It's an classic problem in manufacturing. Where do you put the in-process inventory? Near where it's made? Near where it's used? Someplace else? What batch size do you use for each part? Bigger batches produce lower costs, but you have to put the stuff somewhere. There's software for optimizing this. But over-optimizing that can be a mistake.
Part of the Toyota system is to deliberately keep sources of confusion away from assembly, because confusion in assembly, where many parts come together, is likely to affect both quality and cost. It's better to have trouble upstream, where there's more buffering.
Tesla is in this mess because their production line is running far slower than planned, and their production planning was for a higher rate. So parts manufacturing and final assembly are out of sync.
JIT just pushes the costs on to smaller suppliers who a large company can effectively bully in the same way big supermarkets bully framers.
This is not how the TPS works (At least not in theory. In practice, bullying may or may not happen - I don't know about any). The idea behind JIT is that you can react to changes in external circumstances faster.
Also, the suppliers should also not have huge stockpiles of parts. Then they can, too, react fast to changes in demand. But this can also cause problems: http://sloanreview.mit.edu/article/the-toyota-group-and-the-...
The optimal batch size at the stamping plant may be quite different from the batch size the assembly plant wants. That means in-process inventory, stored somewhere.
In the scale of auto body parts, dies weigh many tons and cranes are needed to do a die change. Clamping a die into position is complicated. There are quick-change systems. They're expensive.[1][2]
What's probably choking Tesla is that they planned their stamping batch sizes for a much higher production rate than they can currently achieve. Much of the planning for the Model 3 production line was based on the assumption that not much would go wrong. They made hard tooling for parts before using prototype soft tooling to save time.[3] That's a win only if you get it right the first time.
(I once worked in an R&D facility for heavy hydraulic equipment. They had to deal with problems like this. People talked about "setup" a lot. A basic fact of manufacturing is that setup and changes are much more difficult than production. Once you get a production machine set up, it's mostly putting in blanks and taking out product, and that's often automatic. Setup requires skilled people. There's been a lot of talk about "agile production" and a "lot size of one", but the real economies still appear only when you make large numbers of something. Look at the pricing for some on-line service that makes custom things for you. There will be a big price break as the quantity goes up.)
[1] http://www.thefabricator.com/article/stamping/10-common-quic... [2] http://www.thefabricator.com/article/stamping/how-quick-die-... [3] https://insideevs.com/tesla-insider-speaks-out-on-problems-o...
Very similar concept to agile software development — setup everything for rapid iteration.
Tesla’s vertical integration of service centers has an interesting side effect - recalls and repairs are pretty cheap, as the costs of service centers and loaner vehicles they give out are fixed and have been borne already.
I wonder if that makes the company more dismissive of potential issues.
The inefficient dealership network is not something that emanates from Japan, but is Japan Auto adapting to local customs, legal requirements, and standards.
- Honesty is a matter of course
- Make no sacrifice
- Examine quality
- Draw customers without words
- Do not allow your products to be defiled
- Purify yourself
- Do not overcharge
- Sell authentic items at suitable prices
I know this list may sound silly and melodramatic to some ears, but I think people should try harder to listen to lessons from the past—even if the provenance is slightly dubious.
Also, related: https://www.tofugu.com/japan/oldest-businesses-in-japan/
Tesla is somewhat generous about voluntary proactive recalls on their dime http://www.businessinsider.com/tesla-recall-model-x-suv-dont...
With that said, I agree that owning a first-year production model from Tesla is an iffy proposition. Late adopters get a fairly reliable workhorse, judging from owners' forums and anecdotal personal experience.
This is assuming that Tesla's centrally managed monolithic dealer/service network will be run more efficiently than the distributed and competitive dealer networks typical of other car companies.
That assumption runs counter to typical economic thinking that competitive market based solutions will be more efficient.
They are awarded on territorial basis to guarantee some monopoly within a geographic market, and the incentive system (inventory requirements, bonuses) benefits current owners opening in new markets rather than new players stepping in.
Their business model also relies on frequent customer visits for repairs/maintenance and subsequent upsell, so their incentives are somewhat misaligned with manufacturer's. So even if Tesla did a complete 180 on their service centers and started selling dealerships, they might not have many takers considering their maintenance needs and potential revenue https://www.tesla.com/support/maintenance-plans
Lastly, while a network of independent mechanic shops might be an indicator of competitive markets working properly, that market is far from transparent due to information asymmetry between the service buyer and the service provider.
Your wikipedia link on the Toyota Production System says it was developed between 1948 and 1975. So it look 27 years to develop. Working for a startup myself, I'm very aware of Kanban and other TPS-inspired work processes. Sometimes in startup-hair-on-fire-mode that all goes out the window and I think that's where Tesla is in company maturity right now.
If that's how you use it, then you also don't grasp it. The point of a process is that it supports you in hard times. (In easy times, any process will do.) In my view, the essence of Lean is patient incremental gain. Giving in to panic is throwing all its benefits away, and creating both immediate and long-term problems for yourself.
A good counterexample to your theory is Toyota itself. They created the hybrid category from scratch. They are still the market leader. They started by building one. They moved on to a small annual production volume. As they worked out the kinks, they sold more and more, eventually starting to sell in the US. As far as I can tell, at no point did they enter "hair-on-fire mode".
Tesla is basically a really big startup, and has nowhere near the cash reserve to take 6 years to bring the model 3 to market. It it basically get it out NOW, or fold. In 6 years Ford etc. etc. already has a model 3 equivalent for sale and knows how to manufacture. You can't play it so safe in an established market if you want to dive in.
"Getting it out now" means building a very reliable, high-volume production line. If you rush that, you get an unreliable production line, which unless you just ship garbage, will be a low-volume production line or a very expensive production line. Or both.
I understand why "hair-on-fire mode" feels right. One's internal feelings of urgency are matched by dramatic action. The chaos and excitement of red alert gives nobody time for contemplation and lets everyone ignore the uncomfortable feelings that might come from that. But people in that mode make terrible choices, drastically undervaluing the long term.
In software, it's especially easy for us. Our products are rarely life critical, people are used to them being flaky, and we can fix things later by shipping new versions. None of these is true about cars. People used to ship garbage anyhow, which is why the US car market in the 70s and 80s was plagued with problems. And why Toyota and others could make such inroads.
Tesla should absolutely be taking risks, especially product feature and marketing risks. But taking quality and productivity risks in hopes that they can somehow make it up later strikes me as a terrible idea.
1) An app is solid, has awesome QA... but the environment does not. You need to shake out some networking bugs / load bugs, etc.
2) The app is terrible. Every part of it is broke. needs a total rewrite.
To an end user both feel the same, but the expected curve to make it stable is VASTLY different.
In theory Tesla has been in the car business for at least a few years. It MAY be closer to 1. Maybe 2 or 3 vendors are 3 months behind schedule in delivering production machines. Which is bad, but not "we are totally screwed" bad.
On the other hand, it is insanely ambitious. It totally could be a #2 totally screwed bad.
Time will tell.
What's wrong with improving the process as you go, as Toyota almost certainly did?
It's sort of like saying, "Why can't you just clean this enormously messy code base up one bug at a time?" Sure, you can in theory. But it's a very rare project that does it. Even places that do total rewrites often eventually end up with as big a mess as they had before. Why? Because the real problem is the ideas, relationships, and habits that got them the messy code base in the first place. That is extremely hard to change.
> (Ironically, the Fremont plant is prominently featured in “The Machine…” as the locus of the ultimately failed GM-Toyota cooperation.)
> As I watched Tesla’s messy, hiccuping line, with workers dashing in to fix faulty parts in place, my mind travelled back to the Honda plant I had visited years ago in Marysville, Ohio. Clean, calm, everything moved smoothly. I was so shocked by the contrast that I imprudently voiced my concern. That didn’t go over well with my fellow Tesla owners. I was a killjoy, I was calling their choice into question.
So we have an actual first-hand description of the differences between Tesla's and Toyota/Honda's understanding of lean manufacturing. Even in the very same building in fact. And I'm pretty sure the author is describing the production of Model S or X, not the new Model 3. Since those cars have been in production for a while now, there's no way they wouldn't have implemented something like the lean manufacturing system unless they simply don't understand it.
It seems more likely to me their "goal" (pun intended for those readers of "The Goal") is the TPS but their particular supply chain may not currently allow for it.
Fundamentally my question on Tesla is now: if you're priced as a "we have a fundamentally better way of doing things that will disrupt the industry" company, but the core process in your business is still fundamentally inferior, will you have a long enough runway?
I used to be a bear on Tesla for grumpy "I don't like the decisions and design choices they make" reasons, but now the business itself seems a bit underdeveloped.
My theory is Musk knows the only way he'll be able to compete is attracting the best talent through stock incentives, and to do that he needs to jack up the stock price by pushing engineers way past what they are comfortable with.
An underreported stat is the increase in TSLA outstanding shares, which I think is Elon throwing stock incentives at his employees to make them go faster, as a TPS alternative (for now?)
Tesla’s stock behavior is due to the ongoing game of “who blinks first” between TSLA shorts and TSLA short-squeezers. With $10 billion short position the company’s fundamentals are reduced to a side show.
It's primarily from equity grants. Even your article says Musk tacked on his stock option exercise as part of the secondary offering, which increased the dilution.
American carmakers didn't get it for decades. Even when they thought they did, they performed Lean rituals without real understanding. This was despite Toyota trying vigorously to teach them.
(A similar story is told of the spread of the use of Feynman diagrams in physics - I’ve read that their spread was far slower than you’d expect given their utility: For whatever reason it wasn’t enough to be told that they were an incredibly useful tool by people you trusted - it was only when non users actually worked together on problems with Feynman diagram users, or were taught by Feynman diagram users directly that their use actually spread. Both of these transfer methods required physical proximity which restricted the use of Feynman diagrams to a small number of labs for a surprisingly long time.)
There's a This American Life episode (centered around what is now Tesla's plant) that covers this well: https://www.thisamericanlife.org/radio-archives/episode/561/...
“Spreading the Tools of Theory: Feynman Diagrams in the USA, Japan, and the Soviet Union” by David Kaiser, Kenji Ito and Karl Hall
http://web.mit.edu/dikaiser/www/Kaiser.SpreadingTools.pdf
must have been the original source research.
edit: Also, there’s a book (by one of the authors of that paper): “Drawing Theories Apart: The Dispersion Of Feynman Diagrams In Postwar Physics” https://www.amazon.co.uk/dp/B002Y5W2X2
<blockquote>
... Thus, when [Freeman] Dyson arrived at the Institute for Advanced Study in September 1948, Feynman diagrams in hand, he immediately joined ten fellow ‘postdocs’ in theoretical physics, the largest cohort yet.
[J. Robert Oppenheimer had reorganized the Institute to emphasize post-doc training for physicists; he was quoted as saying that "the best way to send information is to wrap it up in a person."]
In fact, the new building that had been planned to hold offices for the newly expanded ranks had not been completed on time, so all of the postdocs wound up sharing desks in one large room for the first 6 weeks of their stay – an architectural arrangement perfectly suited for the sharing of tacit knowledge.
They huddled around large tables swapping ideas, pressing Dyson to explain the details of how to use Feynman diagrams. He delivered several sets of long lectures to the group, supplemented by constant, informal conversations.
By the winter of 1948–49, Dyson had organized them into several groups, each working on distinct diagrammatic calculations. One pair, Kenneth Watson and Joseph Lepore, worked with Dyson to calculate high-order terms in a certain model of nuclear forces, while another duo, Norman Kroll and Robert Karplus, took up Dyson’s lead to undertake similarly complex diagrammatic calculations in QED.
So effective were Dyson’s tutorials that Wolfgang Pauli wrote to another of the young Institute theorists that May, asking what Dyson and the rest of the ‘Feynman-school’ were working on.
</blockquote>
(Extra paragraphing added.)
This reminds me of a certain other industry cargo culting a certain other methodology, but I cannot put my finger on it ...
Okay so our products are cheap and simple. But our yearly totals of 50,000 to 100,000 units is comparable to the what Tesla is hoping to get out the door. Tesla's numbers for previous models weren't nearly that high. Gonna be hard to just in time that.
Personal feeling. What an automaker is doing by outsourcing parts to suppliers and then making them manage inventory is. a) they as someone mentioned, keep the inventory management away from the production line. And importantly allowing suppliers to apply higher levels of scale then you could other wise. Important because likely many of Tesla's suppliers production machinery has vastly more capacity then Tesla needs. Telsa and other automakers are thus 'renting' time on those machines.
Frankly I think Tesla is ramping to this scale for the first time, it's bound to be a mess.
Tesla wants to ramp up production as their primary goal right now; if their supply chain is preventing them from doing so then they've funamentally failed.
> I really find it hard to believe they don't understand the TPS.
I agree, but that makes this more confusing and (at least potentially) more damning for Tesla. It's not that hard a concept; if you're running a manufacturing company but you screw up your supply chain badly then...I dunno. Maybe you're bad at running a manufacturing company? Supply chains are really important when you're doing something like making cars. And people having been advocating TPS since before I was even born.
So...it's a conundrum.
Even a couple of years back when Toyota passed GM as highest volume maker of vehicles, the execs at Toyota said there was a moment where they took their eye off clean processes and Toyota quality suffered a bit. Some of that I suspect had to do with the timing of the retirement of a generation of very experienced people in Toyota. I think they've since been on a corrective course.
I think it makes sense to consider the state of the art for reliable production lines. But NIH syndrome will have its way.
Elon's software background, plus SpaceX which is fundamentally a very different kind of manufacturing, probably works against him here. Software startups essentially hate process (for often-good reasons) and view process as mostly a thing you put into place out of necessity after you've successfully navigated the growth phase. The stories from their production floor are the single most worrying thing that's come out about Tesla - bad vendor quality control, overengineering a la the Model X, all those other things are comparatively simple to grasp, learn from, and overcome. Why wasn't "learn how to mass-manufacture reliably" a much earlier goal?
Just because a system took 27 years to develop doesn't mean it takes 27 years to learn.
Seems like blind replication isn't the answer alone ( http://www.tandfonline.com/doi/abs/10.1080/00207540701223493... ), which is unsurprising, but Tesla's sales targets and their current production line process as-reported seem at odds.
This stuff is very well-studied and making basic errors seems like a problem of hubris coming from a software industry that's in a much less mature state.
I'd imagine that is hugely affected by the lead time on the parts in question. If you're in a Japanese industrial town and every part's manufactured within 20km of your factory, it'd be a perfect approach. If you're in Palo Alto, California and some of your parts are manufactured in a Japanese industrial town, then "just enough" stockpiled parts may well be a three month supply.
This isn't just a minor set of sins. Tesla apparently fundamentally does not grasp the concept of Lean Manufacturing+ as developed by Toyota, also known as the Toyota Production System++.
+https://en.wikipedia.org/wiki/Lean_manufacturing
++https://en.wikipedia.org/wiki/Toyota_Production_System