Lessons In Small Scale Manufacturing From The Othermill Shop Floor

Othermachine Co. is not a big company. Their flagship product, the Othermill, is made in small, careful batches. As we’ve seen with other small hardware companies, the manufacturing process can make or break the company. While we toured their factory in Berkeley California, a few interesting things stood out to us about their process which showed their manufacturing competence.

It’s not often that small companies share the secrets of their shop floor. Many of us have dreams of selling kits, so any lessons that can be learned from those who have come before is valuable. The goal of any manufacturing process optimization is to reduce cost while simultaneously maintaining or increasing quality. Despite what cynics would like to believe, this is often entirely possible and often embarrassingly easy to accomplish.

Lean manufacturing defines seven wastes that can be optimized out of a process.

  1. Overproduction: Simply, making more than you currently have demand for. This is a really common mistake for first time producers.
  2. Inventory: Storing more than you need to meet production or demand. Nearly every company I’ve worked for has this problem. There is an art to having just enough. Don’t buy one bulk order of 3,000 screws for six months, order 500 screws every month as needed.
  3. Waiting: Having significant delays between processes. These are things ranging from running out of USB cables to simply having to wait too long for something to arrive on a conveyor belt. Do everything you can to make sure the process is always flowing from one step to another.
  4. Motion: If you have a person walking back and forth between the ends of the factory to complete one step of the manufacturing process, this is wasted motion.
  5. Transport: Different from motion, this is waste in moving the products of each individual process between sections of the assembly.
  6. Rework: Get it right the first time. If your process can’t produce a product that meets specifications, fix the process.
  7. Over-processing: Don’t do more work than is necessary. If your part specifies 1000 hours of runtime don’t buy a million dollar machine to get 2000 hours out of it. If you can find a way to do it with one step, don’t do it with three.


The first thing that stuck out to me upon entering Othermachine Co’s shop floor is their meticulous system for getting small batches through the factory in a timely manner. This allows them to scale their production as their demand fluctuates. CNCs and 3D printers are definitely seasonal purchases; with sales often increasing in the winter months when hackers are no longer lured away from their workstations by nice weather.

As the seven sins proclaim. It would be a bad move for Othermachine Co. to make too many mills. Let’s say they had made an extra 100 mills while demand was at a seasonal low. If they found a design or quality problem from customer feedback they’d have to commit to rework, potentially throwing away piles of defective parts. If they want to push a change to the machine or release a new model they’d either have to rework the machines, trash them, or wait till they all sold before improving their product. Even worse, they may find themselves twiddling their thumbs waiting for their supply to decrease enough to start manufacturing again. This deprives them of opportunities to improve their process and leads to a lax work environment.

One way to ensure that parts are properly handled and inventory is kept to a minimum is with proper visual controls. To this end, Othermachine Co has custom cardboard bins made that perfectly cradle all the precision parts for each process in their own color coded container. Since the shop floor is quite small, it lets them focus on making spindle assemblies one day and motion assemblies another without having to waste time between each step. Also, someone can rekit the parts for a recently completed step easily without interrupting work on the current process going on.


It’s hard to define what’s over processing and what isn’t. My favorite example of what isnt, and something I’ve fought for on nearly every factory floor I’ve worked on is proper torque limiting screwdrivers. They’re a little expensive, but they are a wonderful tool that helps to avoid costly rework and over processing. For example, let’s say you didn’t have a torque limiting screwdriver. Maybe your customers would complain that occasionally a screw came loose. Now, one way to solve this would be the liberal application of Loctite. Another way would be an additional inspection step. Both of these are additional and completely uneccessary steps as most screws will hold as long as they are torqued properly.

In one factory I worked in, it was often a problem that a recently hired worker would overtorque a screw, either stripping it or damaging the parts it was mating together. A torque limiting screwdriver takes the worker’s physical strength out of the equation, while reducing their fatigue throughout the day. It’s a win/win. Any time a crucial step can go from unknown to trusted with the application of a proper tool or test step it is worth it.

Another section where Othermachine Co. applied this principle is with the final machining step for the CNC bed. The step produces a large amount of waste chips. Rather than having an employee waste time vacuuming out every Othermill after it has gone through this process, they spent some time designing a custom vacuum attachment. This essentially removed an entire production step. Not bad!

IMG_0396 (2)

With the proper management of waste it is entirely possible to save money and improve a process at the same time. It takes a bit of training to learn how to see it. It helps to have an experienced person around in order to learn how to properly respond to them, but with a bit of practice it becomes a skill that spreads to all areas of life. Have any of you had experience with this kind of problem solving? I’ve really enjoyed learning from the work stories posted in the comments.

44 thoughts on “Lessons In Small Scale Manufacturing From The Othermill Shop Floor

  1. “Don’t buy one bulk order of 3,000 screws for six months, order 500 screws every month as needed.”

    And pay too much for your 500 screws, increasing the cost of your product.

    I do agree about not buying too many parts in advance, but screws are the worst possible example because 3000 screws doesn’t take more space in a factory than 500 screws. It’s just a taller and heavy container.

    1. Depends on how you negotiate your contracts. I specifically chose screws because most people think it’s only a taller and heavier container rather than a way of thinking that often ends up locking up thousands of dollars in inventory that isn’t actually doing anything.

      For example, what if you drop 10k on springs for the year, saving a penny a spring. Sure you saved ten cents but now you find a month in that half the springs aren’t treated properly. If you don’t have a working contract with a supplier. You check per shipment but the handful of sample parts at the top of the box were all good.The standard contract says accepting delivery is accepting the goods. So, Without having moving inventory then it’s super likely you’ll end up eating 4990 dollars in bad springs plus the cost in support, testing, engineering, re-ordering, etc. This sort of thing happens all the time.

      Then you’re left with an even worse choice. You’ve now estimated your BOM cost and final cost of the product on a price you got for a year’s worth of parts on every part of the device because that’s the way the factory operates. Well, you can either change the price of your product to reflect your new overruns from inventory failures(, redesigns, optimizations, etc) or you can eat into your margin and profit.

      Plus, what happens if you find out one month that switching to torx screws will save you five minutes per product and 1% rejection? Do you now hold off on the change until you’ve torn through your inventory? No you toss it away and eat the loss. Imagine when it’s 10,000 stampings, or three thousand bearings…

      1. The “seven wastes” that you list should be tacked on every startup door – it’s a good starting point for thinking about how to put this together.

        Unfortunately thought article is spot on in what startups should be doing (and a very good example of one that’s doing a good job for its size), my guess is that a lot of people considering (or involved in) a startup will look at it, go glassy eyed and panic because something has to be shipped, answered or assembled in the next 30 seconds…or else.

        In my experience working with a different industrial sector, the problem isn’t so much the specific number of screws (or CPUs) but the capacity to consider all of the pieces and to start with a workable algorithm (often some flavor of JIT/KanBan/Lean) for managing it all that it doesn’t soak up one of the more-precious resources – time and personnel; something that small companies and startups are notoriously short of.

        A bare-bones startup will be shredding itself just getting the product out the door, and they intuitively know that if they’re short on screws or some physical item *everything stops*, so excess inventory begins to accumulate. The other aspects of lean manufacturing and efficient materials management will creep in quickly, but very few entrepreneurs start out with this as part of their business plan – it’s typically evolved into systems as they recognize the cost and shortcomings of not doing them.

        These same people instinctively recognize good balance of inventory of other things in their life (coffee supply or rolls of toilet paper in the home as an example) but putting a whole inventory management system into play in a system they’ve just invented will fall down the list quite a ways. And the rework/return policy is icing on the cake. Springs can at least be recycled for their metal, but what happens when you not only have too much inventory, but it’s defective and can’t be disposed of easily?. I’ve also done work in the food processing industry – what do you do with an entire tank-car load of chocolate that isn’t up to spec? Quite a problem.

    2. It’s not just that. It’s about having cash on hand. How much money are you actually saving by buying 3,000 compared to that 500 and how long will it take you to use up those 3,000? Then how long until you recover that money from selling product that has those screws in it? What are you going to do when you need cash for something unexpected/overlooked but your money is tied up in a ton of pieces parts that you saved X amount of dollars on? Take out a line of credit? It’s good to save money but you never want to be caught without enough cash to cover unexpected expenses. Screws are a good example. It’s something that people don’t think much of. You could have thousands tied up in small consumables like screws, printer paper, toner, etc. I’ve seen many small businesses fail because they tie all there money up and have nothing left for an emergency. The only other way to get around that is to have enough cash flow or maybe a backup plan where you can liquidate certain assets in a short amount of time.

      1. Indeed, for really low cost parts, it can be better just to have “over stock”. It’s no fun to have your assembly stopping because you are out of a 1 cent part. This happens… and the hours of your labor force are much more expensive then the parts in these cases.

        (I’ve seen it happen at Ultimaker in the early days. But we are no longer in the “small scale” category now)

        1. While it is easy to dismiss the point in this example, inventory is about more than cost; it is about control. Screws cost a penny, but suppose they cost a nickel… or a dollar? The answer is not “so what? I can afford 1000 more than I need”, it’s about developing a system that assures you have adequate without extra. Benjamin Franklin said it well, “Watch your pennies and your dollars will watch themselves”.

    3. You have to take a look at the price break point, *shipping charges* and overheads for ordering to figure if it is worth buying. Sometime on digikey, if you want 75+pcs, you are better off buying the 100pcs due to price break.
      There is probably no much more to buy 3000 screws than 1000.

    4. The formula of optimal order quantity is sqrt(2*D*Ce/C*h).
      D is the expected annual demand (units/year), Ce is the cost of placing and handling an order (shipping charges, one time fees, etc) ($/order), C is the item cost ($/item) and h is the money cost as percentage ($/$/yr), usualy the bank interest rate. The factor C*h is the inventory holding cost.
      This means that as the holding cost increases, order size should decrease leading to more frequent smaller orders. Inversely if the cost of placing an order increase, the order quantity increases too leading to less frequent larger orders

      1. Interesting formula. I assume that minimizes the total yearly cost, trading off ordering overhead against the opportunity cost of capital locked up in inventory (that could otherwise be making interest in the bank).

        However it doesn’t deal with volume discounts, which should logically shift you towards making fewer but larger orders. And it doesn’t deal with any of the factors that Gerrit mentions above (like the chance that you might have to throw out your inventory because of a bad batch or because the design changed), which should shift you towards making more frequent but smaller orders.

        1. There are ways of taking quantity discounts into account with EOQ. Here is a link that covers it: http://www.inc.com/encyclopedia/economic-order-quantity-eoq.html
          EOQ is a good tool but needs some common sense applied. For example you might find the EOQ for packing material exceeds the volume of your building. If you are cash starved you may have no other choice but to buy just enough to get by.
          One thing that is an eye opener for many is the cost of placing an order. It can add up when you consider the hourly rate for someone to place the order, pay the invoice and other accounting steps, and receive and put the order in inventory. Even if you don’t follow EOQ knowing all your costs is important.

    5. I have to agree, especially with “consumables” likes screws & fasteners. I used to keep buying the odd packet of 50, 100, or whatever screws/nuts/washers every few months, and eventually realised that was stupid – I was always having to keep an eye on what I was getting low on etc. So I took the plunge and ordered a shitload of common screw sizes (e.g. 2000x 16mm M3 screws, 5000x M3 nuts, and so on). The bulk of it is sitting happily in a box in the shed, and I just refill my “using” stock from that. Should take a good few years before I need to worry about it again ;-)

      Also, that last photo – is that a piece of timing belt wrapped into a spiral, or is it a brush? I can’t quite make out from the perspective.

      1. Maybe you will worry sooner. Where I live, the invading mice will gnaw a hole in the container and crap all over the screws – because they can and rusty + smelly looks better to them ;-).

  2. Inventory is a blessing and a curse. Buying screws 500 a month or 3000 per half year might either cause you a waste of storage space or give you a discount and a guaranteed supply during those 6 months, not to mention the elimination of 5 times ordering, paying, receiving and storing a new batch of screws. So the trick is to balance. You can (and maybe should ;-)) maintain a larger inventory of cheap standard stuff and a smaller inventory of more expensive specific stuff.
    But even on the specific stuff you will need to have your supply lines secured. It is rather awkward to find that next weeks batch of expensive parts won’t be delivered due to customs taking more time/a power outage at the parts factory/strike at the airport/etc… Ideally you have thought upfront about alternative parts should supply of your components run short.

    1. And it seems that consistency of parts has become a problem for most small parts suppliers. Last time I bought a batch of 3000 screws, the recorder had them slightly different causing a redesign to use them. I really miss the days of having precision small parts that were made with quality in mind instead of profit.

        1. I once had some bolts from China and the socket head would occasionally pop off when tightening, leaving a weird threaded shaft with a mushroom on top. It was as if they attempted (and failed) at friction welding a piece of threaded rod and a stamped socket head together. Really weird stuff.

        2. I’ve seen crap like that often enough, needed some grade 10 50mm M8s threaded all the way to replace original parts, and most local suppliers if they had grade 10, the 50s only had 40mm of thread on. This was in a clamping application, you ended up with most of the thread exposed the other side of the nut, but you needed the length to start them… I had to use 40mms in the end to get thread all the way to the head, and use other means to compress it that first few mm so I could get them together.

          Then 5mm M3 or something around there, ordered some of those, and they were 5mm of thread plus a freaking 1mm dome on the bottom, meaning they bottomed out in my 5mm hole, gah.

          Lastly, a batch of self tapping metal screws, what do you use those on, thin sheet right? Yeah the thread didn’t go to the last millimeter and close to the head, so they’d go in and just spin, needed to buy washers to use them, they were the pan heads you don’t think you need washers for too.

    1. Nope. Gerrit just went by their factory for a tour, so he had stuff to say about it. We thought it’d be interesting to y’all to hear about how a small-scale, hacker-based company works.

      We still haven’t and don’t do sponsored content. We run ads, the old-fashioned way.

      In my personal opinion, the whole sponsored-content thing reflects a lack of editorial integrity and is a tremendously worrying trend in journalism when firms like the New York Times follow suit. Who are you gonna believe, when everything’s for sale?

      (Ironically, Hackaday. Go figure. I hope we can continue this way indefinitely. I’m tremendously proud of us in that respect.)

        1. One only has to disclose things when there _is_ a material or monetary compensation related to an article: the Law.

          There is no such requirement for disclosure when there is _no_ compensation. (Would that be negative disclosure?) I only responded to [J]’s question because he asked it, and the presence of sponsored content elsewhere really grinds my personal gears.

          And so it is with the store. There’s no connection between the fact that we sell Othermills in the store and Gerrit writing this article. No connection = no reason for (negative) disclosure.

          Except to reassure HaD commenteers, naturally. And because we live in a crazy world where sponsored ads are becoming the (dystopian) norm, I don’t mind following up on things like this at all, or being extra explicit.

          There is no influence from the store on the topics we cover, or the projects we pick up for the blog. We (the blog) don’t get money from the store, and certainly not more or less depending on their sales. In short, there’s nothing to disclose, but if there were, we’d let you know.

          Now get back to talking about how many screws to buy!

  3. The criticism of overproduction needs to be balanced with cost reductions associated with scale production.

    I had something like 3 years’ worth of Crazy Clock boards manufactured at one go, but it took that to reduce the unit cost to the point where it made sense to have them manufactured.

    Now, you can talk about the opportunity cost of that much inventory, certainly, and maybe that was the point.

  4. Reducing the number of different parts in a BOM for a priduct or share them amount your products:
    – if you can live with a common value of decoupling caps, pull up, series termination, common uC, chips, drivers, PCB, screws etc for different models etc.
    – If there are mechanical parts that can be rotated/flipped even if that requires drilling the extra holes would still make sense.

    1. I just read an article “Design for Disassembly” that touches on that. Reducing/sharing parts helps a load things like stock, assembly, disassembly, recycling and waste disposal. Good for the manufacturer and good for the DIY-minded user

  5. A very good article as was the previous one on Othermill. It raises important points, but is a bit doctrinaire. Ultimately the requirement is to not do something foolish. I would put buying Chinese fasteners at the head of that list, but that may just be me. I’ve seen too many otherwise good products spoiled by crappy fasteners.

    A very important aspect of JIT is the vulnerability of society to a disruption of transport. What happens when the trucks can’t run because they can’t get parts? And in an inflationary environment, inventory is beneficial whereas in a deflationary environment it’s harmful. If you cannot get what you need when you need it, disaster will follow close behind.

    At the household level JIT advocates buying fasteners in those wretched little bags at the big box stores. I’ll do that, but only if I can’t get a box of 100. In my case, I know I’ll need the extra, just not when.

    At the end of the day, what matters is not how much inventory you do or do not have, but that you understand the costs and risks associated. If demand is very seasonal you may have to produce product and warehouse it in order to retain your workforce. You do not solve that by not carrying excess inventory, but by recognizing it is part of your overhead and must be factored into your selling price. If you imagine that skilled, motivated employees will sit at home waiting for you to call them in, you’re deluded and will get what you deserve. If the answers were simple, no one would fail.

    1. JIT only works if you use a standard design inventory that will not risk EOL during production runs.

      Note a $12 spool of resisters is negligible compared to labour expenses, and inventory prices can fluctuate wildly with the US currency exchange rates.

      The trick is to minimize the number of design part values, and standardize around a few package sizes.
      i.e. The inventory is feeding the machines 24 hours a day, labour is actually adding value, and product sub-components are transplantable.

      The company in the article will not survive more than a few years if it keeps gambling on JIT.
      Reduce… reuse.. recycle… is probably better advice than some kids playing start-up tag.

    1. Free cash flow is what keeps a business alive. That’s the point of JIT manufacturing to maximize it — to not have racks and racks of half assembled product that aren’t yet sellable but have tied up your capital.

      Batch n push for example is a waste because there are bottle necks in every process.

      Besides acting as a commercial for Othermill, I don’t know if this is targeting the right audience to teach about efficient production techniques.

  6. It depends how you pitch it, and to whom you pitch it. Accountants almost always want to minimize stock. Ive seen cases where keeping 2 months stock can reduce that cost to 1/5 (not by, to) but the accountants don’t want it, perhaps theres a marginal case where it makes sense but generally I’d say thats stupid. On the other hand, people that create stuff generally want stock, the argument against stock most likely to resonate with them, would be that stock keeping reduces your agility, you simply can’t implement that change or improvement until existing stock has to be used up first.

    One other argument for keeping stock is that it enables you to design with parts you should otherwise not touch. Some parts are supply sensitive, like all those wonderfully cheap Chinese modules available today. But tomorrow a newer module may suddenly become the flavor of the month and suddenly those wonderfully cheap parts that you have on your PCB suddenly becomes unavailable. I like to keep 3 months worth of stock, of those parts that may suddenly disappear. It is a once off expense, but allows enough time to redesign without disrupting the manufacturing in the event of a source suddenly drying up. Keep in mind, that module cost I am talking about here, is often 4 times cheaper that the components on the module alone. (at the volumes of a small scale manufacturing).

    Another supply limitation is politics. In the early 2000s, tantalum became scarcer than hens teeth. One reason it that the demand increased massively because of cellphones, but also, at that time, the vast majority of the worlds supply came from the Congo. As the conflicts in the area moved to the coltan mining regions, suddenly the supply dried up.

    Personally I don’t design with tantalum capacitors at all, not until the “resource curse” has ended in the Congo and the majority of the people on the ground start reaping the benefits rather than the politically connected few and a few multinational corporations.

    For those unaware of the true cost of the materials that designers specify, I’d recommend Adam Hochschild’s “King Leopold’s Ghost”.
    The only thing more shocking than the history of the Congo is the apathy and unfamiliarity around the world.

  7. This is a tricky issue. My experience is that for hardware if yo can’t buy it in a local hardware store, try very hard to avoid using it, Avoid putting any electronics in a product where obsolescence might be an issue in the product lifecycle or deliveries and when ordering for custom parts consider setup costs. This has been my long experience and was validated by my last job where ALL of these became 7 figure issues, at the same time. That was NOT a fun year.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.