where they come from · 4 September 2026
  • In 2025, Amazon spent $131,819 million buying property and equipment, against $82,999 million in 2024, mostly for technology infrastructure to support AWS and more capacity for its fulfillment network (Amazon’s annual report).
  • Amazon’s operations brought in $139,514 million of cash in 2025, more than in any year before, according to the same report.
  • Federal crop insurance bases a farm’s coverage on its own record, the average of its last four to ten years. The Risk Management Agency lets farmers drop a year when their county fell at least half below its ten-year average.
  • In mid-December 2025, Craig Myhre, who has grown corn and soybeans in western Wisconsin for nearly 40 years, still had his 2025 crop in storage, waiting for a better price, he told Wisconsin Public Radio.

Bufetico puts every company on more than 30 exchanges through the same six layers. Prices update after every market close, and company financials once a month. What passes is decided by thresholds you set yourself. See it.

← All articles

Sold too early, or held on too long?

Join the waitlist and I'll email you when Bufetico opens, plus a new article or tool now and then.

No more than one email a week. You can unsubscribe anytime, and the Privacy Policy explains how we handle your name and email.

Can't tell a cheap stock from a dying one?

The Bufetico Filter checks more than 30,000 companies against six layers and shows you only the ones that pass. You decide how strict each layer is. Bufetico opens soon.

Join the waitlist »