2015 Q4 discussion thread for delivery numbers

[Auzie]

Great post, Auzie. I like that you are going for 52,500 and not 50,000 - which I think many bulls would be happy with given the current sentiment.

For the record, leasing was first announced in April 2013, I think it was part of the 5-part trilogy: Tesla Announces Model S Leasing From $500/Month* - But Not Really (Video With CEO) | Inside EVs

And business leasing seems to have been announced April 2014: Tesla's new $408 per month business lease deal for Model S reminds us of last 'revolutionay' deal

Thanks, Mike. If business leasing was announced in April 2014, that aligns with the leased cars being reported for the first time in q2 2014.

If you are looking for another vin data point, I just confirmed an order for a 90d with a vin of 114390. December delivery. I'm assuming the vin everyone quotes is just the last six digits of the number under the picture of the car on the my tesla page right?

Thanks for the data point.

It is possible that some higher VINs get delivered before some lower VINs get delivered.

VINs can get scrambled for various reasons: cars batching in production, shipment schedules to various destinations, preference given to some models, etc. Such scrambling would imply that the delivered number of cars is lower than the highest VIN delivered, as there might be some lower VINs that get delivered later for various reasons.

Having said that, if there is no single VIN reported in the 114000+ that means that the delivered quarterly number will fall shorter than my estimates above. The appearance of your VIN does not guarantee my estimated deliveries, but it allows the possibility of hitting the estimated targets.

Can we really read anything definitive into the vins? Has anyone successfully drawn any conclusions in past quarters studying vins? I'm sure similar analysis has been attempted in the past. I've exited all my options positions prior to earnings but I could be persuaded to get back in...

VIN counting was effective during the early Model S days, but after a lot of people got burned on a bad ER, the general consensus was that the method was not accurate enough anymore. Still, I'd like to see some members see if they can find any flaws in Auzie's logic here.

It would be interesting to establish if these incorrect past estimations correspond to Tesla's expansion offshore and the scrambled VINs due to shipments schedules. I was not really following the deliveries based on VIN estimations in the past so I am not able to correlate the past estimates to the possible events that caused the estimation errors.

It might be impossible to get accurate calculations/predictions of quarterly deliveries based on reported VINs due to VIN scrambling, as I stated above. As the number of delivered cars/quarter increase, the VIN scrambling effect is likely to play a relatively smaller role and the delivery estimation error might become smaller. Variation of several hundred or even few thousand cars is introducing relatively smaller error when Tesla starts delivering tens of thousand cars/quarter.

I also expect the inventory cars to go up in proportion (not linear though) to rising deliveries from quarter to quarter. More cars are needed for more stores, service centers and to fill the larger transit pipeline.

It would be helpful to try to estimate the VIN band - the difference between the highest delivered VIN and the lowest not delivered VIN in a quarter. The size of that band might help estimate the size of the delivery prediction error.

Or am I going insane with this, not sure

 

[Quant]

......

Or am I going insane with this, not sure :biggrin:

Recently ( in the last 2 weeks ) . Elon Musk was quoted as saying current production rate was about 1,200 a week . So, I'm not sure your numbers and expectations jive with current production rates.

Sorry, I don't have the quote in the interview handy....

 

[mrdoubleb]

Thanks, Mike. If business leasing was announced in April 2014, that aligns with the leased cars being reported for the first time in q2 2014.



Thanks for the data point.

It is possible that some higher VINs get delivered before some lower VINs get delivered.

VINs can get scrambled for various reasons: cars batching in production, shipment schedules to various destinations, preference given to some models, etc. Such scrambling would imply that the delivered number of cars is lower than the highest VIN delivered, as there might be some lower VINs that get delivered later for various reasons.

Having said that, if there is no single VIN reported in the 114000+ that means that the delivered quarterly number will fall shorter than my estimates above. The appearance of your VIN does not guarantee my estimated deliveries, but it allows the possibility of hitting the estimated targets.





It would be interesting to establish if these incorrect past estimations correspond to Tesla's expansion offshore and the scrambled VINs due to shipments schedules. I was not really following the deliveries based on VIN estimations in the past so I am not able to correlate the past estimates to the possible events that caused the estimation errors.

It might be impossible to get accurate calculations/predictions of quarterly deliveries based on reported VINs due to VIN scrambling, as I stated above. As the number of delivered cars/quarter increase, the VIN scrambling effect is likely to play a relatively smaller role and the delivery estimation error might become smaller. Variation of several hundred or even few thousand cars is introducing relatively smaller error when Tesla starts delivering tens of thousand cars/quarter.

I also expect the inventory cars to go up in proportion (not linear though) to rising deliveries from quarter to quarter. More cars are needed for more stores, service centers and to fill the larger transit pipeline.

It would be helpful to try to estimate the VIN band - the difference between the highest delivered VIN and the lowest not delivered VIN in a quarter. The size of that band might help estimate the size of the delivery prediction error.

Or am I going insane with this, not sure :biggrin:

Just a word of caution: we have discussed several times how VINs are no longer a good method to predict delivery/sales as Tesla has taken precautions against what you are doing nine this analysis. As far as I know VINs are no longer strictly sequential and do not correlate to production order. Also, Tesla may be skipping VINs, so just because you se No. 114,000 issued, it does not mean 114k have been misused, they may have skipped a few thousand numbers.

I am by no means an expert on VIN, so you may be a step ahead of my understanding of this, just wanted to make sure you are aware of some of the difficulties of predicting sales by VIN.

 

[esk8mw]

Thanks, Mike. If business leasing was announced in April 2014, that aligns with the leased cars being reported for the first time in q2 2014.



Thanks for the data point.

It is possible that some higher VINs get delivered before some lower VINs get delivered.

VINs can get scrambled for various reasons: cars batching in production, shipment schedules to various destinations, preference given to some models, etc. Such scrambling would imply that the delivered number of cars is lower than the highest VIN delivered, as there might be some lower VINs that get delivered later for various reasons.

Having said that, if there is no single VIN reported in the 114000+ that means that the delivered quarterly number will fall shorter than my estimates above. The appearance of your VIN does not guarantee my estimated deliveries, but it allows the possibility of hitting the estimated targets.





It would be interesting to establish if these incorrect past estimations correspond to Tesla's expansion offshore and the scrambled VINs due to shipments schedules. I was not really following the deliveries based on VIN estimations in the past so I am not able to correlate the past estimates to the possible events that caused the estimation errors.

It might be impossible to get accurate calculations/predictions of quarterly deliveries based on reported VINs due to VIN scrambling, as I stated above. As the number of delivered cars/quarter increase, the VIN scrambling effect is likely to play a relatively smaller role and the delivery estimation error might become smaller. Variation of several hundred or even few thousand cars is introducing relatively smaller error when Tesla starts delivering tens of thousand cars/quarter.

I also expect the inventory cars to go up in proportion (not linear though) to rising deliveries from quarter to quarter. More cars are needed for more stores, service centers and to fill the larger transit pipeline.

It would be helpful to try to estimate the VIN band - the difference between the highest delivered VIN and the lowest not delivered VIN in a quarter. The size of that band might help estimate the size of the delivery prediction error.

Or am I going insane with this, not sure :biggrin:

I should also mention that I was quoted December delivery when I ordered the car - I do not yet have a firm December delivery date. Another post in the Model S forum indicates that delivery dates are being pushed back for some people so it's entirely possible that my 114 VIN will not be delivered in 2015 - I haven't heard either way yet.

 

[hobbes]

It would be helpful to try to estimate the VIN band - the difference between the highest delivered VIN and the lowest not delivered VIN in a quarter. The size of that band might help estimate the size of the delivery prediction error.

Or am I going insane with this, not sure :biggrin:

No, you´re not! Just thinking a bit farther than everyone else. I like the idea of the VIN band. The hard question is as always with those approaches: how to get the data?! Is the anecdotal data points from posts enough?

 

[hobbes]

Wow, found a very interesting post from MrBacardi (the guy who collects and posts the Norwegian registrations on elbilforum.no)! From this it seems at first sight that they do not skip any VINs and they are definitely batched by color. Also, the cars being batched by color would mean they are produced in the order of the VINs, at least for Norway - though Norway is a special case as they get their cars directly and not via the Netherlands, it is likely that they produce a whole batch for them at a time and put them on a ship. Post is of Aug 27. So I would guess those were cars delivered in Aug.

Just to check order of magnitude 114390 (highest known VIN scheduled for delivery, at late Dec according to post above) - 82616 (highest VIN of the Post quoted, assuming early Aug delivery) = 31774 VINs in about 5 months, about 6000/month, 1500/week - that is too much for it to be actual cars, but order of magnitude seems about right. So maybe the VINs are without gaps for a batch like the one reported from Norway, but there are gaps in between batches?

I have attached a list that shows a VIN-sequence of Norwegian 70D cars. The list contains only 70D cars.

As you can see, the VINs are grouped by car-type, then by color.

But the 9th character in the VIN seems to be incrementing along the VIN, and then "wrapping" back around. Is this some kind of check-digit?


View attachment 92009

Pic from the quote:

 

[FredTMC]

Wow, found a very interesting post from MrBacardi (the guy who collects and posts the Norwegian registrations on elbilforum.no)! From this it seems at first sight that they do not skip any VINs and they are definitely batched by color. Also, the cars being batched by color would mean they are produced in the order of the VINs, at least for Norway - though Norway is a special case as they get their cars directly and not via the Netherlands, it is likely that they produce a whole batch for them at a time and put them on a ship.

interesting. Color grouping makes sense since Body-In-white (metal shell) goes to paint shop. I'd think it's more efficient for the paint shop to shoot a batch of cars of the same color. I have no specific knowledge. If anyone knows, let us know. Thx!

 

[esk8mw]

interesting. Color grouping makes sense since Body-In-white (metal shell) goes to paint shop. I'd think it's more efficient for the paint shop to shoot a batch of cars of the same color. I have no specific knowledge. If anyone knows, let us know. Thx!

Well, mine (114390) is a blue 90D, for what it's worth. Tesla reassured me mid-December is a conservative estimate, though I'm not in production yet.

 

[Auzie]

Just a word of caution: we have discussed several times how VINs are no longer a good method to predict delivery/sales as Tesla has taken precautions against what you are doing nine this analysis. As far as I know VINs are no longer strictly sequential and do not correlate to production order. Also, Tesla may be skipping VINs, so just because you se No. 114,000 issued, it does not mean 114k have been misused, they may have skipped a few thousand numbers.

I am by no means an expert on VIN, so you may be a step ahead of my understanding of this, just wanted to make sure you are aware of some of the difficulties of predicting sales by VIN.

I doubt that Tesla pays attention to speculating investors, and I doubt that Tesla would go out of their way to mess up their complex scheduling/supply chain/production for the sole reason of circumventing speculators speculating. Speculators will speculate no matter what...

My guess is that VINs are assigned in a particular order unknown to us, no skipping, as such skipping serves no purpose and would be likely to introduce issues in Tesla's supply chain and production scheduling.

Scrambling of VINs is more likely due to various reasons mentioned up-thread imho.

Wow, found a very interesting post from MrBacardi (the guy who collects and posts the Norwegian registrations on elbilforum.no)! From this it seems at first sight that they do not skip any VINs and they are definitely batched by color. Also, the cars being batched by color would mean they are produced in the order of the VINs, at least for Norway - though Norway is a special case as they get their cars directly and not via the Netherlands, it is likely that they produce a whole batch for them at a time and put them on a ship. Post is of Aug 27. So I would guess those were cars delivered in Aug.

There is likely to be over-laid batching on top of some color batching due to other reasons as well.

Color batching makes sense for older manufacturing facilities. The design/layout of the paint line will determine the need for batching and the optimum batch size. It is possible that Tesla has overcome the need for large batching or reduced the batch size with paint line layout (workflow) design. I read somewhere that Tesla has paint robots that use quick exchange paint cartridges. If the paint booth is small, that would allow for efficient quick exchanges and efficient workflow. The size of the cartridge determines the batch size. It could be just a few cars.

It is possible that what appears to be skipped VINs are inventory cars (loaners, stores, sc), transit cars, or anything else.

I would be very surprised if Tesla would skip VINs because there is no upside in skipping them and there are downsides.

My understanding is that Tesla uses in house developed ERP system (Enterprise Resource Planning). For people not familiar what ERP is, SAP is more common industry used ERP. SAP is awfully complex, inflexible, prone to errors, hence Tesla developed their own system. ERP system must use VIN as quite relevant data point, and messing with VIN would either throw any ERP out of the whack (very costly to business) or would require extra programming to counter messing with VINs. I consider both of these unlikely.

My view is that the VINs are assigned in an orderly manner but the delivery gets scrambled to an extent, say few thousand VINs may get out of order at delivery. That VIN band error is huge on the delivery of several thousand cars/quarter but is not so significant when Tesla is delivering say 20k or more/cars per quarter.

 

[mrdoubleb]

Auzie, you may have missed "the great VIN counting fiasco" here at TMC a year or two ago. I don't have links to the old threads, but the story was basically:
- Tesla VINS were completely sequential and the TMC community did a darn good job of counting them and predicting sales, so much so, that some media sites started to rely on this analysis too.
- At some point things started to go south when Tesla decided it did not want half the internet tracking it's sales like that - they were much more secretive about this back then.
- While obviously they never released a PR citing TMC as the cause, at some point they changed the sequential method and some people here started to get too excited about big VINs coming out and overestimated sales by a good margin.
- Since then it is "common knowledge" in the investors' forum that predicting sales by counting VINs is not a reliable method.

Maybe a more senior TMC member recalls more details and can provide you links to the old threads. (I may be wrong on some of the details, this is just my recollection of events).

These days we get quarterly delivery numbers straight from the horse's mouth, plus we have solid registration data collected all over Europe every month and insideEVs seems to be OKish for ballpark US figures.

 

[hobbes]

My view is that the VINs are assigned in an orderly manner but the delivery gets scrambled to an extent, say few thousand VINs may get out of order at delivery. That VIN band error is huge on the delivery of several thousand cars/quarter but is not so significant when Tesla is delivering say 20k or more/cars per quarter.

Throwing this out for discussion, shows VINs vs. their delivery date in Q3 (generated from Model S delivery update google sheet)


<update: the was an error in the plot I made which I just realized a day after making it: the time axis was not really a linear time axis, but just was in fact determined by the line number in the table I used, but labelled with the according dates. Will make a new one later.>

 

[doctoxics]

Not sure Tesla gets to 50,000 for 2015. Appears that the current Model S line is maxed at approximately 1,000 cars/week. S production in Q4 is not likely to be greater than 1,000/week unless the new S/X line is used to make Ss; that does not seem likely considering the need to work out the kinks in X ramp-up on the S/X line. Is there any basis for thinking that the old S line has been increased or that the S/X line is making both Ss and Xs? If not, seems like 47,000 cars in 2015 is the maximum that can be expected (assumes no more than 1,000 Xs made in Q4).

 

[FredTMC]

Not sure Tesla gets to 50,000 for 2015. Appears that the current Model S line is maxed at approximately 1,000 cars/week. S production in Q4 is not likely to be greater than 1,000/week unless the new S/X line is used to make Ss; that does not seem likely considering the need to work out the kinks in X ramp-up on the S/X line. Is there any basis for thinking that the old S line has been increased or that the S/X line is making both Ss and Xs? If not, seems like 47,000 cars in 2015 is the maximum that can be expected (assumes no more than 1,000 Xs made in Q4).

tesla will absolutely make more MS than 1000 per week in Q4. Proof: Here's a snippet from their ER from last quarter. Notice than one item is the 'Model S Body Center'. Allows for increased model S Body-in-white production.

We just concluded a planned one-week Fremont
factory shutdown and made changes in stamping, Model S
body center, drive unit production, battery module and pack
production and general assembly to allow for an elevated level of production and efficiency.

 

[Auzie]

Auzie, you may have missed "the great VIN counting fiasco" here at TMC a year or two ago. I don't have links to the old threads, but the story was basically:
- Tesla VINS were completely sequential and the TMC community did a darn good job of counting them and predicting sales, so much so, that some media sites started to rely on this analysis too.
- At some point things started to go south when Tesla decided it did not want half the internet tracking it's sales like that - they were much more secretive about this back then.
- While obviously they never released a PR citing TMC as the cause, at some point they changed the sequential method and some people here started to get too excited about big VINs coming out and overestimated sales by a good margin.
- Since then it is "common knowledge" in the investors' forum that predicting sales by counting VINs is not a reliable method.

Maybe a more senior TMC member recalls more details and can provide you links to the old threads. (I may be wrong on some of the details, this is just my recollection of events).

These days we get quarterly delivery numbers straight from the horse's mouth, plus we have solid registration data collected all over Europe every month and insideEVs seems to be OKish for ballpark US figures.

I am not saying my estimations based on assigned VIN are correct, it is just an attempt at estimation. These estimations are likely insufficiently precise to provide meaningful actionable information in regard to confirming that Tesla will reach the target.

If we had a situation in which the only VINs we were seeing on the threads were in the say 109000 range, then the above estimations can be used to deduce that Tesla for sure is likely to miss the target. In that respect, assigned VINs do provide some information.

Regardless of the theory validity, I have difficulty believing that Tesla deliberately scrambles VINs to confuse public. Do we have any evidence or data to confirm that? I find it more likely that Tesla becomes more secretive with their data. Their process is complex enough without extra deliberate attempts at making it more so by playing games with VINs. My 2c and I can be wrong.

Not sure Tesla gets to 50,000 for 2015. Appears that the current Model S line is maxed at approximately 1,000 cars/week. S production in Q4 is not likely to be greater than 1,000/week unless the new S/X line is used to make Ss; that does not seem likely considering the need to work out the kinks in X ramp-up on the S/X line. Is there any basis for thinking that the old S line has been increased or that the S/X line is making both Ss and Xs? If not, seems like 47,000 cars in 2015 is the maximum that can be expected (assumes no more than 1,000 Xs made in Q4).

If Tesla's run rate is only 1000cars/week, then they are unlikely to hit the yearly target.

 

[doctoxics]

Improvements in model S Body-in-white production:

Why then did they only deliver ~11,500 Model Ss in Q3? Seems that Model S production has been about 1,000/month for the past six months.

 

[FredTMC]

Why then did they only deliver ~11,500 Model Ss in Q3? Seems that Model S production has been about the 1,000/month level for the past six months.

The factory retooling was in that Quarter. Lost over a week of production.

They produced over 12800 which was a blend of before and after. As they've said, they were driving for "efficiency". Not max throughput.

I toured the factory a month ago. Not my first time. I witnessed the changes made.

 

[schonelucht]

Knowing there were 11,580 deliveries in Q3, I dare to say that the VINs and number of vehicles delivered are correlated. Would be interesting to make the same graph with production complete dates etc... Maybe tomorrow.

Very interesting graph! Thanks. Also interesting that the slope of the fitted lines 91 cars per day suggest (real number of deliveries were 125). If you graph the same for earlier quarters, is the same pattern there with the higher variation near the end of the quarter?

 

[hobbes]

Sorry, removed the plot I made above until I have time to fix it:

Throwing this out for discussion, shows VINs vs. their delivery date in Q3 (generated from Model S delivery update google sheet)


<update: the was an error in the plot I made which I just realized a day after making it: the time axis was not really a linear time axis, but just was in fact determined by the line number in the table I used, but labelled with the according dates. Will make a new one later.>

 

[hobbes]

Throwing this out for discussion, shows VINs vs. their delivery date in Q3 (generated from Model S delivery update google sheet)


<update: the was an error in the plot I made which I just realized a day after making it: the time axis was not really a linear time axis, but just was in fact determined by the line number in the table I used, but labelled with the according dates. Will make a new one later.>

O.k., fixed. Doesn´t really look all that different only that the dates are evenly spread their axis now. Also added a rectangle with a height corresponding to the 11580 deliveries in that quarter (its vertical placement is random):

 

[MikeC]

I am not saying my estimations based on assigned VIN are correct, it is just an attempt at estimation. These estimations are likely insufficiently precise to provide meaningful actionable information in regard to confirming that Tesla will reach the target.

If we had a situation in which the only VINs we were seeing on the threads were in the say 109000 range, then the above estimations can be used to deduce that Tesla for sure is likely to miss the target. In that respect, assigned VINs do provide some information.

Regardless of the theory validity, I have difficulty believing that Tesla deliberately scrambles VINs to confuse public. Do we have any evidence or data to confirm that? I find it more likely that Tesla becomes more secretive with their data. Their process is complex enough without extra deliberate attempts at making it more so by playing games with VINs. My 2c and I can be wrong.

From this Norwegian data, it appears that VINs are assigned in an orderly manner based on region, car type, and then color: Decoding Tesla Model S VINs - Page 33

I understand MrBacardi who posted that has a list of every Model S delivered in Norway? It would be interesting to have a bigger sample size, with different types of cars delivered in Q4.

It looks like only one VIN out of 50+ was scrapped probably due to some type of production problem (or maybe a single orphan car was urgently made for some reason). I think the biggest risks of VIN counting is the filling of the pipeline and the widely varying lag between production and complete for different cars. But I agree with Auzie that it can give us viable information about whether production in the quarter was sufficient, at least.

Also, is there a tracker like this but for Q4? Model S Order Delivery - Google Sheets