FIELD NOTE 001 · AMERICAN APPAREL
A sweater, a factory, and a 90% collapse.
Framework made my new sweater in Los Angeles. I followed the yarn—and the data—to see where that production model could actually work.
Drawn in your browser, one loop at a time. This is a procedural interpretation—not Framework's machine sequence. See the actual sweater.
The opening illustration constructs a stylized sweater from the hem to the collar as the reader scrolls. The animation is illustrative and does not reproduce Framework's manufacturing process.
The sweater is navy, cotton, and almost comically direct. An American flag is knitted across the chest. REINDUSTRIALIZE runs underneath it. Turn the collar over and the label says “Made in USA by Framework.” It is industrial policy you can put through a delicate wash cycle.
Framework says the sweater came off an 8-gauge knitting machine in one piece—no cut panels, no side seams. I liked it enough to ask what the slogan would require economically.
The missing factory
Automation is often blamed for the loss of American factory jobs. Apparel’s numbers point elsewhere. Since 1990, inflation-adjusted U.S. garment purchases have more than tripled. Over the same period, domestic apparel output fell by 90%. Labor hours fell by almost exactly the same amount.
THE GREAT SEPARATION
We bought more clothes. We made 90% less here.
Index, 1990 = 100. Hover, tap, or use the arrow keys for annual values.
The aggregate data do not show a productivity-led contraction. Output and labor hours fell together, while output per hour rose only 1.4% from 1990 to 2025. At the national level, the output disappeared with the hours.
How import-dependent is the market now? For 2021, USITC economists estimated import penetration by value at 90.3% for apparel knitting mills and 94.8% for cut-and-sew manufacturers.
100 STITCHES
Domestic supply barely registers.
Framework is entering an industry whose domestic base is now tiny. Its plan depends on a new production system, not rebuilding the labor-intensive one that disappeared.
The valuable thing may be time
Wages dominate the usual comparison between Los Angeles and Asia. Framework’s case depends on lead time.
In a 2018 paper, the USITC described one U.S. manufacturer’s experience: less than one month for domestic orders and at least six months for orders from China. Six months before a selling season, a brand has to guess what people will want. How many navy sweaters? Which sizes? How much enthusiasm for giant lettering?
A cone of yarn has options. A finished medium navy sweater does not.
THE COMMITMENT CLOCK
When does flexible yarn become a fixed bet?
In the Boathouse case, under one month: the brand could wait longer before committing yarn to a size and style.
Apparel is visibly inventory-intensive, although public data do not isolate why. In 2024, clothing and clothing-accessory stores had an average published inventory-to-sales ratio of 2.28. The ratio for retail overall was 1.31.
INVENTORY-TO-SALES RATIO · 2024
Source: author's average of 12 seasonally adjusted monthly ratios in the 2024 Census retail benchmark. These are store categories, not apparel sold through every channel; inventories are valued at cost while sales are receipts. The comparison does not isolate the effect of lead time.
This is the part of Framework founder Denver Rayburn’s argument that I find interesting. In essays on the complexity tax and cost physics, he argues that local, automated production lets brands postpone the moment yarn becomes a particular style, color, and size. His formulas remain conjectural. The underlying mechanism is ordinary inventory economics: postponing that decision could offset part of the domestic cost premium.
Framework’s contract clarifies where the inventory risk goes. Customers generally fund the yarn. If the agreement is terminated, they may also owe for unshipped products and unused Framework-supplied materials. Some exposure can remain as more flexible yarn; finished-goods risk does not disappear.
What the machine actually changes
An older process comparison gives the machine a more modest scale. A 2012 doctoral thesis modeled two ways to make a 12-gauge sweater. The conventional, fully fashioned process took 123 minutes. The nearly one-piece process took 113. Knitting itself was actually three minutes slower. The win came later: less cutting and sewing.
WHERE THE TEN MINUTES GO
The modeled gain was ten minutes.
The modeled saving was 8.1%, concentrated in finishing and assembly. The machine reduced handoffs far more than it accelerated knitting. If the factory is close enough to the customer, it can also move a consequential decision closer to demand.
There is no uniquely American machine here. SHIMA SEIKI sells whole-garment equipment worldwide. An overseas factory can buy the same class of hardware. Framework therefore has to outperform on operations: utilization across small runs, fast program changes, defect control, yarn tracking, and fulfillment.
What I want to know next
The sweater proves that Framework can make a very good sweater in Los Angeles. It does not prove that the factory can make money doing it at scale. Five lines of operating data would tell most of the story:
- 01Domestic cost per sellable garment
- 02Matched landed cost from offshore
- 03Machine utilization and first-pass yield
- 04Median and worst-case order-to-ship time
- 05Customer markdowns and stockouts versus the old system
As of August 2026, I could not find public Framework figures for them. Those are the numbers that separate a clever factory from a great photograph.
The strongest candidate is apparel at its most annoying: uncertain demand, many sizes and colors, small opening orders, frequent replenishment. Commodity basics are the harder case. If you know you will sell a million identical white T-shirts, the giant offshore factory has plenty of time to be giant and cheap.
The sweater is still a small thing. It is also a physical answer to a question the data make hard to dismiss. America did not lose apparel because Americans stopped wanting clothes. We lost the factory while demand exploded.
My sweater will get worn. Framework’s business model will stand or fall on whether customers reorder faster, mark down less, and save enough to cover production in Los Angeles. Those are the next numbers worth publishing.