Hacker News new | past | comments | ask | show | jobs | submit login
Startups Can’t Manufacture Like Apple Does (2014) (bolt.io)
341 points by bootload on May 28, 2016 | hide | past | favorite | 81 comments

This is a good list. Let me add one to it. I manufacture products in China and I consistently see new designs for injection molded parts where the designer has unrealistic expectations for the tolerances that can be achieved on molded parts. While the tolerance we can achieve will change proportionally with the size of the injection molded part, don't design parts that rely on precisions of 0.01mm. Hard to do this unless you are Apple.

Scott Miller of the Bolt team has a great video on injection molding.

Design for Manufacturing Course 5: Injection Molding - DragonInnovation.com https://www.youtube.com/watch?v=jx5_gO9LTf8

After watching it I decided to use commodity enclosures with custom milling and avoid injection molding for now.

This video https://www.youtube.com/watch?v=RMjtmsr3CqA was a series of revelations as to how injection molding works.

Another interesting resource is "Guerrilla guide to CNC machining, mold making, and resin casting" @ http://lcamtuf.coredump.cx/gcnc/

How did you find your commodity enclosures? It seems like there are a lot of random manufacturers out there making ugly enclosures.

I prefer Takachi enclosure. Some of their plastic is translucent so your LEDs will shine through the enclosure. You can see an example here: https://hackaday.io/project/6357/gallery#67d3556e38f0b46c486...

I am fresh out of college. I work for a very big OEM and make drawings for plastic parts. i have no clue of the tolerances which suppliers are capable of holding. how does one learn that? I usually tolerance based on our objectives and source a supplier who says yes.

You call the supplier, tell them you know nothing, and ask them to help you understand what they're capable of. You should find a mentor, this is the kind of retroactively obvious advice I got from mine.

Where does one find a mentor?

Find the guy whose work you like the most at your company and bother him all the time. Most good technical people love to help, to be asked questions, and ultimately to be valued for the experience they've worked hard to get. In the end mentorship is something that you seek out and get for yourself. It's rare that someone will actually attach the title mentor to a such a relationship.

No more than two doors away.

any time you make something, you find that out from the people that will make it. usually they can also improve your designs to make then cheaper to produce as well.

... I thought that was standard teachings in mechanical engineering?

Hey - this is a long shot, but hoping to connect over possibly moving PCB fabrication, assembly, and test over to Asia from the US. Let me know if it'd be Ok to chat. If not, that's cool too.

Hey - we've been performing PCBA in China for our own commerical project and some open source projects for a few years. Outside of that, have been working in traditional manufacturing in China for many years. Feel free to reach out, I'd be happy to help or point you in a direction.

logistics(a)parklink.asia (apologies for the non personal email address, don't want to put my personal on the public forum)

His contact info is in his profile. Hopefully you were planning to pay for his consulting services. Moving fabrication and supply chain to another country doesn't sound like something solved with a brief chat.

Thanks for the heads up on the contact info.

I remember coming across this some time ago on HN and being really interested by the white plastic bit. It clicks into place Apple's ability to set itself apart visually during the white earbuds era of iPod advertisements. PCs were black and gray and beige, Apple was a shining white.

And no one else could mimic their style, since it was just too difficult.

This makes me think of Peter O'Toole wearing that shining white ... garb in the middle of "Lawrence of Arabia" in defiance of all the filth in the desert.

Lean did it because he could and because it reinforced the messianic oveure into which the character had fallen. And because it looked amazing in all that light. It still does. The failed theater at which I first saw this on posters kept that image for ten years or more.

But understood in the context of all that history, we gain that Lawrence was the agent of so much destruction. Destruction that we barely understand now.

So I hope all that gleaming white was worth it. Because I'd hate to think of the price we paid to get away from putty color.

Deserts are dry, not filthy. White is the traditional colour for a tunic worn in those areas, according to this: http://www.pbs.org/lawrenceofarabia/revolt/clothing.html

Where does this "filth" idea come from?

The tunic got quite dirty later in the movie, and the character kept calling the desert "clean". Deserts have considerable quantities of dust.

My take on the movie is that this was being communicated as a character detail of Lawrence, part of the messianic nature of the character.

Deserts are one of the cleanest natural environments. It's pretty easy to keep white stuff white there in comparison to grassland or forest (just to pick two at random)

Well, there is a whole lot of awful dust in the desert.

The earbuds (and the translucent iMac that predates it) aren't the best example of "no one else could mimic their style".

When the first iPod came out in 2001, Apple wasn't the behemoth it is now; its market cap was around $7 billion (http://www.wikinvest.com/stock/Apple_(AAPL)/Data/Market_Capi...). At that time, many companies could have paid as much attention to design as Apple did.

"And no one else could mimic their style, since it was just too difficult."

The magic of Apple isn't their design, it's the ability to re-create and distribute good design cheaper than Braun. [0]

[0] https://twitter.com/HannuRytkonen/status/731511690911711232

That's just designer nods to earlier work/aesthetics, and happen all the time in industrial (and graphic) design.

Especially since we're talking about 20 and 50 year old designs at the time -- it's not like copying this year's hot product to sell a cheap imitation. Totally different era, totally different functionality, totally different devices -- but borrowing the old visual style.

The magic of Apple is the functionality and thinking behind their devices. There are tons of same looking competitor devices that get it wrong.

E.g. the whole idea behind the unibody design is not the seam-less look, it's the sturdiness and tolerance of the materia. Competitors copy the overall look and color, but leave out both the seam-lessness AND the sturdiness.

Or take touches like the mag safe port, the touch trackpad, the no-button trackpad, the first to come out with the hi-dpi display -- plus the thinking that goes into ensuring good battery life, small weight, and thinness along with said sturdiness. Competitors often leave out one or more of these -- which are all important for a laptop.

Where Apple fails, OTOH, is their occasional cheapness, still selling embarrassing starter configurations (memory, disk-size wise). Of course if they do sell these, and buyers have no problem, kudos to them, but one would expect more, since that's too is part of the whole experience, even if the buyer thinks 128 GB SSD is good for them.

My wife got (on my recommendation) the starter config Air, and it's never occurred to me that there was anything inadequate about it (to be fair, two years in, she ran out of disk space and had to move a stash of photos to an external hard drive, and for me, I can't stand the tiny display, but it works for her, so that's great).

I use a 128GB SSD, have done for nearly two years.

This is my main machine (a laptop) for work and home.

Never had a problem and I don't expect to have a problem before the laptop fails.

As for the entry level MacBook Air, for my work the CPU is a little slow, and I would probably boost the RAM for the occasional "lets open everything", but I see no reason to increase the SSD size unless it is trivially cheap. Each to their own though.

I genuinely can't tell how much of that is supposed to be a joke.

The H&M of electronics.

Any manufacturing experts can help us understand why white plastic is difficult?

White plastic highlights defects including contamination with other colors (or materials), burn marks, and shrinkage. It is common for one injection machine to be used for many colors and types of plastic, but since white is such a pure color, it is very easy to see swirls or traces of other colors in the part. Even different batches of white plastic can cause unsightly marks; getting some nylon in your ABS part will certainly ruin it. Burn marks (which are dark) obviously stand out more against white than any other color, and are generally caused by either the mold or plastic being too hot. Shrink marks are caused by uneven cooling of the plastic because of machine settings or part design, and are most visible on white because of uneven reflections and small shadows. Even tiny scratches can ruin the glossy finish of a smooth part.

Since Apple has huge volumes, they can dedicate machines to injecting a single mold, and ensure that plastic colorants are matched across batches; the rest of us (with volumes of less than ~100-500k/yr) have to settle for grey.

If anyone wants to chat about part design and manufacture, let me know.

Is this still a thing?

I ask in seriousness, as commodity smartphones with small production runs, less than 100k, and earphones, even USB cables, now exist at very low prices, production runs and seemingly margins.

Or has the mold part got easier, allowing different shapes/sizes without burning, and the colourants / dedicated white production line now standardised?

Smartphone enclosures themselves are usually quite simple, and if you look closely, you will see that most small devices are designed much more simply than Apple products. Apple dedicates more effort into modeling plastic flows, shrinkage, and other manufacturing issues than most other OEMs. Apple also dedicates more effort to refining production techniques than most other manufacturers. Companies like Samsung and HTC have volumes almost as large as Apple's, and are not far off.

Molding parts has gotten a little easier over the last 10 years. Molding machines are more easily tuned, and CFD tools have gotten much better and easier to use. The Bolt post is addressing new companies with 1k-100k/yr volumes and less expertise than Apple, and the post is largely correct.


After reading further comments I decided to compare all of the white plastic devices I have, and they're all a often noticeably different colour, even on different parts of the same device.

I think several of us are interested in some insight into this world. It seems like you have it and are able to convey it in an easy to understand, compelling way.

I suggest you write a series of blog posts that explain interesting bits and quirks of manufacturing processes that the rest of us outside of the industry don't really think about.

It's a way that you can help inform the world, and also capitalize off of your knowledge! Please consider it :)


We manufacture large parts in Apple white, and it's not caused a huge number of problems. At least not yet.

What about white with texture? Definitely understand white shows artifacts on glossy textures very readily, but we've had decent luck with Mold-Tech textured whites at admittedly very small volumes (1-5K).

Color match is still rough though.

Texture makes any color much easier to deal with, as it makes mixed colors, materials, and shrinkage less visible.

Really glossy any colour is difficult. Glossy black shows sink much worse than light colours in my experience. Texturing fixes so many difficulties in injection molded parts.

I certainly am! Please tell more. :)

Black makes it easier to hide defects. White makes any tiny issue show up very clearly, it's hard to get the exact shade of white right, and you'll need to run lots of plastic through the machine to make sure no little bits of other colors are still waiting in there to ruin your new pieces.

I'd rather go back to what came before Apple: Next cubes. They were sleek, black cubes with better usability and performance than many competitors. Xbox also did great-looking, black boxes. So, one can still differentiate with a good form of black.

I ranted a bit on this as well on Hackaday http://wp.me/pk3lN-PBv . Apple and Foxconn just have an unimaginable amount of capital. However, I don't think hardware designers should despair. There are many ways to design things, and constraint is the mother of innovation. As the author mentions, it's entirely possible to make products just as appealing with a fraction of the cost.

Great post. I don't normally read Hackaday because I can't stand white-on-black text (there's a reason nobody else does that anymore) but your post and your earlier one from March 7 were well worth it.

One particular bit of this makes no sense:

> What happened when Apple wanted to CNC machine a million MacBook bodies a year? They bought 10k CNC machines to do it.

CNC milling scales linearly. If you want to make 1k things per year, you can probably do it with one CNC machine. I know a startup that's using CNC-milled enclosures and that's probably the single easiest part of their production.

Sure, startups won't buy 10k CNC machines, but they won't need them either.

Yes, and here's a CNC machine shop in Germany machining a MacBook body, or something very close to it: [1]

It's hard to believe that Apple needed one CNC machine for every 100 (was 10, oops) MacBooks. The MacBook isn't that expensive.

[1] https://www.youtube.com/watch?v=Ob-RBntcZc8

I think you missed a zero, it'd be 100 a year, and then of course they don't need to buy new ones every year.

I'd imagine they probably use them for more than just MacBooks as well but I'm just inferencing and don't have any real idea.

It's not just Apple vs startups. Large customers like Apple or Samsung have big advantages over smaller phone manufacturers.

Apple can always call Foxconn and tell them that they want even this barely visible detail fixed. Foxconn comes up with number for changes in the manufacturing processes ($10 million for better tooling for example) and it's done. Small manufacturers who have low margins can't justify similar attention to detail.

Or look at phones vs. other appliances that are made for smaller markets. Garmin is the dominant incumbent in portable GPS devices and even their handsets feel like garage-made prototypes compared to even the worst phone, despite being exactly the same technology except for the backlight diffuser film.

> CNC machining is fantastic for prototypes [but] it is not for consumer devices. Figure out a way to cast your metal parts.

Why is that?

I'm trying to build a better quick release plate for DSLR cameras, compatible with Manfrotto tripods but that lets you do a couple of other things (like attach a hand strap to it directly).

I had prototypes made in China with CNC machining, and they are of a very good quality (superb, even, it seems to me).

For production, injection molding is of a reasonnable price but the result would not be the same quality (plastic is a poor choice for this).

Die casting is too expensive for the volume, given it's really a niche product.

So I was thinking of doing short runs with CNC: what' wrong with that option?

There's nothing wrong with CNC for that type of product, and it is necessary as well. Many of the higher-end products in that market use nearly 100% CNC, whereas lower-end mix casting and CNC.

I'm not sure the quoted statement is fully correct, CNC and Casting are processes and each have advantages as well as capabilities the other cannot easily replicate. (Try casting threaded holes -- there are ways to add threading to a casting process, but it's not the same.)

I think the original quote intended to convey the following idea, instead of never CNC: if you have a high-volume product, while CNC machining can effectively handle shaping operations, casting will be less expensive for shaping in the long run, leaving CNC to more difficult to cast operations and final product fit.

In the camera market, there are tons of companies selling equipment in thousands and tens of thousands of units that are fully-CNC'ed and no casting. (See Kessler, Dynamic Perception, CamBLOCK, etc.)

The problem is people build a little widget that would happily fit in a box from Hammond, but they would rather have it customised at great expense and it eats into their profit. CNC is pretty time inefficient and expensive when you get to very large volumes.

The major advantage of CNC though is that if you mess up the first time, you haven't ruined an entire batch. You just send the new model to the company. With molding or die casting, fuckups can easily cost tens of thousands.

If your margin can absorb the cost of CNC - and for photography it's probably a marketing feature - go for it. Plenty of companies sell CNC machined housing simply because they can say it's CNC'd and people go "ooooh".

Oh okay, so it's just a volume thing?

Does it make sense to do metal casting for low volume (under 1k)? It would be economic nonsense anyway...?

I'm mostly familiar with plastic molds, but for durable/production ready molds, you're normally starting off at like 20k per mold. You can usually bring your part to manufacturers (both CNC and casting shops) and ask for some quotes.

Have you looked at Really Right Stuff and Kirk Enterprises, they are machining the parts you are making, basically a sick release plate. But they are now ubiquitous enough that I would hate to see yet another quick release style. Never been a fan of the large ManFrotto/Bogen style. Take a look at the two aforementioned, I think they take a cue from Arca/Swiss, and are some of the nicest designs in quick releases that I have seen.

Thanks for your comment but why would you "hate" to see another design?

We're not creating a whole new system, just an alternative. And as a start we will be Mamfrotto compatible only, since that's what we use. I could be wrong but Kirk doesn't do Manfrotto at all.

Well, the truth is that very few companies can manufacture like Apple does. Between the obsession with design, details, and having the resources to have the manufacturers create entire new production processes just for them. And often, the companies who do the above are usually not consumer facing.

Does anyone know what a "CM" in this context is?:

"Unless you’re a billionaire genius, your product will have noticeable ejector pin marks. A good CM knows how to hide these well. Nearly zero CMs hide them as well as Apple does."

I wondered the same thing. My guess is "custom molder" but that's all it is.

Contract manufacturer

It's pretty interesting to think that quality can actually go up as volume increases. This seems to run counter to general perception that small/hand batches can be of incredibly high quality and once it goes to mass production quality goes lower.

We've never really seen production at Apple/Samsung's scale before and I wonder if this quality curve is something that is all that well understood.

Take a look at the Toyota Production System. A large part of the reason they won (at least temporarily) was that they forced all their suppliers and factories to produce high quality parts.

Then take a look at Six Sigma.

Small batches can be made by experts and thoroughly inspected; but it is prohibitively expensive to scale up artisans and inspections, both in complexity and quantity.

This is not surprising. Cobbled together prototypes and first runs are usually not pretty. It's like the first pancake or waffle, usually not quite right. A machine will certainly be quicker and more consistent once setup correctly.

Perhaps there are artisan niches where a machine cannot do the work, and a human slaves over the minute details, but those are usually luxury items.

" It's like the first pancake or waffle, usually not quite right"

Very much OT, but that's because the pan or iron isn't hot enough. Ever since I started first putting the pan on the fire, then mix my batter while the pan heats up, my first pancake has been as good as every one after that.

(sorry to contaminate the discussion with this, but I found it much more satisfying than I would think is normal when I got a process down that stopped me from looking like an amateur pancake maker, which is why I figured I'd share :) )

My first pancakes 'fail' (no-one minds eating them :) not because of that but because I need to get the feel for the batter spoon in my wrist, get a feel for the cooking time and get a feel in my wrist for the pan & flipping motion. I do not bake them that often so that feeling needs to come first. After a few it's there and then I can bake 100s of the same ones. Which I do at (beer) festivals.

Also sorry to contaminate the thread.

When you manufacture at the scale Apple does, you can justify more expensive tooling as the cost gets distributed over many more parts. It isn't that you can't do the high quality at low quantity, just that the fixed costs make it infeasible.

Same with code.

Abstracted code often leads to less efficient performance. With large volume, every bit of optimization matters, and Apple has that money to squash the last bit out of it.

I bought a Hand ground coffee grinder last year. It's a Kickstarter project and we should actually have received the grinder last year some time, but in the spirit of Kickstarter I'm patient.

I think the challenges they've had are similar to what is touched upon in this article.

They've had a ton of clearance and molding issues, and dozens of design prototypes and mismatching parts. To add to the difficulties, the coffee grinder has moving parts that take a lot of stress, as opposed to just a box with a circuit board inside.

* http://handground.com/

Posts like that are why I come to HN. Even if my field of expertise have absolutely nothing to do with manufacturing, its nice to read about problems and solutions of huge industry explained in simple and clean language.

What are high margin parts? ...asking for a friend

Parts you can sell for much more than they cost to produce

A while ago Andrew 'bunnie' Huang (Remember chumby?) gave a talk on hardware manufacturing in linuxconf-AU. Could not find a link but it covered a lot - why you cant do it like apple or samsung. His book is out now!

Apple outsources its manufacturing.

A small startup can't just get Samsung and Foxconn to build their parts. This is true.

Yes, but Apple still has significant say in how its products are made. In many cases they use their own production engineers to figure out the best way to make things. In other cases, they set insanely hard specifications and use their massive buying power to entice their partners to figure it out.

Source: I've worked with ex-Apple engineers at a company that mostly used Apple's hardware processes.

There was a long article (maybe Vanity Fair or Esquire?) that pointed out how Apple put up 1B to Foxconn for building an Apple dedicated factory where the factory processes would be dictated fully by Apple.

In the lost interview of Steve Jobs, Jobs says this:

I’ll give you an example. When we were building our Apple computers in a garage, we knew exactly what they cost. When we got into a factory in the Apple II days, the accountants had this notion of a standard cost, where you kind of set a standard cost and at the end of the quarter, you would adjust it with a variance. I kept asking: why do we do this? The answer was, “That’s just the way it’s done.”

After about six months of digging into this, I realized that the reason they did this is that they didn’t have good enough controls to know how much it’s going to cost. So you guess. And then you fix your guess at the end of the quarter. And the reason you don’t know how much it costs is because your information systems aren’t good enough. But nobody said it that way.

So later on, when we designed this automated factory for the Macintosh, we were able to get rid of a lot of these antiquated concepts and know exactly what something cost.

If you know what exactly something costs during manufacturing, now you can optimize it.

Not to mention, they were making equipment that made it easier to know such details.

That's definitely part of it, although that's certainly not all of it.

(2014), but still as relevant as ever.


Applications are open for YC Summer 2023

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact