Tesla's Secret Second Floor

[mspohr]

Why Tesla makes more parts in house... And why other companies should do the same

Inside Tesla's Secret Second Floor | Backchannel

The answer is simple: Our goal wasn’t to build the best electric vehicle. It was to build the best premium car in the world that just happened to be an EV. This meant integrating technologies that were not readily available. It also meant pushing the boundaries of what was considered “normal” for the design and manufacturing of a car. Furthermore, we needed to do this on an accelerated timeline that most automotive suppliers could not fathom. So, in many cases, this meant building components ourselves. Building your own core components has obvious benefits, but there are some other advantages that you might not immediately recognize.

"Up to 50 changes a week"

 

[McRat]

Why Tesla makes more parts in house... And why other companies should do the same

Inside Tesla's Secret Second Floor | Backchannel

The answer is simple: Our goal wasn’t to build the best electric vehicle. It was to build the best premium car in the world that just happened to be an EV. This meant integrating technologies that were not readily available. It also meant pushing the boundaries of what was considered “normal” for the design and manufacturing of a car. Furthermore, we needed to do this on an accelerated timeline that most automotive suppliers could not fathom. So, in many cases, this meant building components ourselves. Building your own core components has obvious benefits, but there are some other advantages that you might not immediately recognize.

"Up to 50 changes a week"

You mean like the 1800's, right? Screw JIT and risk based supply chains when a master craftsman can make each part by hand with a file in the workshop.

Up to 50 changes a week is funny.

Sort of like saying I can change all 4 tires on my car 10 times a day. OK...

 

[EinSV]

Why Tesla makes more parts in house... And why other companies should do the same

Inside Tesla's Secret Second Floor | Backchannel

The answer is simple: Our goal wasn’t to build the best electric vehicle. It was to build the best premium car in the world that just happened to be an EV. This meant integrating technologies that were not readily available. It also meant pushing the boundaries of what was considered “normal” for the design and manufacturing of a car. Furthermore, we needed to do this on an accelerated timeline that most automotive suppliers could not fathom. So, in many cases, this meant building components ourselves. Building your own core components has obvious benefits, but there are some other advantages that you might not immediately recognize.

"Up to 50 changes a week"

As this piece summarizes, Tesla’s philosophy of constant innovation has been a key part of their success. I wouldn’t be surprised to see more automakers begin abandoning the traditional model year to try to keep up.

 

[SO16]

Up to 50 changes a week is funny.

Sort of like saying I can change all 4 tires on my car 10 times a day. OK...

Um no.

 

[scaesare]

You mean like the 1800's, right? Screw JIT and risk based supply chains when a master craftsman can make each part by hand with a file in the workshop.

Up to 50 changes a week is funny.

Sort of like saying I can change all 4 tires on my car 10 times a day. OK...

This feels like you didn't even read the article...

 

[McRat]

This feels like you didn't even read the article...

I read it. Vertical integration was dropped on complex manufacturing long ago.
What changed? CAD/CAM interconnection software. QMS implementation. Parallel development.

If you have to 'fly over to correct supplier problems' you can save your plane ticket by just firing your procurement team. They cannot successfully validate or build a robust supply chain. ie - They are behind the curve, outdated, inefficient.

EDIT - In the past, the goal of the 'buyer' was to get the lowest price on a batch of parts. Today, it's to improve long term profitability and growth. They are two different concepts.

 

[scaesare]

The discussion of "layered" capability in that article addresses that. You act like they are suggesting that they are ignoring outside vendor integration entirely.

And while procurement teams managing supply chains does work, it's not necessarily the most agile and effective method if fast iteration and new feature adoption are paramount.

There's likely some very valid reasons why it takes Detroit several years to introduce features and changes to a new vehicle model whereas Tesla does it in months. I suspect having the design, engineering, manufacturing, and production teams in the same foxhole is one of them.

 

[PhatCat]

A secret second floor? Keep your eyes out for "Being Elon Musk" at a theater near you.

 

[jaguar36]

There's likely some very valid reasons why it takes Detroit several years to introduce features and changes to a new vehicle model whereas Tesla does it in months. I suspect having the design, engineering, manufacturing, and production teams in the same foxhole is one of them.

Sure, Tesla just skips the whole validation and testing part. or rather, they use their customers to do that.

 

[Rashomon]

I

I read it. Vertical integration was dropped on complex manufacturing long ago.
What changed? CAD/CAM interconnection software. QMS implementation. Parallel development.

If you have to 'fly over to correct supplier problems' you can save your plane ticket by just firing your procurement team. They cannot successfully validate or build a robust supply chain. ie - They are behind the curve, outdated, inefficient.

EDIT - In the past, the goal of the 'buyer' was to get the lowest price on a batch of parts. Today, it's to improve long term profitability and growth. They are two different concepts.

There's a lot of discussion in the business theory world of when it's appropriate to be integrated or when to outsource. See Clay Christensen and Horace Dediu in particular for this; Horace has looked at it relative to Apple and phones, and noted the advantages that Apple has had in being particularly integrated. If you're doing something completely new, it's usually better to be integrated. If you want a better, higher-current, faster-acting contacter than anyone offers, and your need is unique, you may have trouble finding a supplier for such a thing -- particularly if you're only making 50k vehicles a year. It's probably better to do it in-house. Same thing with controllers and likely e-motors. It's certainly true of vehicle control software.

But when you're choosing to make seats, which almost every other auto company in the world buys from very efficient suppliers, and which really aren't any different from what everyone else is doing, it probably speaks to a broken design/sourcing system. Integration, which made sense in the cases of new technology, becomes an excuse not to fix the broken parts of the system. In this particular example it seems particularly true as there is little evidence that Tesla seats are even as good as those of its German competitors, let alone better.

 

[scaesare]

Sure, Tesla just skips the whole validation and testing part. or rather, they use their customers to do that.

There's certainly some argument to be made for that for a number of things.

But there are a while host of things where Tesla simply iterates and implements changes (many of them not user facing), that I suspect would be rather difficult if they relied on outsourcing nearly as much as traditional auto manufacturers do... who largely have outsourced the entire car (save the engine) to 3rd part suppliers.

 

[jaguar36]

But there are a while host of things where Tesla simply iterates and implements changes (many of them not user facing), that I suspect would be rather difficult if they relied on outsourcing nearly as much as traditional auto manufacturers do... who largely have outsourced the entire car (save the engine) to 3rd part suppliers.

Most (all?) other automakers iterate and improve on components that have issues over the life of the car as well. This is not unique to Tesla. It doesn't really matter if its a 3rd party who is making the component or not. What is unique to Tesla is that they will obsolete a huge feature set with no warning at a random time.

 

[ItsNotAboutTheMoney]

You mean like the 1800's, right? Screw JIT and risk based supply chains when a master craftsman can make each part by hand with a file in the workshop.

Up to 50 changes a week is funny.

Sort of like saying I can change all 4 tires on my car 10 times a day. OK...

Reading the article provides the context.

 

[scaesare]

... they will obsolete a huge feature set with no warning at a random time.

Cool, huh?

 

[McRat]

The discussion of "layered" capability in that article addresses that. You act like they are suggesting that they are ignoring outside vendor integration entirely.

And while procurement teams managing supply chains does work, it's not necessarily the most agile and effective method if fast iteration and new feature adoption are paramount.

There's likely some very valid reasons why it takes Detroit several years to introduce features and changes to a new vehicle model whereas Tesla does it in months. I suspect having the design, engineering, manufacturing, and production teams in the same foxhole is one of them.

The information is not handled by robots. It is not hand carried. It takes just as long for a Revision C dataset to go from one office at Tesla HQ to the next room over 20' away, as it does to go to 2000 miles away. It takes a human just long as long to make a decision whether the decision is made in house or 2000 miles away.

Welcome to the Internet Age and data interoperability.

"Detroit" (or whatever city you want to use) often puts millions of road miles on a new design before release. Skip that, and you get major features that do not work, or fail quickly. No, it's not cheaper to fix it later.

 

[scaesare]

The information is not handled by robots. It is not hand carried. It takes just as long for a Revision C dataset to go from one office at Tesla HQ to the next room over 20' away, as it does to go to 2000 miles away. It takes a human just long as long to make a decision whether the decision is made in house or 2000 miles away.

Welcome to the Internet Age and data interoperability.

"Detroit" (or whatever city you want to use) often puts millions of road miles on a new design before release. Skip that, and you get major features that do not work, or fail quickly. No, it's not cheaper to fix it later.

The barriers aren't just distance.

They are different orgs with their own communications styles, management structures, methodologies, priorities, resourcing policies, etc...

There's much more to dealing with change request than simply pressing "Send".

 

[McRat]

The barriers aren't just distance.

They are different orgs with their own communications styles, management structures, methodologies, priorities, resourcing policies, etc...

There's much more to dealing with change request than simply pressing "Send".

Yeah... OK. I send a video microscope image, IGES point data, B/P, cavity ID at 8am outlining a deviation. I get back an ECN that opens the tolerance. I submit the final report and send. Then I have lunch. The parts are being produced at the rate of hundreds per hour that day. Seriously.

This wasn't always the case. I opened my doors and 99% of my drawings were paper, and CAD files had little to no interoperability. Today, JPL sent us a CAD model, we quoted it, and they will have everything back by Monday. 1% of our drawings are paper, and there is an internet and email.

You know at least 2 of my customers besides JPL. Guaranteed. NDA restrictions do not allow me to say which 2.

 

[scaesare]

Yeah... OK. I send a video microscope image, IGES point data, B/P, cavity ID at 8am outlining a deviation. I get back an ECN that opens the tolerance. I submit the final report and send. Then I have lunch. The parts are being produced at the rate of hundreds per hour that day. Seriously.

This wasn't always the case. I opened my doors and 99% of my drawings were paper, and CAD files had little to no interoperability. Today, JPL sent us a CAD model, we quoted it, and they will have everything back by Monday. 1% of our drawings are paper, and there is an internet and email.

You know at least 2 of my customers besides JPL. Guaranteed. NDA restrictions do not allow me to say which 2.

I've no doubt it's possible to have responsive partners and manage change. Certainly that happens. But for every account like yours, there are accounts of red tape and finger pointing.

I also don't discount the value of collaborative partnerships across orgs at a distance. I also don't discount the value of having design and production folks on the same floor. Tesla is by far not the first org to have discovered this.

 

[mspohr]

Yeah... OK. I send a video microscope image, IGES point data, B/P, cavity ID at 8am outlining a deviation. I get back an ECN that opens the tolerance. I submit the final report and send. Then I have lunch. The parts are being produced at the rate of hundreds per hour that day. Seriously.

This wasn't always the case. I opened my doors and 99% of my drawings were paper, and CAD files had little to no interoperability. Today, JPL sent us a CAD model, we quoted it, and they will have everything back by Monday. 1% of our drawings are paper, and there is an internet and email.

You know at least 2 of my customers besides JPL. Guaranteed. NDA restrictions do not allow me to say which 2.

I think you've missed the point. It's not about just making something to spec. It's about the (much more important) design process which requires complex interaction among engineers and production people.