A Robot of the Week, Revisited

In a robots of the week post a while back, I mentioned two companies that are attempting to automate the apparel-production process. Recently, one of these companies, Softwear Automation, announced that its Sewbot product is now also available on a rental basis under the banner Sewbot as a Service.  (SaaS, playing off the acronym for the currently-hot field of Software as a Service.)  From the SaaS announcement:

From 1994 to 2005, the United States lost more than 900,000 textile and apparel jobs to offshoring.

Fast-forward to 2018. The pendulum is swinging back and textiles are returning as lean, highly automated, environmentally conscious production facilities. Within the last six years, there have been significant announcements by foreign-owned textile companies investing in the United States, with site selection choices clustered in the Southeast including the first Chinese owned Cut Make Trim factory in Arkansas.

Despite this industry reversal, the seamstresses are not returning. While the knowledge can be shared to upskill workers, people don’t have the  desire to work in a traditional textile factory.

To solve this and accelerate the growth of US based textile manufacturing, Softwear Automation is announcing SEWBOTS-as-a-Service, a rental lease service to allow manufacturers, brands, and retailers to source and manufacture here in the US at a lower cost than outsourcing and with greater predictability and quality. While we understand the benefits of “Made in America”, the focus of this program is to offer US textile manufacturing more control, greater margin, faster turn times and less inventory.

The rental rate for Sewbot is quoted as starting at $5000/month, which comes to $55/shift for a three-shift operation.Softwear is also now offering production-rate estimates for various kinds of textile products. For microfiber towels, a single operator supervising 6 robots can make 2880 towels in an 8-hour shift, compared with 223 towels for a single operator performing traditional manual activities.   Other product types which the company sees as suitable for Sewbot automation include mattress covers, pillows, automotive floormats, t-shirts, and shoes (uppers).

Most aspects of the apparel supply chain have long been highly automated: indeed, the mechanization of spinning and weaving was the hallmark of the Industrial Revolution.  The sewing process, however, has remained stubbornly labor-intensive, largely because the flexible nature of fabric makes it hard to handle mechanically.  Softwear Automation’s solution involves the use of machine vision for precise fabric positioning.  This article at IEEE Spectrum explains a little bit about how it works.

Depending on how well these systems turn out to work in practice, and how the technology evolves, they may turn out to be not only the robots of the week, but the robots of the year or even the decade.  Apparel-making is a vast industry, concentrated in nations which are not-so-well-off economically, and employs a large number of people. A high level of automation would likely result in much of this production being relocated closer to the markets, thus saving transportation costs and shortening supply cycles.  The consequences for countries like China, Bangladesh, and Sri Lanka could be pretty unpleasant.  For the US, the onshoring of the work would seem clearly to be beneficial.

I don’t know enough about the industry to analyze the economics of Sewbot vs low-wage-country production in any depth, but back-of-the envelope for one product type (the towels) suggests that on a pure direct labor cost per unit basis, a US-based Sewbot can still be undercut by human labor rates below about $4/hour.  (Calculated using the rental rate:  for many companies, purchase may offer better economics.)  But production isn’t the only factor in the product cost equation, of course, and in many situations proximity to end markets will be of considerable value: especially simpler inventory control and faster response to style changes. And a Made in the USA label is surely worth at least something.  Also, the economics may be different for some of the other product types…for the t-shirts, the company is citing a unit cost of $.33 for US-based production using  Sewbot…this compares with something around $.22 for a country such as Bangladesh, and is probably cheaper than China at the current wage rates.


Most likely, some product types will move to automated production in the US and other high-wage countries fairly rapidly and some will take a lot longer; also, if the trajectory of this form of automation is similar to that of many others, then unforeseen problems will occur and will slow things down. Developing a sales and support capability for the robots in third world countries will also take some time.  And high-end products are apparently a long way off from being robot-produced.  So, an Apparel Apocalypse for the low-wage countries probably will not occur in the short term.  (Here’s one company that is remaining focused on cheap human labor rather than robotics, at least in the near term.)    It would certainly be wise, though, apparel-manufacturing-intensive countries to focus on developing broader employment bases.

While there has been much valid concern about bad working conditions–sometimes positively dangerous conditions–and extremely low wages in the offshored production facilities serving major American brands, it is also true that for many people, work in this industry has been better than their other available alternatives.

In her book The Travels of a T-Shirt in the Global Economy, Pietra Rivoli cites a North Carolina woman from 1899:

We all went to work in the Amazon Cotton Mill and we all worked there all our lives.  We were all anxious to go to work because, I don’t know, we didn’t like the farming.  It was so hot from sunup to sun down. No, that was not for me.  Mill work was better…Once we went to work in the mill after we moved here from the farm, we had more clothes and more kinds of food…And we had a better house.

She follows this passage by quoting a present-day Chinese woman on the advantages of factory work versus her previous farming life:

(Farming) is really had work.  Every morning, from 4am to 7am, you have to cut through the bark of 400 rubber trees in total darkness.  It has to be done before daybreak, otherwise the sunshine will evaporate the rubber juice. If you were me, would you prefer the factory or the farm?

If I were the leader of, say, Bangladesh, I would be studying the capabilities and economics of the apparel-automation technology in depth, and probably losing some sleep over it.

Overall, I continue to believe that the employment/productivity effect of today’s automation technologies will be a continuation of long-run trends rather than a sharp upward break.  See my post series Attack of the Job-Killing Robots for some history of previous automation technologies and associated automation panics.  Also, Peter Gaskell’s 1836 book, which I reviewed here, provides an interesting contemporary view of the original Industrial Revolution.

Here’s a good WSJ video on the automation of apparel production; also this one, which is specifically focused on Softwear’s product.  And here’s a short Fox interview with Softwear CEO Palaniswamy Rajan.


6 thoughts on “A Robot of the Week, Revisited”

  1. Textiles have strange quality variance issues that I don’t understand–I have work shirts I’ve worn for 15 years that are still in good shape, and I’ve had ones that literally frayed to pieces within 6 months. My understanding is that women’s clothes are typically made quite cheaply because most women won’t wear the same thing for years anyway, due to fashion changes. It would be interesting to know more about how robots might either standardize quality, or otherwise impact this issue…

  2. Brian…one factor in the quality valuations might be that brands use a lot of different subcontractors to do work on the same product…although I’d think that use of the same fabric would be required for all of them.

  3. I’m not sure how any economic forecast can reasonably include the effects of automation (or robotication) moving from factory to home.

    Historically the sewing machine may provide an example, but in much more recent times the PC, fax, scanner, copier, and color-printer all have similarly revolutionized both office and home. Coming right up we already see hints of how 3-D “printers” and other fabrication systems are moving in. Any wanna-be craftsman can set up a corner of the garage with welding gear, band saws, air compressors …

    The “criket” style paper cutting machines and rug hooking machines and embroidery machines and sequin applique machines are all selling at hobby shops. Most of the stuff either craftsmen or hobby-ladies actually make is not up to the standards of the factory-made stuff. But that stuff can be made at ALL is new and different and people seem to like doing it — work become recreation, and how does standard economics account for such a change?

    And once in a while some sort of hobbyist develops a product that creates a market — like Columbus marks a new land with his tiny little flagpost. Making single board computers or cabbage patch dolls or bean bag stuffed animals or carved salt lanterns … then fleets of factory worker replace the explorer and the new market is colonized faster than Sooners moving into Oklahoma.

  4. I am terribly suspicious of Hardware as a Service. It smells more like installment sales because we can’t find anybody to do lease financing on our unproven and otherwise worthless technology and no customer will sign a long term agreement. If these guys really have the better mousetrap, and I believe someone will, they can take off. But they have to find the customers who can make money with it. A lot of contingent risk. But bespoke clothing for the masses will come.

    Thanks for trying.

  5. They do not currently seem to be focusing on bespoke clothing or other items….though that seems like a logical extension…but about making mass-market products less labor-intensive.

    Making the Sewbot available on a lease basis seems to me like a rational thing to do in the name of risk reduction. I don’t know how they are financing it, it’s possible that they have a leasing company behind the scenes.

    To be truly analogous to SaaS as the term is used in the IT industry, I think you’d need a centralized facility operated by a service provider such that customer companies could send in the files describing the product and the quantities desired, and get the products back, shipped either to their distribution points or drop-shipped to retailers or end customers. There are businesses now that do machining and 3d printing on this kind of basis. Don’t know whether it’s practical with the current Sewbot or not.

  6. No sure what the current status is, but as ofvthe mid 80s, as i understood it, some absurd amount of inexpensive SOCKS was coming, not from China, not from Malaydia, but ALABAMA.

    There were machines for making tube socks that cost about 10k. They were supposedly fairly reliable — you just stuck one in your garage and regularly fed it spools of yarn, as it dumped completed socks in a box.

Comments are closed.