FRAMEWORK AUTOMATION · LOS ANGELES
The sweater
and the calendar.
Framework made this sweater in Los Angeles. I wanted to know whether a later production decision could justify making it here.
REPORTING AND DATA UPDATED AUGUST 30, 2026
The sweater that arrived at my house is navy, midweight, and unusually declarative. REINDUSTRIALIZE runs across the chest. The flag and letters are knitted into the fabric; the label says it was made in the United States by Framework.
Two fits and five sizes make ten possible garments before color or design changes. Compatible yarn can serve more than one of them; a finished sweater serves one fit and size. A shorter lead time matters if delaying that conversion prevents enough mistakes to cover any added cost.
Procedural illustration. See the actual sweater.
1990 → 2025 · OFFICIAL SERIES
U.S. apparel manufacturing output shrank to a tenth.
Set 1990 to 100. Real garment purchases rose to 317.5. Domestic apparel output fell to 10.0; labor hours to 9.9. Productivity ended just 1.4 percent above its 1990 level. Output and hours fell almost together: the data describe a shrinking industry, not an automation-led productivity boom.
PURCHASES317.5
OUTPUT10.0
HOURS9.9
1990 = 100THE MACHINE
Ten minutes.
A published 12-gauge comparison—not Framework's 8-gauge sweater or its undisclosed machine setup—answered a narrower question. Cutting and sewing fell by thirteen minutes; knitting added three. Total time moved from 123 minutes to 113. Most of the saving came from cutting and sewing that no longer had to happen.
THE CALENDAR
At least five months.
An at-least-five-month shorter lead time can move the commitment closer to the selling season, when a brand knows more. In one USITC case, Boathouse Sports reported under a month for domestic orders and at least six months from China. Framework has not published a comparable result. Founder Denver Rayburn describes the same wager as holding value in material until demand clarifies the product.
AN ILLUSTRATIVE MODEL · NOT FRAMEWORK DATA
What is waiting worth?
I priced the wait with an illustrative inventory model. Set six-month demand uncertainty to a 30 percent coefficient of variation. A one-month source can cost up to 5.42 points of retail more than a six-month source before expected profit falls below it. The allowance drops to 2.72 against three months and 1.53 against two.
THE CONTRACT
Somebody still owns the yarn.
Framework advertises “zero inventory risk.” Under its standard public terms, customers generally prepay yarn and can remain liable for unused material or unshipped goods. Flexible yarn may be easier to reuse than finished size-and-color inventory, but the customer still carries material risk.
VS 6 MONTHS5.42
VS 3 MONTHS2.72
VS 2 MONTHS1.53
FINISHED SKURISKPREPAID YARN
THE OPEN QUESTIONS
What this sweater cannot show.
To judge the economics, we still need five measures: matched unit cost, first-pass yield, utilization beside lead time, full-price sell-through, and the change in markdowns and stockouts.
I’m keeping the sweater. I’d still like the spreadsheet.
- 01
- 02
- 03
- 04
- 05
REPORTING NOTES
What the pictures mean
The archive date keeps this design experiment below newer posts; the reporting and data were completed in August 2026. No 2026 source is presented as information available in 2024. Framework's product page describes an 8-gauge, one-piece cotton sweater made in its Los Angeles factory. The garment and its “Made in USA” label establish the object; they do not establish factory cost, scale, or customer results.
The disappearance graphic plots annual observations from BEA real garment purchases and BLS real apparel output, labor hours, and labor productivity. Each series is rebased to 1990. Purchases include imports, so the comparison shows divergence—not import share.
The 123-to-113-minute comparison is an illustrative 12-gauge case in Joel Peterson's 2012 thesis, not a test of Framework. SHIMA SEIKI introduced whole-garment equipment in 1995 and sells it globally; Framework has not publicly identified its machinery. The lead-time comparison comes from Boathouse Sports in a 2018 USITC staff paper. Boathouse averaged 17 garments per order. It is one case, not an industry estimate.
The inventory model assumes a retail price of 100, landed offshore cost of 35, end-season recovery of 10, lognormal demand, no financing, and forecast error that declines with the square root of lead time. None of these are Framework estimates. The two- and three-month cases show that a faster source captures much of the modeled timing value; this model contains no observed nearshore costs or lead times. Denver Rayburn's essay on product complexity motivates the late-binding interpretation, but its formula is a heuristic rather than a measured law. Framework's manufacturing agreement supplies the yarn-payment terms discussed above.
Model table and reproducible files
| Six-month demand CV | 20% | 30% | 40% | 50% | 60% |
|---|---|---|---|---|---|
| Versus 6 months | 3.63 | 5.42 | 7.13 | 8.73 | 10.20 |
| Versus 3 months | 1.82 | 2.72 | 3.59 | 4.43 | 5.22 |
| Versus 2 months | 1.02 | 1.53 | 2.02 | 2.50 | 2.95 |
Download the scenario data or inspect the calculation.