I'm been reflecting on how my analysis of all of the good data crowd sourced here failed to forecast this delivery miss. What I realized is that we have some good indicators of demand, and of demand relative to production, but not of the production rate, with one exception. Going through the metrics:
- VIN counting - a good indicator of absolute demand. VINs are issued for customer orders and for inventory cars that Tesla expects to sell or which will replace existing demos that will in turn be sold. But that rate that VINs are issued does not tell us the rate at which they will be produced
- Tesla website wait times - a good indicator of demand relative to production. Wait times can increase because demand picks up or because production slows down and there is no way to tell which it is.
- ev-cpo historical data - shows how many inventory cars are being sold relative to prior quarters, but does not capture all of the sales. Also there is no way of knowing what Tesla's planned mix is between inventory sales and custom orders.
- Factory tour observations - anecdotal evidence of the assembly line not operating during tours or the parking lot being relatively empty can tell us if the factory is running. However when it is running we have no way of knowing how many cars are being produced per day/week/month.
- Delivery spreadsheets - I was looking at these to see whether Tesla was pressing to deliver December orders before quarter end. When they did not I assumed it meant they had the quarter made. Come to find out it was because they didn't have enough production to fill those orders. However
@bonaire pointed out today that the relative number of spreadsheet entries between Q3 and Q4 correlated highly to the absolute delivery numbers in those quarters. This will definitely be one to watch.
- InsideEVs monthly reports. Looks like these are pretty accurate. I thought they were missing the extreme geographical batching and the shift toward more inventory cars being sold. I guess not.
My other observation is that every time Tesla does a hardware transition they miss their forecast delivery number. This happened in 4Q14 when the D and AP1 hardware was introduced, in 2Q16 with the exterior refresh, and again this quarter with the AP2 upgrade. It just takes longer than Tesla believes to "slipstream" these hardware changes into the manufacturing process.
Contrast this with traditional auto makers who shut down the factories for weeks once per year for the model year changeover. My experience has been that if you try to order a new model year BMW in July in the US you are lucky to get it by Thanksgiving. The assembly line is shut down at least for August, starts producing the new model in September, and of course there is the shipping delay. Tesla believes they can do these kind of changes in 10-14 days and it just hasn't happened that way. Tesla, please do not refresh the interior until you figure this out.