Just curious, does your model differentiate between "days of inventory outstanding" between the Model 3 and the S/X?
In particular we have a pretty good notion of how Model 3 delivery latencies look like currently, if we look at the
"VIN to Delivery time in the last 3 weeks" chart of Troy's tracker.
While the sample of people reporting is self-selecting and thus not representative statistically, it's unlikely to be particularly biased by expected delivery time - in fact longer delivery time gives people more time to register their VIN in the tracker so I'd expect the bias towards entries with longer delivery times.
According to the histogram there the average delivery time for Model 3's is somewhere between 10-14 days from VIN to delivery. To that we have to add a couple of days of manufacturing time - which should be no more than 7 days (possibly much less). I.e. 20 days looks like an upper bound for M3 days of inventory.
For the Model S/X 75% of the demand (rest of the world) is at least 1 month of shipping distance away from Fremont, plus European deliveries also have a (re-)assembly step. DIO should be a lot higher, especially considering that Tesla cannot stop the Fremont factory at all, so cars are being produced whether it's conveniently sold by the end of the quarter or not. I'd guesstimate 45 days on average.
Another factor is that the Model 3 is optimized for mass manufacturing: fewer components and faster assembly. This could shave off another couple of days from the typical "new order" -> "delivery" delay and reduce the DIO figure.
A very rough DIO estimate for the last 3 quarters plus Q3:
MSX at 2k/week rate:
- MSX DIO is 45 days
- MSX ASP is $105k
- Cost basis is 75% with a 25% gross margin
- At 2k/week production outstanding inventory is 45/7*2,000*75%*$0.105m = $1,012m
This is close to observed/reported MSX finished goods inventory levels.
M3 at 2k/week rate (end of Q1-ish):
- M3 DIO is 20 days
- M3 ASP is $59k
- Cost basis is 120% with a -20% gross margin at a 0.8k/week run-rate (end of Q1)
- At 1k/week outstanding inventory is 20/7*800*120%*$0.059m = $161m
- Cost basis is 100% with a 0% gross margin at a ~4k/week run-rate (end of Q2)
- At 2k/week outstanding inventory is 20/7*4,000*100%*$0.059m = $674m
- Cost basis is 85% with a 15% gross margin guided at a ~5k/week run-rate (end of Q3)
- At 5k/week outstanding inventory is 20/7*5,000*85%*$0.059m = $716m
Cross-check of these estimates:
- 2017/Q4 finished goods estimate: $1,012m, Q4 reported value: $1,013m (99% accurate)
- 2018/Q1 finished goods estimate: $1,173m, Q1 reported value: $1,125m (96% accurate)
- 2018/Q2 finished goods estimate: $1,686m, Q2 reported value: $1,721m (98% accurate)
- 2018/Q3 finished goods estimate: $1,728m
So by separating out the Model 3's different manufacturing and delivery characteristics we can match the actual inventory levels pretty well (which is not a surprise, there's more parameters to the linear equations than data points
![Wink ;) ;)](data:image/gif;base64,R0lGODlhAQABAIAAAAAAAP///yH5BAEAAAAALAAAAAABAAEAAAIBRAA7)
), but interestingly the Q3 inventory estimate is only 2.5% higher than Q2 finished goods levels.
Even if we factor in a few other expected changes for Q3, such as a slightly higher MSX rate, I don't see a bigger inventory growth than 5%. One reason is that M3 gross margins improve, which reduces the CoG calculation of the inventory valuation, the other reason is that the 4k/week exit rate at the end of Q2 only increases by 25% to 5k/week at the end of Q3.
While you are calculating +40% based on DIO - which is an order of magnitude difference.
Which is weird, and I love weird results!
Any idea what is going on here?