Model 3 Tracker Misses Tesla’s Production Shortfall

(Bloomberg) -- Bloomberg’s Tesla Model 3 tracker blew it this time.

In early 2018, an experimental tool developed by Bloomberg started tracking one of the biggest mysteries in the auto industry: production of Tesla Inc.’s all-important Model 3, the electric automaker’s first car aimed at the mass market. After four straight quarters of forecasts coming closer than Wall Street estimates, Bloomberg’s model finally missed one. In a big way. 

On Wednesday, Tesla reported that production had reached a new high of 62,950, far less than the 79,130 projected by the Model 3 tracker. (You can read about Tesla’s other numbers here.)

The tracker relies on two sets of vehicles identification numbers, VINs: those Tesla registers in batches with the U.S. government prior to production and those submitted by new owners to Bloomberg after delivery. Tesla’s expansion into overseas markets throughout the quarter created increased risk, as we warned readers, that the model may be running too hot. There was no way to be certain without additional verified data from Tesla, which came from the company with Wednesday’s production report. 

As expected, Bloomberg’s estimates were too high—and, in fact, missed Tesla’s actual production by a margin larger than anticipated. Tesla’s production came in below Wall Street analyst estimates, too, although not by as much. 

It appears the root of the Model 3 Tracker’s miss was caused by a change in Tesla’s VIN registrations that came with international deliveries.

Tesla registers VINs in batches prior to production. It uses unique sequences of VINs for each drivetrain type—rear-wheel drive or dual motor, for example—and keeps a extra cache of each VIN type that never gets fully depleted. When additional international deliveries started this past quarter, that required new strings of international VINs for each type of drivetrain, doubling the caches of extra VINs on hand. Tesla registered more than 140,000 VINs for the quarter, an unprecedented haul.

This registration pattern affected the tracker’s other dataset as well, which relies on VINs reported to Bloomberg by new car owners. In the chart below, large gaps opened up in the dataset.

The cluster of North American VINs at the top of the above chart marks the start of Tesla’s new lower-priced Standard Range Plus model. The tracker aims to exclude the largest gaps like this, but the exclusion rules appear to have not been sensitive enough. Tesla tends to deliver VINs to individual countries in sequential batches, which may have created additional pockets of unused VINs. 

So what to do next? There are several methods that we previously identified to address the problems above, which have now been integrated into the model. In the registered VIN model, the tracker can distribute the impact of sudden spikes of VINs over a longer time horizon. In the reported VINs model, the tracker has increased sensitivity to exclude gaps in the dataset. There is sound reasoning behind these changes, but making changes this significant introduces a new level of uncertainty going forward. 

It’s always difficult to know when a project like this has run its natural course. When the tracker was launched, Tesla’s production was a deep mystery that frustrated investors and customers. In early 2018, Bloomberg correctly projected that Tesla was severely underperforming the guidance it had given investors. In mid-2018, the tracker correctly anticipated the Model 3’s sharp turnaround to become America’s best-selling luxury vehicle.  

As production levels of the Model 3 stabilize, interest naturally turns instead to the total number of deliveries, as that is the point when a company recognizes revenue. Scale has made deliveries easier to estimate, without resorting to estimates that rely on VINs. It’s also possible that as the Model 3 spreads out globally, this VIN-based forecast simply won’t be able to keep up. 

The Model 3 Tracker will remain a work in progress for now. After five quarters of producing cars, however, it may have passed its peak usefulness.

To contact the editor responsible for this story: Aaron Rutkoff at arutkoff@bloomberg.net

©2019 Bloomberg L.P.