With Model 3, Tesla Is Taking An Unusual Production Risk

Tesla Model 3


Is Tesla taking a production risk with the Model 3?

Recently, Reuters set out to explore and answer this exact question and the conclusion was that Tesla is pushing the boundaries with the Model 3 and it’s a risk that will either make or break Tesla.

Tesla Model 3

Tesla Model 3

Reuter states:

“Most automakers test a new model’s production line by building vehicles with relatively cheap, prototype tools designed to be scrapped once they deliver doors that fit, body panels with the right shape and dashboards that don’t have gaps or seams.”

“Tesla, however, is skipping that preliminary step and ordering permanent, more expensive equipment as it races to launch its Model 3 sedan by a self-imposed volume production deadline of September, Musk told investors last month.”

Basically, Tesla was a bit behind schedule with the Model 3, so in order to catch up, the automaker is skipping a step, but we’re not so sure it’s a necessary step anyways.

Tesla seems confident it can bypass the ordering of prototype tools, stamps equipment, etc and move right into building what it refers to as release candidates. If the automaker successfully pulls this off, then it may well change how future automobiles are developed.

Reuters adds:

“Musk’s decision underscores his high-risk tolerance and willingness to forego long-held industry norms that has helped Tesla upend the traditional auto industry. While Tesla is not the first automaker to try to accelerate production on the factory floor, no other rival is putting this much faith in the production strategy succeeding.”

Tesla is always looking for ways to revolutionize the automobile and this time the revolution is happening behind the scenes.

If it works, Tesla will be applauded. If it’s a total failure, well…

Source: Reuters

Categories: Tesla

Tags: ,

Leave a Reply

47 Comments on "With Model 3, Tesla Is Taking An Unusual Production Risk"

newest oldest most voted
The article says soft tooling hurt the Model X because Tesla had no time to incorporated what was learned from soft tools into the order for hard tools. This is not a fault of soft tooling and it isn’t soft tooling hurting Tesla. That is simply not putting enough time in your schedule to do things properly. This process is actually used in airplane production. It does allow them to start production faster, but it also results in a lot of travelled work. That is work that must be redone because the copy in question was produced using a process/tool which was later found to be wrong. So the copy is held at the factory and reworked. On planes where production is relatively slow that hurts less. But if you bang out 30,000 cars and then have to fix them all before sending them out it can greatly increase costs. And it also gives incentive to ramp up slowly so that any mistakes you make result in fewer copies to repair. Also, I was at the Tesla factory recently and I was surprised how much rework they had to do there. It’s not that it was amazingly high compared to… Read more »

Agreed. You can skip prototype tooling, but you have to deal with the ramifications if you find a problem down the line. The reason automakers don’t skip the step is they’ve been burned before. The cost/time of doing it is a lot cheaper than not doing it.

Hey unlucky,

Did Tesla have any “soft” tools for the model 3?? They must have used something to make the first few prototype model 3’s.

They just made those by hand.

(⌐■_■) Trollnonymous

Hands washed with Palmolive.
Those are really soft tools…..lol

I don’t know if they had soft tools. I can only say it isn’t completely necessary that they do. You can hand-beat metal panels without any tooling at all. If you make a wooden or other similar form you can do it without any soft tooling, although I would suggest that even a basic form is tooling, just not soft tooling (because soft tooling means pre-hardened steel dies in this case).

Beyond that I couldn’t say, because I don’t have any more info.

Well, I have one more piece of info. That is Tesla showed stamped aluminum Model 3 parts when I was there. They weren’t body panels, more dashboard panels, etc. There were no steel panels of any sort in evidence, but they said the Model 3 line which would do it was almost put together, just hidden from visitor view.

You couldn’t safety test an aluminum Model 3, so there must be a bunch of steel 3s floating around.

If the story of there being about 100 Model 3s at the test facility back east is true, not to mention all the pre-production candidates running around the Bay Area, they must be stamping and forming lots of near production ready cars. Forming that much steal by hand would be nearly impossible.

Oh, there are steel 3s I’m sure. But they didn’t show any of the parts used to make them. I took this as meaning the plant was not yet capable of making them on even a semi-production line.

My tour of the plant was before the 3 sightings were as common as they are now. There could have been fewer then. As I said, the line which was going to make the steel stampings was supposedly nearly put together, surely it has been able to make parts for some time now. With hard tooling, not soft.

All die tooling is disposable at some point. Dies can only go a certain number of cycles (stamping) before they need to be reworked, and eventually get thrown away. There was a push to make temporary dies out of, well, aluminum but of course stamping aluminum with aluminum does not work (like metals are sticky). In any case they use an advanced alloy like fortal (aluminum with most of the strength of steel) for temp dies.

Aluminum actually is easier to press than steel sheet, but has a higher springback ratio, which means you have to make the die shape for what the part looks like after springback is complete. Its an ugly problem, but computer modeling has helped a lot, and we assume Tesla has mastered the process.

Finally, it is not for sure that much or even most of the M3 is aluminum.

“Aluminum actually is easier to press than steel sheet, but has a higher springback ratio, which means you have to make the die shape for what the part looks like after springback is complete.”

I think you mean before springback is complete. Figuring out what shape the die needs to be after springback is complete — that is, the intended final shape — would be pretty straightforward. Trying to figure out what shape it should be while still under pressure from the stamping process, before the piece is released, would be much harder.

Aluminum also can’t be bent as sharply as steel, nor stretched as far, and in some cases (perhaps not at Tesla) the stamping has to be done multiple times on the same piece to achieve the final shape.


The parts I saw were all aluminum. They were mostly Model S/X parts with a few Model 3 parts mixed in. The Model 3 parts might even have been mockups, I dunno. The S/X parts were parts stacked up for use on cars. Some for rework, some ready to go.

I posted a link to a Forbes article at the bottom of the comments that claims automakers haven’t used “soft tooling” for decades. Can anyone confirm whether this is true?

Do these issues [cost aside] arise with carbon fibre panels? What about glass fibre or even plastics, as used on scooters & motorbikes? Such panels mounted on energy-absorbing spaceframes a possibility?
Certification by crash-testing would be a problem but surely could be overcome.
Steel panels just seems SO ‘Ford Model-T’ in 2017!

I say the article is BULL.
With computer modeling, you don’t need this step anymore.
If you’re computer modeling the factory.

Soft tooling isn’t as great as some think. It:

1. Costs extra money.

2. Takes more time, particularly if you wait until releasing hard tooling.

3. Can mean you end up working on two different products – the prototype and the production version.

Per #3: We once had a product (plastic encased body-worn electronics) that we soft-tooled “to work out the bugs first”. Instead, we encountered an entirely different set of problems once we went to hard tooling, because the hard tools performed differently. So we spent a lot of time trying to determine whether the issues were with the design, the tooling, the process, or everything. Ironically, in this case, all soft tooling did for us was cost more time & money, and produce more engineering changes.

Yeah, traditional methodologies develop because they meet a need. Once that need can be realized by cutting out part of that methodology, since things are changing all the time, then they have an advantage in the time it takes to come to market, and cost.

Also, as you point out, soft-tooling is not a panacea.

The idea of soft tooling is a hold over from manufacturing of the past.

It will be just like Model X, a lot of problems with initial production, but they will work them out gradually. I wouldn’t buy the first year production unless it is absolutely necessary to get the tax credits.

Right! This is why I’m happy to be ~< 200,000 on the list. 😉

(although I'm sure the California location moves me up)

I guess if you don’t do soft tooling, then you want to ramp up slowly in case things need to be reworked.

Also if you blow it, it takes longer to fix your tooling. Soft tooling is easier to modify and replacing your tooling with new soft tooling is also quicker.

To change a hard tool can take weeks. Tool steel is very hard, grinding it down takes a long time.

Musk has high confidence they can get their dies right the first time. You don’t need soft tooling if you can get it right the first time. But if you think you can and then find out you can’t it can cost you a lot of time in making new tooling you didn’t expect to make.

maybe Tesla altered the hard tooling machine to make it easier to change.

They are constantly talking about the machine that makes the machine.

It’s not a problem with the machine. Dies are mostly made by hand anyway. The problem is tool steel is very hard. It takes a long time just to grind it.

Computerized modeling greatly reduces the risks of skipping a step that has been done for a 100 years…
Also not having an ICE and transmission also greatly reduces the risks vs a legacy car design…
Plus the interior is simplified enough there wont be any vent or gauge problems I guess??

It’s just playing the odds. Fewer parts, and simpler parts, means fewer things to go wrong. But it certainly does not guarantee that nothing will go wrong. Even a relatively simple car is still a very complex machine, with thousands of parts. And every place one part interfaces with another is a potential failure point.

It’s almost mathematically certain that things will go wrong when Model 3 production first starts. Realistically, Tesla can reduce the number of production problems by good planning and lots of computer modeling, but not eliminate them entirely.

The Model 3 will be built by robots.

If some of the robots don’t work properly Tesla will simply tweak the software (this was demonstrated in a recent video).

With reprogramming, those same robots can also be used to build other models.

Where’s the risk?

Most car assembly is automated except for final assembly. Go to any US auto plant and there are few people standing around.

Robots weld. The conversation is about stamping.

Yup. Doesn’t do any good to re-program the robot if the part is the wrong shape to fit properly or the wrong shape to work properly.

pshaw. If the part doesn’t fit, bring in a Bender robot and give it some beer. It’ll make it fit.

(⌐■_■) Trollnonymous

The risk is sole sourcing vendors.

The whole Model 3 enterprise is a giant risk, from the size of the battery factory to the investment in the production facility for 400,000/year. I sure hope they pull it off.

The real question is whether they can make money by selling these cars for $35-$40k. I seriously doubt Tesla can make these cars profitably in the high cost area where they factories are located. Let’s say a Model S has 25% gross margin. Assuming average selling price is $80k, that means on average the car costs $60k to build. Now in order to make Model 3 profitable they need to cut cost in half, to $30k. Tesla has not been good with their price projections before. Both their cars have been more expensive than the initial announced price.

I think you mean all three of their cars. Tesla raised the price 10% on the Roadster shortly after it first went on sale.

Certainly there is a very real risk that Tesla has underestimated the cost of producing a unit of the Model 3. With Tesla’s previous cars, initial underpricing wasn’t that important, because the “premium” market segment is relatively immune to raising the price by a few thousand dollars. As I recall, when Tesla raised the prisce of the Roadster, it lost almost no reservations at all.

Things are very different with the Model 3, not only because Tesla is aiming at a significantly lower market segment, but also because Tesla has repeated the $35 figure so often that the market is going to see that as a promise rather than an estimate. If Tesla raises the base price for the Model 3, it’s going to lose a lot of customers.

I hope that Tesla has planned ahead for this contingency, by making some things it plans to put into the Model 3 optional; things which can be left out at the last minute (or made optional rather than standard equipment) if the unit cost proves to be higher than estimated.

I think it’s clear that Tesla has underestimated the cost to build the Model 3. Why else would Musk already be saying that the Model Y will not use the Model 3 platform, and that he needs a new plant to do it? OK, I can think of one reason…another capital raise, but I suspect that building the Model 3 is costing them a lot more money than anticipated and cost savings are not possible without redesign and a new, more modern production facility.

Yeah, it will be a long time, if ever, that Tesla produces a base Model 3 that sells for a profit at $35,000 all things considered. Maybe 4 years from now when mass production has spread out the cost of development and capital investment, and when batteries have dropped in price and Tesla has been ironing out costs for 4 years.

The good news is that most people will order upgraded Model 3s: the bigger battery version, and all-wheel drive will both be popular, and even people ordering the base model will often opt for at least a couple thousand dollar option package to get a little more luxury in the car.

A Model 3 P75D with Ludicrous mode will probably sell for $70,000+ if they decide to build and sell it.

I think average sale price of a BMW 3 series is $43k. This includes very generous lease offers, that bring down the real price to about $37k. So I think as long as $7.5-$10k subsidies exist we can see >$45k transaction prices in large volume, once it is gone either they have to cut price or production.

Heads would roll, at a major OEM if something were wrong after skipping steps. At Tesla, as in the past, they’ll lean more on improving, on the fly, like they do everything else. They’ll also slow, or nearly stop production, if problems are bad. Model X real production started quite late.

Lastly, I bet Tesla will respond with a mid-year update if anything about the car proves wildly unpopular.

I would agree their cars see greater improvement over time. A sword with two edges. It’s not like we won’t all know, by ~October.

Apparently, the Reuters article is nonsense and completely wrong. Automakers don’t currently use soft tooling and haven’t used soft tooling so for decades. From the Forbes article linked below: “The Reuters article suggests that Tesla plans to go straight from digital model to final production dies. This elicits yawns at global OEMs. ‘For a long time, we have gone from digital straight to mass-production tooling,’ said my European contact. ‘If you can digitally test your design with a level of confidence, there is no reason not to go straight to mass-production tooling. We are not the only ones who do this.'” “Even makers of rather low cost Chinese cars appear to be far ahead of what is celebrated as ‘pushing the envelope’ at Tesla. Last week in Shanghai, I could witness in Geely’s design studio how digitally designed sheet metal parts were transferred into testing samples with a high-speed CNC machine, and fitted to a confirmation setup. Once approved, the digital data go to the toolmaker for the final production dies without creating fanfares on international wire-services.” . . . “Tesla’s alleged ‘pushing of boundaries’ may impress gullible reporters, or sell-side analysts. For large OEMs, it comes down to ‘nothing… Read more »

Risky but only to a point. this from another Reuters article:

The auto industry’s incumbents have not been standing still. Volkswagen AG’s Audi division launched production of a new plant in Mexico using computer simulations of production tools – and indeed the entire assembly line and factory – that Audi said it believed to be an industry first. That process allowed the plant to launch production 30 percent faster than usual, Audi said. An Audi executive involved in the Mexican plant launch, Peter Hochholdinger, is now Tesla’s vice president of production.”

So Tesla hired one of the people who has successfully applied the simulation technology.



If Tesla wanted to do everything just like the ICE car makers, they would have built a nice safe 75 mile range slowish compact car with 6.6 kW “high speed” charging. And EV’s would have languished with a horrible reputation of being electric golf carts.

Be different, Tesla, be different.

You mean except for the ICE car makers making 238 miles not slowish compact cars with 50kW (apparently) high speed charging I guess?

Or for that matter the ICE car makers making 135 mile semi-slowish compact cars with 100kW high speed charging I guess?

If Tesla wants to own the (nearly) affordable EV market they have to follow the ICE companies which preceded them into this market long ago and continue to iterate.

It is becoming more and more difficult to separate my desire for Tesla to succeed with my dislike of Elon Musk.

Now, solely because Elon has made these ridiculous promises of whimsical production goals, Tesla is skipping a step intended for quality assurance to meet unrealistic production dates.

Apparently, the lesson Elon took from the launch of the Model X is that Tesla wasted too much time in quality control.

How many ASSUMPTIONS have you made here??

As noted, no different than Audi which successfully did the same think in Mexico plant with new model.

And the first 25,000 are to employees as test bed before they get to the public so Tesla will have six months of 2017 to find and correct flaws before the general public starts getting cars in early 2018.

“Basically, Tesla was a bit behind schedule with the Model 3, so in order to catch up, the automaker is skipping a step, but we’re not so sure it’s a necessary step anyways.”

You may not be sure, but every other OEM is (based on a centuries of design/manufacturing experience). Tesla is more than a little behind; the steps their skipping, if performed properly, probably add up to 8-12 months of validation and testing that is being skipped.