Separate names with a comma.
Discussion in 'Model S' started by AnOutsider, Jan 4, 2012.
It will. One way or the other.
Hence my concern...
Why, oh why?! It's a pleasant, sunny 65 degrees out here today! No way your Punxsutawney Phil can assure you of anything like that till May?! :smile:
I don't mean to dampen the mood here but, letting Tesla take their time might actually be a good thing; the sigs are bound to have minor niggles (that can most definitely be fixed easily afterwards) even without the added pressure of having to deliver by a date set in stone - you sig res holders are the guinea pigs for the benefit of the rest of us gen prod folks who'll get 'perfect' cars! :wink:
No product (and I'm speaking for the software industry to be specific) tends to come out in a perfect form when first officially released; there are bound to be compromises made (or minor blemishes overlooked consciously or unwittingly) to get to a reasonable delivery date. A piece of heavy machinery such as a car - albeit with far fewer moving parts than otherwise - from a first-time assembly line car maker coming off a brand new prod line is that much harder to "nail" from the get go...
Are we not seeing the same with the Roadster? Up until today they are fixing things on the roadster.
But take a step back.. Isn't it total madness that we are SO full of a car yet to come out? We are putting so much effort in this car and helping tesla (I assume they read this forum) with giving them feedback.
If it was BMW or Audi who was bringing out an EV which did not meet my expectations i'd simply ignore them, but what are we all doing?
So if this forum turns into a frenzy around Juli that should be a real big honor for Tesla!
How about celebration in August because it does meet expectations?
Yes, on this forum the assumption was that the signature would not only be a special line to skip the first-come-first-served principle, but also coupled with a fixed yet unknown package deal, top-of-the-line model. That always implied a risk. Now you don't even have to pay for performance if you don't want to. As far as I can tell, tires are the only sig no-cost option where you need to pay for something less expensive. Maybe Tesla gets a better price for the turbine wheels if the volume is larger, or they want the first 1000 to have a common showcase "signature" look, which they consider the turbine wheels to be a part of. In addition to having plans that require money, such as surviving by continuously introducing new products.
Back to the main topic:
Nope. Although numbers of 6000 and even up to 7000 are sometimes mentioned as possibilities, the goal is 5000. Five thousand.
Don't pre-program the next disappointment.
Wow, that huge mass of really cool, useful information DiscoDucky takes the time to post, and that is what you choose to take from it? Not only do you single out just that one line, but you see it as some sort of deliberate attempt to set up things for disappointment.
I understand being a Tesla fan, but he's just sharing his manufacturing experience. Was there something wrong with just saying something like "Neat, but a quick correction. The target goal is 5k, not 6k." ?
You are assuming that the 5k correction is the only thing I would have to say if I quoted the whole thing. While it's an interesting read, and while I appreciate someone giving background info about computer manufacturing, I don't think you can apply that 1:1 without having any specific knowledge about what Tesla is doing, and he doesn't seem well informed about that at all. In November Tesla still implied that their mid-2012 target is a plan that doesn't require them to rush things in terms of quality, and since they employ a highly qualified team with lots of industry experience, I don't think any of DiscoDucky's information would be surprising news to them.
If you are asking for it, I can make three more quotes:
1000 cars in July/Aug? For testing or for delivery? Neither one, unless I'm missing something big. Unless things are ten times better than to be expected based on public info. I don't even understand how this is supposed to work. It's not like Tesla could throw away 500 cars from QG4 if they don't turn out too well. Or could they? If you add QG1 up to QG5, it implies Tesla would build 1626 cars until end of August, doesn't it? If that somehow makes sense, then it requires a lot more explanation. Based on comments we heard so far, we'd expect Tesla to start with at most few hundred in the first month of delivery (more or less equals production according to a Tesla rep), which AFAIK nobody expects to be before July.
Where does "(8 to 12 weeks)" come from? The smallest combination of numbers is 25 weeks. But perhaps this relates to text which was deleted before posting.
is later followed by
With a list of 6 items, none of which comes with any information indicating it would be particularly difficult to believe they could achieve it, if they put their minds and hearts to it. Which they do.
There seem to be others who think Tesla's schedule is very aggressive for what is common in the industry, and I don't have any information to counter that. But I still don't, in neither direction. Maybe they will make 5000 in 2012, maybe not. Who knows.
I found DD's posting interesting not for the volumes, but the description of the process.
Re numbers: DD's from the computer industry, and the raw material cost of a computer, less the salvage cost of parts you can pull from a QC-rejected computer, is pretty small; also, the objects themselves are small. I took with a healthy dose of skepticism that Tesla could ramp up as quickly as DD suggested.
Re process: This description, especially of early-stage development, struck me as highly plausible and would explain why "getting the line up" isn't enough to instantly go from 0 to 400 cars per week. Helpful insights into the world of manufacturing for those of us far removed from making things on an assembly line.
That was what I got from it, that the process was interesting.
As for # of cars built, Discoducky isn't talking about shipping during the rampup, just run through the line. Nothing that comes out of a line that fails a QC gate can be shipped and you go back to QC gate #1 because the line still has systemic issues. They'd be disassembled for parts I'd guess though I suppose they always have the ability to do a complete reinspection and could decide some were fit for shipping.
In order to ship 5,000 cars, Tesla will have to build significantly more than that. I have no idea how many extra builds would be required, but I figure Discoducky has a much better insight than most of us. Tesla seems convinced that they'll make it, so maybe they're ahead of the curve Discoducky noted.
With a car, you also have to guarantee a much higher quality for each single car that gets delivered. You can't just build a bunch and make statistical tests to see where that goes.
Regarding the process, I guess I would expect them more to, for example, progress from 'low speed/low failure rate' towards 'high speed/low failure rate'. Instead of starting with max speed and then reducing the statistical failure rate, which would be from 'max speed/high failure rate' towards 'max speed/low failure rate', in so far as I understand the description.
Perhaps they would use that QG process multiple times, each time they increase the line speed, however with fewer cycles, and more testing, on average... or apply it to sub-assembly-steps as those are improved, or something like that.
Based on what experience do you make that guess? That would have been my naive guess as well, but having now heard from someone with roughly 1000x more manufacturing experience than me, I believe my gut instinct was probably wrong.
Actually, I'd guess they run at low speeds during the debugging process of actually getting the robots to work at all, but that's sort of in the "development" stage. In order to pass the quality gates, you run at full speed.
Which makes perfect economic sense really from a risk perspective. You absolute, positively don't want to find a timing bug at full speed when inspecting a batch of 500 cars... You want to find it much, much earlier in the process when you're running a batch of 1/100th of that quantity.
I would expect that if they make a batch of cars and there is a failure in quality control on that batch that they would fix what went wrong and not throw all those cars in the used parts bin and start over.
What constitutes a failure in quality control? If the cars come off the line and the right front quarterpanel is misshapen then just re-do them once the quarterpanel machine is working properly.
If the battery pack (the heart of the car) does not pass quality control then just replace with a battery pack that is working properly.
My guess is that every Signature edition will be a car that did not come pristinely off the line but needed some re-work to make it perfect.
Good question and no idea. I'd guess not all problems are equal.
I also have no idea what the regulations are. If one of the air bag components on the line has some sporadic failures and isn't passing the quality control checks, do you really want one of the cars from that run? Can they even legally ship them? I'd want a car that passed from a clean run. On the other hand, if it's a slight gap in how the left fender lines up that's a matter of turning a screw three times, maybe that's perfectly fine (and legal) to hand fix that batch.
I wish I knew someone knowledgeable in the area. After Discoducky's post, I'm rather fascinated by how all those sorts of things are handled.
Agreed. I think making a car is vastly different than motherboards or computers. There, you lose $1,000 or so in a batch. Not so with a car. I'd lean more towards Norbert's thinking as well, which is similar to DD's but seems more applicable to a more expensive item. Perhaps they do that when machining the smaller bits, but not to whole cars.
I have no illusions that regulations = common sense (they rarely do) but I would think that common sense would dictate that in that scenario Tesla could/would/should replace the airbag system in every one of those cars.
The best example of what happens when "low speed/low failure rate" doesn't match up with "high speed" http://www.youtube.com/watch?v=HnbNcQlzV-4 :tongue:
Yes, due to the cost of failures, they will need to fix most problems in development runs with *very* low volume (maybe just : build one car, test one car). And even before that, build and test sub-assemblies only.
However, we heard that there will be a "ramp-up" of production while deliveries have already started, over a period of several months. Since they can't afford high-failure rates of completed cars, the "ramp-up" can't be from high failure rates to low failure rates. So there must be something else changing over those months, something which hasn't been described yet. My guess would be that at each manufacturing step, there will be more manual intervention, testing, adjusting, non-automated actions, perhaps also slower robot movements, in general, until they know which movements can be done faster, or how they can be done faster. All that will increase the time taken for each car. Then, over time, they can remove, optimize, and automate more and more of those steps. While that's just a guess, since you say "my gut instinct was probably wrong", we seem to agree that at least the descriptions so far seem to be missing something.
Oh my god, so funny! Fond memories of watching the show with my grandmother.
Beats my attempts of rationalization... :biggrin: