Bloomberg Model 3 Tracker

[jhm]

It may be good to devote a thread to following the Bloomberg Model 3 Tracker.

Bloomberg Model 3 Tracker

Current as of Aug 24, 2018 indications are 6278 per week and 75,910 cumulative.

What might this tell us about Q3 production? Cumulative at end of Q2 was 41,030. So 54 days into Q3 has 34,880.

This is an average rate of 646/day or 4521/week. Extending this average rate out over 91 days in Q3 arrives at 58,779 as a conservative projection.

As an alternative, suppose that for the remaining 37 days of Q3, Tesla averages the current rate of 6278 per week. This is 33,148 over the remaining 37 days, leading to a Q3 total of 68,084.

Tesla guided 50k to 55k for Q3. So far 35k is done 54 days into the quarter. The two extrapolations above suggest that 59k to 68k are plausible.

 

[jhm]

Bloomberg Model 3 Tracker

My extrapolation of the Bloomberg Model 3 Tracker suggest that Q3 production could reach 59k to 68k.

 

[Fact Checking]

Bloomberg Model 3 Tracker

My extrapolation of the Bloomberg Model 3 Tracker suggest that Q3 production could reach 59k to 68k.

Note that the Troy/TesLike registrations project is currently estimating Q3 Model 3 production of 51.5k:

Model 3 Order Tracker (Published Web Version)

They have a good prediction track record of 96% and 98% for Q1 and Q2.

Tesla guided 50-55k Q3 production - which is conservative at 4k/week, and which might be the reflection of some known component bottleneck: for example battery packs.

So I'd caution against excessive optimism regarding Q3 production, unless Tesla can keep up the high pace of production for the whole of September as well.

 

[Johan]

Note that the Troy/TesLike registrations project is currently estimating Q3 Model 3 production of 51.5k:

Model 3 Order Tracker (Published Web Version)

They have a good prediction track record of 96% and 98% for Q1 and Q2.

Tesla guided 50-55k Q3 production - which is conservative at 4k/week, and which might be the reflection of some known component bottleneck: for example battery packs.

So I'd caution against excessive optimism regarding Q3 production, unless Tesla can keep up the high pace of production for the whole of September as well.

I'd tend to agree with this. Something else that would support this notion of caution is the fact that the 5000/week push apparently came at some cost when it comes to quality (high reworking percentage on cars that left the line during that time), which is also supported by anectodal evidence from the forum here from people taking delivery shortly after the push.

Tesla have no pressure to exceed their guidance when it comes to production in Q3 if that means taking shortcuts when it comes to quality control. Again, a perfect example of why it's only natural for Elon want to delist the company.

 

[KarenRei]

Bloomberg tracker predicts a significant production rate drop over the next three weeks. The reason why isn't clear.

 

[Chickenlittle]

Bloomberg tracker predicts a significant production rate drop over the next three weeks. The reason why isn't clear.

What does that tell you about the tracker. Supposedly measure deliveries and now claims to predict the future deliveries by sampling owners?

 

[generalenthu]

According to skabooshka, the factory slowed down in the last few days, but before that, there were some rumours of Tesla hitting 6k model 3s a week.

This fits well with the Bloomberg narrative if you assume that this lags true production by a week to 10 days. Bloomberg is projecting a drop as well. This actually coincides with Elon sleeping at the factory again, and I suppose efforts are under way to de-bottleneck and make a run for 7-8k.

 

[KarenRei]

What does that tell you about the tracker. Supposedly measure deliveries and now claims to predict the future deliveries by sampling owners?

1) You're grossly oversimplifying their model.
2) They've been accurate about such things in the past.

 

[Papafox]

 

[MP3Mike]

What does that tell you about the tracker. Supposedly measure deliveries and now claims to predict the future deliveries by sampling owners?

The Bloomberg tracker tracks/predicts production not deliveries.

 

[dondy]

1) You're grossly oversimplifying their model.
2) They've been accurate about such things in the past.

"accurate"? Let see last week of Q2. Tesla reported 5000 (a couple more), Bloomberg 3900.

More of it when you check web.archive.org snapshots you will see they do all the time backlog "patching".

Their "predictions" correspond to the current USA sells.

 

[dondy]

The Bloomberg tracker tracks/predicts production not deliveries.

They base their estimations on deliveries, at least that's what they claim.

 

[Orbit]

Extending this average rate out over 91 days in Q3 arrives at 58,779 as a conservative projection.

So they might actually beat guidance on M3 production. Likely to beat it even.
Strange feeling....
After all the M3 misses it somehow felt like they would obviously miss again, at least to me.

 

[RobStark]

What does that tell you about the tracker. Supposedly measure deliveries and now claims to predict the future deliveries by sampling owners?

It tries to predict production not deliveries.

They have been pretty accurate last 3 quarters predicting the entire quarter.

We will never know how accurate they are week to week because Tesla will never publish those numbers.

 

[KarenRei]

"accurate"? Let see last week of Q2. Tesla reported 5000 (a couple more), Bloomberg 3900.

More of it when you check web.archive.org snapshots you will see they do all the time backlog "patching".

Their "predictions" correspond to the current USA sells.

Bloomberg has some "smoothing"; it doesn't respond to "instantaneous" changes in production rates (like a 5k burst). It's been this same way since the beginning. You can complain about that from a long side ("it's not showing our new instantaneous peak!"), but it also bites shorts, too ("It's still showing production even though they're down!").

Remember that Tesla went from that 5000 Model 3s into a downtime the week after. Combined with smoothing, you end up with only about 4000 vehicles around that point in time.

1) Web.archive.org doesn't work on the Bloomberg tracker, so you've clearly never tried this. Go do so. It gets confused and reports over 8000 vehicles per week. Something about the scripting or something.

2) I'm comparing to past screenshots right now, and not seeing any "backlog patching", at least not to any meaningful degree. Example here.

3) You sum up the old numbers and they match up quite well with reported figures.

I'm sorry that you don't like what the Bloomberg tracker numbers say, but take solace in the fact that the shorts don't either.

 

[Cloxxki]

From the thick many dots of VIN reports up to 80,000 I suspect they may have passed 80,000 already. Much higher have been reported as well. The tracker seems to still be catching up to reality. I'd not be surprised to see 66,000 Model 3's produced in Q3. Could end up higher also.
And then there are the in-transit cars from late Q2. If Tesla manages to have a small amount in transit (all Cali deliveries last 2 weeks), they might hit close to 100,000 cars sold in Q3. As opposed to 53K last quarter.

 

[Big Earl]

Bloomberg tracker predicts a significant production rate drop over the next three weeks. The reason why isn't clear.

What does that tell you about the tracker. Supposedly measure deliveries and now claims to predict the future deliveries by sampling owners?

The tracker uses some historical modeling to estimate what's going to happen next. If you look at all of the past production spikes, they're usually followed by a dip (for various reasons). The model is expecting another dip after this spike. A few more consecutive weeks of steady production will smooth out the trend.

 

[pnungesser]

Bloomberg has some "smoothing"; it doesn't respond to "instantaneous" changes in production rates (like a 5k burst). It's been this same way since the beginning. You can complain about that from a long side ("it's not showing our new instantaneous peak!"), but it also bites shorts, too ("It's still showing production even though they're down!").

Remember that Tesla went from that 5000 Model 3s into a downtime the week after. Combined with smoothing, you end up with only about 4000 vehicles around that point in time.

1) Web.archive.org doesn't work on the Bloomberg tracker, so you've clearly never tried this. Go do so. It gets confused and reports over 8000 vehicles per week. Something about the scripting or something.

2) I'm comparing to past screenshots right now, and not seeing any "backlog patching", at least not to any meaningful degree. Example here.

3) You sum up the old numbers and they match up quite well with reported figures.

I'm sorry that you don't like what the Bloomberg tracker numbers say, but take solace in the fact that the shorts don't either.

My only complaint is they went 10 days without updating the data. The VIN's pulled data was 18k behind and the reported VIN's in the wild also had not been updated. (you can hover over the dots and see the date it was added). Having the most current data certainly does not slew the conclusions, especially with the smoothing included in the model.

 

[beachbum77]

My only complaint is they went 10 days without updating the data. The VIN's pulled data was 18k behind and the reported VIN's in the wild also had not been updated. (you can hover over the dots and see the date it was added). Having the most current data certainly does not slew the conclusions, especially with the smoothing included in the model.

The tracker largely depends on reader-submitted data on VIN sightings. However, they began realizing (and reported) that people were submitting VINs at time of assignment not knowing whether the cars had actually been built yet. It seemed Tesla started pushing out VIN notifications quickly presumably since that was the determinate in reservations becoming non-refundable. That practice has since stopped. While VINs are still being assigned buyers are not being given their VINs until just days before the scheduled delivery. So the data should become much more reliable going forward. That could explain the drop in reported VINs for the coming weeks.

 

[KarenRei]

My only complaint is they went 10 days without updating the data. The VIN's pulled data was 18k behind and the reported VIN's in the wild also had not been updated. (you can hover over the dots and see the date it was added). Having the most current data certainly does not slew the conclusions, especially with the smoothing included in the model.

It explains how they handle VINs on the right-hand column; small VIN batches are grouped together to avoid skewing the algorithm.