In the Toyota / lean-theory rework is the start of an near-exponential process where different tasks are waiting for the other to be completed ad infinitum. If you've ever worked at an old-school firm you'll recognize this: the quality of your input, a colleagues output (say, a dataset) is sub-par, so you start checking it regularly. If it falls short to your standards, you ask for a new set and wait. He does his work again, and you restart your task and start checking his work even before you start your real work. That's a lot of time ill spent.

Multiply that over departments and many processes and the result is great inefficiency. So FPY is the KPI to stop all that. FPY is not really valuable by itself, but hopefully by the culture it supports. Swapping FPY for # delivered is false management and points to cultural issues. In this case: managing by the wrong target (# delivered instead of the "total worth of the value chain").

Nobody would care if Tesla would just deliver growing amounts of cars at falling costs. It's the bootstrapping (in the Baron of M-sense) towards false metrics that gets the poison going and management is totally working within that frame of mind. They should find their own frame, and operational excellence just might be one. (Apologies for the MBA bs-bingo, but in this case I think it's to the point.)

Telsa really should aspire to Toyota levels of efficiency, but you also have to account for the fact that Toyota has had decades of experience as well as a stable and long-term relationship with their parts suppliers. Stopping the line and reconfiguring that plant in a Toyota plant is likely a minor rework for efficiency. But when a model and line and entire system is all new, a stoppage might not be a small rework despite best efforts to have arranged for it to be so ahead of time. Even the timeline and difficulty to setup a new Toyota plant vs a new Tesla plant aren't really comparable, because Toyota also has and excellent and stable set of design practices and supplier relationships. If two to three years from now, the rework ratio hasn't dropped considerably then Tesla is doing something wrong, but right now I think it's understandable that Tesla is just getting on it's legs wrt truly mass production. You always have to learn new things at each order of magnitude of production - even with products and methods that have been successful at the previous production level.

A thought: We're at a point in time where Toyota employees and management (and many other manufacturers' employees and management) have that experience. This experience is not something that has to be learned with every new iteration of the car production line - it's something that can quite literally be bought.

The question that follows on: What internal process has prevented that experience from being bought? Instead Tesla seems to have opted to gain that experience from scratch at great cost to investors (and consumers).

Some expertise can be bought, but some of the support is built over time, intrinsically outside the factory into the overall Toyota system. But ironically, thinking one can just buy an expert is to apply the same short-term mentality to the process that Tesla is accused of misapplying to product.

When Toyota first started creating their own US factories, they took the steps of shipping floor employees in the US to japan a few at a time and working for weeks of training each class. When other auto companies tried to replicate Toyota processes in their own factories, it didn't help yield in the factories in many cases because often the input supplies needed so much rework. And then they had problems changing suppliers configurations because those relationships were contentious, and not of the nature of Toyota to it's suppliers. When you're trying to stand the whole system up at once, it's not just a matter of "don't write bugs" in one part of the final assembly.

When Boeing reworked it's production design process for the 787, they had years of production problems to re-iron out, despite having years of great process experience previously.

The problem here seems to be that they seem to be striving for significantly short term goals, like meeting production targets, rather than really doing the work change the entire system and culture to one that would focus more on quality.

That doesn't really bode well for their future if they consider their current production rates that imminently important to their success.

But who is going to give them money to shut down the production line again and retool everything to increase FPY? And will they do it again when more issues are found?

Isn't swapping FPY for # delivered while they determine the root of the issue and fix it a good plan? It keeps the money flowing in, it helps them find other issues in parallel, and keeps cars getting delivered.

I see a lot of people critisizing Tesla here, but my question is what would you do? You are already in a situation where your plant is producing cars with flaws, no pretending that you don't make mistakes or are infallible and would never be there.

Garbage is flying off the line, what is the solution that you would go with that is better than what Tesla is currently doing?

Agree. FPY might not be the metric fitting the current status of the company best. It's all situation-dependent. FPY of 14% is clearly unsustainable though. That's because quality is not one or zero, but a continuum. So what's the SPY, TPY, etc.

Yeah, i'd love to hear final numbers on how many have issues once in the hands of customers, or what kind of issues they are fixing, or how long they plan on having this "manual" patch in place.

Without that information, it seems kind of silly seeing everyone proclaim how Tesla is done because of this... It's like telling your doctor that you don't want the medicine because the better solution is to just never get sick in the first place... yeah, everyone agrees, but you are sick, so here take this.

> But who is going to give them money to shut down the production line again and retool everything to increase FPY?

I imagine that it's significantly cheaper to fix the problem (and identify a root cause) when the car is still in pieces than to disassemble the finished product to fix the problem.

As for re-tooling, that can be done down the line with a manual fix in place in the short term (after all, they're able to fix it manually after the assembly line is completed).

Much like how in software development it's cheaper and faster to fix a bug caught in development than it is in production.

Yes, but if you find a bug in development and you are days away from shipping, and you have a solution that fixes the problem but doesn't address the root cause, are you going to delay the whole launch so you can fix it before hand? or are you going to release with the patch, and fix the root cause later?

Now software is different than cars, but I'm gonna say that I'd go for the latter just about every time. Both as a developer, and as a user.

> all of that just influences manufacturing rate and cost per vehicle.

It seems like you answered your own question? These are really important for a company that's trying to make money.

Unless it's found that it's cheaper and faster to crank out the cars with a poor FPY, and fix most of them up afterwards.

Especially if the problems are known problems. If a specific step needs a person or 2 to fix a problem on 90% of cars, but afterwords it works 100% of the time, what's the issue? Changing that step to fix that problem might introduce other issues, or cause problems with the supply chain, or could possibly just not work.

Hiring a few extra people to fix the problem that you know about in order to keep moving forward not only gives you the time to really design and think about a "proper" solution, but also keeps the money flowing in.

Its not cheaper, and it won't result in better quality. Any time rework is done outside of the normal process it is a big deal. There are whole organizations meant to make sure things like that happen as little as possible.

It seems significantly cheaper than stopping production to fix the initial problem, test it, ensure it doesn't cause any other problems, ensure it doesn't upset the supply chain, ensure it is reliable, and more.

Yeah, it is cheaper than stopping the world. It is much more expensive than not making the mistake in the first place. So much so that I would guess that Tesla is losing money on each car that they had to rework. And remember, Tesla had so many production issues that they _did_ have to stop the world and move into a tent of all places.

You think Tesla is losing money because of 37 minutes per car that needs rework? Roughly $15 per car of labor?

Not making the mistake in the first place isn't possible at this point... Your options here are "stop everything and fix it at the source, burning piles of cash every second it's down" or "hire some workers to spend 37 minutes per car to fix these known issue while you work on solving them at the root".

And going with option 1 seems like a fast track to bankruptcy.

Do you know how long it takes to build an entire vehicle? If its 60 hours, then 37 minutes probably isn't breaking, but if its 8 hours, then that 37 minutes is 7.5% of a vehicle's construction time. If Tesla is wasting 7.5% of its time fixing defects, its running at at least 7.5% less than peak efficiency, and if margins are less than 7.5%, that does indeed mean that the vehicles are being produced at a loss.

I'm sorry but that math makes no sense at all... It's a matter of throughput vs latency.

Who cares if it takes an extra 37 minutes per car, if they can maintain the same throughput, it won't matter a single bit (except possibly someone having to wait an extra 37 minutes to get their car...)

The extra time spent on manufacturing has as much to do with margins as the delivery truck getting stuck in traffic does. As long as it's not holding anything up, then there isn't an issue.

And it sounds to me like this is a temporary solutions while they fix the root cause, which seems like a really goddamn good idea vs just shutting the whole operation down and giving up.

Exactly, it's a question of capital costs vs burn.

If you need say 100 million in hardware and 1,000$ in manpower to build a car then spending an extra 100$ in manpower is cheap even * (5,000 * 0.14) = 70k per week. Meanwhile you might be paying 70k per hour the line's stopped.

Granted, you don't want to wait a full year to fix a production issue, but you can batch things so you are fixing several parts of the line at the same time. And if one of those batched fixes was not 100% correct but is re-workable then you can wait another few weeks for the next batch of fixes.

Conversely, if you where producing 50,000 cars per week then rework costs go up and shutting things down quickly to make a minor fix is more viable.

Making up numbers isn't helpful here. TFA says the average is 37 minutes/car. I am not sure where you get the $15 from, but usually manufacturing companies have an internal hourly rate for different stations within the plant. For example, the machine shop at the last company I worked for was ~$400/hour. Further, it is likely that there is also significant material cost associated w/ these reworks. Given that the model 3 is a low cost vehicle, it is probably also pretty low margin (I haven't looked). I would guess that this means that they are selling the model 3 at a loss, especially if you throw in the costs for the recent retooling.

I got the $15 from other areas that have said the cost of labor is roughly $30/hr for stuff like this.

But regardless, if you really think a 37 minute fix is costing them enough that it means they are selling the car for a loss, then I can see why people are upset by this news, as it means that shutting down the entire plant would make more financial sense than selling at a loss.

But since we are both working from a guess of a guess of a guess, I'm fairly certain that an extra 37 minutes per car of work isn't eating over 100% of their profits, and that their choice of not stopping production over small fixable issues is a smart choice. And even if it were, i'd argue that keeping your workforce from getting other jobs because they can't get work at your company while you endlessly tinker with the production lines to try and achieve what only a few companies in existence have done before (almost 100% FPY) isn't good business sense either...

But then you want to throw in the cost of tooling changes, and now this conversation isn't against the cost of a car, but the entire cashflow of the business, which if you are curious you can lookup. They aren't turning a profit, and nobody thinks they are yet. But if people are holding them to the ability to get everything right on the very first try and the second there is an issue to shut down again, then they are going to be disappointed regardless.

The industry as a whole tries to avoid rework as much as possible, but you are saying that a 20% FPY is fine, and it only costs $30/ vehicle anyway. I am not saying that I know much about the specifics of Tesla's finances, but I can say that you are missing the big picture here.

Having poor FPY means difficulty scaling, which is indeed one of Tesla's problems--their production scales not-quite-linearly with the number of bodies they throw at the problem.

It depends on the difficulty of the "fix". half an hour per car doesn't seem that bad, and while they will clearly try to fix that (and I'm not trying to imply they shouldn't), I don't think the fact that they are continuing to pump out cars with a "flawed" system (while presumably working on solving the problem internally) is any indication that they are unable to scale forever.