Some thoughts from the field on the causes of contract manufacturing defects. Not trying to be narcissistic, just hoping to contribute to the community in some way.
Hi. Author here. Really sorry about all the issues.
(1) Sorry for the bugs -- misconfigured subscribe tool has been removed and you can use http to access the site as there is an issue with the https config on the new landing page
(2) Graphs have the axes explained in the text. Removing them from the figures was recommended as a marketing hack, but very ill-advised. I'll get the figures updated within the next half-hour or so
(3) Advertisement? Sure. But... also a huge dataset and an interesting finding about the relative perceptions. It's not a scientific journal article by any means (I've written some), but it does outline how the data was put together. If anyone is interested in collaborating on a deeper study, I'm very happy to do so!
You're right in that if the factory wants to misbehave and the owner/GM is a tyrant, you aren't likely to directly impact worker treatment (we have in-house auditors in China that can check on the factory for you if you have that specific concern). Fortunately, the overwhelming number of factory owners/GM's are fundamentally decent people caught up in difficult circumstances.
Typical contracts are structured 30% deposit and 70% when it ships. Margins to the factory aren't 70% so they wind up quite leveraged. Most of the scenarios I saw came out of the factory not understanding or executing on the customer requirements correctly. If they build units and the customer rejects them, they have to cover those material costs so they can repair or otherwise try again. This tremendous financial pressure causes otherwise good GM's to cut corners every possible way they can to try to keep the lights on.
GetScale impacts worker treatment in several ways:
(1) it eliminates waste and errors that result from misunderstandings in customer acceptance procedures. After deploying us in a factory you just don't encounter thousands of bad units because the factory is confirming against the exact acceptance procedure as you go (instead of build a lot, ship, customer receives and rejects).
(2) it is possible to audit your suppliers and if you do that, GetScale provides the ground truth of which workers were working at which stations when and for how long (down to the millisecond!). Since you have to login-logout to do any work our records are very accurate.
(3) it enables you to vote with your dollars. Without invasive monitoring (like GetScale) it is very difficult to determine which factories are "good" because reputation is easy to coopt and places like Alibaba can be bought. GetScale data collection is automatic and the data is held off-shore (in the US). We already see factories in China using GetScale as a selling point to foreign clients since it provides quantitative proof of their compliance and performance. A factory that is willing to be open about its performance and let you document workers on the line is the factory proud of its worker treatment, not the one trying to hide it.
As to studies, I have seen some in both the psychology of management literature and the manufacturing management literature, but as to specific citations I will need some time to dig them out of my research notes.
In any case, I hope I've answered your concerns. If I missed anything, please follow-up!
The owners who are "fundamentally decent" are the ones abusing the workers?
Of the three ways that GetScale "impacts worker treatment", the first doesn't mention workers at all, the second only improves record quality (not worker treatment), and the third depends on foreign buyers to demand good treatment of workers (which there is no evidence they'll do) and does nothing to address the single abuse mentioned in the article, higher water bills for "unproductive" workers.
This story seems set on the idea that increasing productivity (benefiting the owners directly) will increase conditions for the workers. I think it's great to increase productivity, but the link to worker conditions still seems like magical thinking, and as a result I find the linked article deceptive.
Hi HN. I'm Jonathan, one of the founders. I'll monitor this thread all day to answer questions. Thanks for taking the time to read about us! It means a lot to everyone here at GetScale.
Do you think your camera system could (eventually) support some combination of Computer Vision and Machine Learning to identify (some) assembly mistakes in real-time and provide feedback/warnings to the assembler?
My thought is improving yields by reducing defect rates caused by those hopelessly imperfect humans (like me) might be a great selling point for some types of manufacturing.
You are describing a type of machine called an AOI (automated optical inspection). It is a common piece of equipment in electronics assembly lines. But AOI equipment is very expensive (~$250k+) and only works on a specific set of products. For example, it doesn't work well when the product contains soft or flexible surfaces, or has a highly non-planar shape.
GetScale terminals can be interfaced with AOI machines in the factory to collect and store the findings along with our human-led inspections and catalog everything by serial number and we do have ambitions to add AOI-like capabilities to our hardware down the line.
Sounds like you have a project in mind. What are you working on? ;-)
I was trying to avoid running off into the endless weeds of details and
acronyms and talking, but oh well... ;)
I'm not very familiar with the more advance microscopy based AOI systems
used on silicon and chip level manufacturing (e.g. wire bonding checks).
I'm also not familiar with advanced AOI used in robotic assembly. On the
bright side, I am familiar with the comparatively "simple" form of AOI
used on PCB assembly lines.
With PCB's most of the AOI identification and spatial orientation issues
are solved through Fiducial Marks [1] and pattern recognition. Some PCB
AOI systems even read/record chip/board lettering with OCR, as well as
handling Bar and QR codes. Most of the rules and routines controlling
the AOI are consistently derived directly from the PCB design files, so
there is little need for custom setup programming on each new PCB run.
That last point is critical; The rules and routines are automatically
and consistently _derived_ from the authoritative data sources. Manually
writing every tiny step in the AOI automation routines for each PCB
design would never be time/cost effective, so it has never been done.
Given sufficient example 2D images, 3D scan models, or CAD files, doing
the object recognition and spatial orientation with OpenCV is reasonably
straight forward, even without fiducials. The extremely tricky part will
be automatically deriving rules and routines for the GetScale AOI-like
capabilities from authoritative data sources. If the human assembly task
is to attach the top cover and secure it with four screws from the other
side, you should _never_ need to manually write the AOI routine for
checking the existence of a screw in each of the four corners when
viewing the bottom of the product.
I'm sure you're really busy, so I think I'll stop here. ;)
If you want to know about the projects I'm working on, email would be a
lot better for me. My address is on my HN profile page. Also, I spotted
a few places where your website could use a little TLC, and it's better
to point them out through email rather than publicly.
We've discussed this possibility of this and I would really like to make an attempt when we have more data!
The practical application is this is sort of an 80/20 proportion though. Currently we see fairly significant improvements in some situations just by the system being there. Factories will pre-test units before passing them over to our system in order to demonstrate better quality, when prior to installation they weren't testing a majority of units at all.
There are metric-truck-loads of existing research on behavior changes
due to known surveillance. Though they might be pre-testing units
off-camera for the sake of giving a good "demonstrating quality" stage
show while being recorded by your system, there's also the effect of
better behavior when knowingly being monitored. Either way, you win. ;)
The problem with pre-test is it can both decrease defect rate _and_
decrease throughput rate. Depending on how much each decrease is, you
can actually decrease yield/time rather than increase it. This exact
problem exists on PCB manufacturing. It's certainly possible to pre-test
every component as it comes off the reel and before it's used on the
PCB. The component pre-test does decrease defect rates of the finished
boards, but it also decreases throughput rate, so it can negatively
impact yield/time. The decision then becomes an investment/accounting
problem. Often, post-mfg circuit/component test automation and reworking
failed boards makes more sense financially.
As far as I know, no one does hard or soft real-time verification of the
human side of manufacturing and lab experimentation, but _my_ knowledge
of the current state of the art is admittedly outdated. Using CV and ML
to spot human manufacturing errors (or inefficiencies) seems feasible if
you throw enough compute power at it, but whether or not it's practical
and financially viable is another (more important) question altogether.
Though CV+ML may not be reasonable in the real world, I agree with you;
it sure does sound like fun to try.
My company, GetScale, offers cheap fixed-price contract work in design-for-manufacturing (identify and avoid the kinds of mistakes mentioned in the article) and going forward my overseas employees can find and monitor suppliers to make sure things run smoothly (like enforcing your quality standards). We only charge a few dollars per unit for that.
I'm an entrepreneur, so it all sounds great for me, but why would investors go for this? It seems like they give up a lot of down-side protection: (1) No ability to convert or abort in the absence of a QFE, (2) no more first creditor protection -- if the company goes under, but also has outstanding loans, investors don't participate in a share of the liquidation proceeds as they would as debt holders (3) no interest = less equity at conversion? Am I wrong here? What am I missing? Thanks YC for your ongoing efforts!
I'm assuming that the theory here is that the good investors are more concerned about being in on the next Snapchat or Airbnb, and a lot less interested in bolstering downside protections that only apply if an investment is going to be one of the unproductive ones anyways.
Meanwhile, the good companies aren't going to be likely to entertain financing on anything but terms like these, so fighting them just incurs an adverse selection penalty.
The same thing seems to have happened with convertible debt, which was preceded by financing mechanisms that were way, way more onerous for entrepreneurs.
Completely agree with you. The article seems to make the point that investors would welcome these changes, when in reality, we will be forcing these changes on investors. That was precisely my experience with the series AA.
With the progressive reduction in the cost of creating a startup, there is a shift in the leverage fulcrum toward the entrepreneur away from investors. This is a milestone that marks this progression. Yes it's marginally worse for investors, but honestly the investors who would care about such minimal edge-case benefits typically don't understand how startups work, and are thus not the ones you want anyway.
Hi from Jonathan (cofounder). Thanks for the awesome detailed feedback. We hear you and are already working on (or have it on the near-term roadmap) almost all of this. I apologize for any frustration thus far. I really do.
Perhaps this will help until then: (1) You can click anywhere that isn't the splash screen (any dark area) to dismiss it. (2) You can jump straight to parts by URL https://circuithub.com/parts or (3) by the big button on the splash screen in the lower right.
Thank you! I hope you will. Things are moving really quickly internally. In other news, we fixed the splash screen. It now permanently dismisses so you shouldn't see it too often. That was our intention we just didn't quite get it right the first time ;-)
I confirm only seeing the splash screen once more. Thanks. :-)
I feel silly even suggesting this, but have you considered some kind of 'like button' and a way to share comments on parts? I swear, I've never asked for such a feature before, I don't even have a FB account! :-)
Is there somewhere else can we circuit dreamers and designers chat about parts by number? For example, this search https://duckduckgo.com/?q=555+timer+comments doesn't turn up much for arguably one of the most well-known part numbers.
