First we monitored factory output. Now we’re taking a broader look at Elon Musk’s make-or-break electric car.
The release of Tesla Inc.’s most important electric car, the Model 3, raised an existential question for the company. Could it figure out how to crank up production before the money ran out? To find the answer, Bloomberg built a data-tracking tool that proved uncannily accurate at estimating output during that critical first year. This is the story of how we did that, why we’re stopping, and what we’re doing next.
The Model 3 Tracker launched in February 2018 after Tesla created something no other automaker could claim: an electric car that hundreds of thousands of people lined up to buy. The only problem, at least in the beginning, was that Tesla couldn’t produce enough of them. Elon Musk had “bet the company”
on mass production—and he very nearly blew it
Tesla reported its production totals only on a quarterly basis, leaving investors to speculate in the dark as the company missed self-imposed targets. Into this fray, Bloomberg’s tracker generated real-time forecasts of the number of Model 3s coming off the lines at the factory in Fremont, California. We did this by combining data from an obscure government registry with reports from thousands of our readers.
The results far exceeded expectations. For four consecutive quarters, Bloomberg’s estimates came closer than Wall Street forecasts. The Model 3 Tracker twice came within 0.5 percent of Tesla’s actual production. (We blew it
on the fifth quarter; more on that below.)
Now it’s time to move on to tracking the next phase in Tesla’s evolution. As the company proved its ability to crank out Model 3s, investors have turned to different metrics of success. Tesla shares have been in a tailspin in recent weeks on doubts about consumer demand for the Model 3—and electric vehicles writ large. Tesla needs to prove it can mass produce cars with the quality, reliability, innovation and service to keep attracting new customers as traditional automakers add more electric options.
This is why Bloomberg’s Model 3 Tracker will stop producing new forecasts. Instead, the data we’ve collected will help launch an extensive survey of Tesla customers. Model 3 owners can participate using the link below.
The Tesla Model 3 SurveyView the survey results
To kick off the survey, we’ve reached out to the customers who contributed to the original tracker. We’re also opening the survey to other Model 3 owners around the world. We want to know about initial customer experiences—the good and the bad—and to follow how cars made by Tesla hold up over time.
While we’ll no longer be modeling overall production, we’ll continue to update the two datasets that Tesla watchers have grown to rely on: Vehicle Identification Numbers (VINs) that Tesla registers with safety regulators prior to production and VINs that new owners submit to Bloomberg.Tesla Registered VINs Tracker(Last updated 12/31/19)
Tesla’s quarterly vehicle totals VINs registered with U.S. regulators9/30/170100200300400
Registration activity sharply diverged from production in Q1 2019
Note: VINs aren’t produced in perfect sequence. Reported VINs always exceed actual production. Sources: NHTSA, Tesla
Reported VINs Tracker(Last updated 12/31/19)
Tesla’s quarterly vehicle totals VINs from social media North American VINs reported to Bloomberg International VINs reported to Bloomberg
Submit your VIN9/30/170100200300400
Note: VINs aren’t produced in perfect sequence. Reported VINs always exceed actual production. Sources: Reddit, Tesla Motors Club, Bloomberg
A Look Back at How the Tracker WorkedBloomberg’s production tracker relied on the two sets of Vehicle Identification Numbers (VINs) presented in the charts above. A VIN is a unique string of letters and numbers assigned to every vehicle manufactured in the U.S. First, we gathered batches of VINs that Tesla registers with safety regulators prior to production (Registered VINs Model). Then we developed our own dataset of VINs sent to us by owners after they took delivery of the cars (Reported VINs Model).
The first dataset was very forward looking. It served as a rough gauge of how much production Tesla anticipated in the near future. The second dataset was backward looking, as Bloomberg doesn’t receive those VINs until after cars are shipped and delivered to new customers. Each dataset individually performed poorly at estimating Tesla’s progress; taken together, however, it worked surprisingly well.
The combined model was updated every day. Just below the headline, readers could see our latest estimates for cumulative production and weekly output. Here’s what our primary charts looked like before we retired the forecasting tool:
Total Model 3 Production Snapshot
(Model retired on 5/28/19)
Tesla’s reportingBloomberg estimates
Production total last confirmed at 218,637 for end of Q1
Weekly Model 3 Production Rate Snapshot
(Model retired on 5/28/19)
Tesla’s reportingBloomberg estimates Raw weekly data
Note: We use a 13-week trailing average to assess the weekly production rate. The raw weekly data are subject to temporary fluctuations and don't represent our best estimate of current output. Source: Bloomberg
The tracker was extremely accurate during the North American rollout. But Tesla’s expansion overseas, starting in 2019, presented challenges that our method could not easily overcome. Bloomberg is among thousands of websites that are restricted in mainland China, the second biggest market for Tesla, which meant very few VINs contributed by Chinese owners.
The international rollout also inflated both datasets as Tesla registered entirely new sequences of VINs bound for overseas markets. This led the model to overestimate first-quarter production in 2019 by 26 percent—a huge miss for the forecasting tool. The problem was evident in the weeks before Tesla reported results, which we relayed to readers in the project’s integrated blog.
Sources: Tesla, Bloomberg
Note: Chart updated to measure Bloomberg estimates against production rather than production against estimates.
We used the tool’s blog to publish more than 100 posts, which are archived here
. The idea was to help interpret results of this data journalism experiment and describe changes made to the model over time. This became even more important after Wall Street analysts, such as those at Sanford C. Bernstein and UBS, began using the tracker results to advise clients. In total, the Model 3 Tracker has been viewed more than 5 million times since its launch. The project helped break news and featured prominently in more than a dozen Bloomberg News stories.
In the early days of 2018, the Model 3 Tracker proved invaluable at forecasting that Tesla was falling well short of Musk’s guidance long before official results were posted. By mid-2018, the tracker also detailed the remarkable turnaround that led the Model 3 to rank among the best-selling sedans in the U.S.Tesla's Weekly Production Rate vs Tesla’s Targets
Tesla’s reportingBloomberg estimate Tesla's 1-week burst rate targets Tesla's targets for sustained production
Tesla’s targets have become more conservative over time
Sources: Tesla, Bloomberg
Now that Tesla’s production is stabilizing, the importance of month-to-month production tracking has diminished. The Model 3, however, is no less important—for the future Tesla and for the adoption of electric vehicles in general. Thank you to the thousands of readers who contributed VINs to our database. Now we’re asking Tesla owners to help us build the next phase of our data-driven inquiry into the world’s biggest electric automaker.
Methodology: Car companies register new batches of VINs before previous batches are depleted, so we’ve adjusted Method 1 accordingly to match Tesla’s quarterly reported production numbers. We’ve also made adjustments for Method 2 to reflect that vehicle deliveries don’t occur in perfect sequence. We’ve removed 400 VINs from the total database to account for Tesla’s test cars. Our final production estimate is an average of the two methods.