Why two-thirds of early Tesla Model S motors will need replacing by 60,000 miles(Read article summary)
Early Tesla Model S motors have had reliability issues. Now, a new analysis of data provided to Plug-In America by 327 owners of early Tesla Model S cars suggests that as many as two-thirds of those early Model S drivetrains will need to be replaced within 60,000 miles.
With almost 100,000 on the world's roads, the Tesla Model S electric car is a remarkable achievement.
It remains the longest-range electric car in volume production more than three years after it launched.
But reliability issues with electric traction motors in early cars--those from the 2012 and 2013 model years--have dogged the earliest owners.
Now, a new analysis of data provided to Plug-In America by 327 owners of early Tesla Model S cars suggests that as many as two-thirds of those early Model S drivetrains will need to be replaced within 60,000 miles.
[EDITOR'S NOTE: The title of this article originally said that two-thirds of the drivetrains would "fail" by 60,000 miles. As several readers pointed out, that's not quite accurate; Tesla's policy is to replace them when they begin to make noise, without waiting for them actually to fail mechanically. To make the title more accurate, we reworded it to say they will "need replacement," to match the sentence above more closely.]
This analysis has not been publicly disclosed before now. Before publishing this story, Green Car Reports asked several specific questions of Tesla Motors to help put the analysis in context.
Tesla declined to answer those questions. Instead, it issued general statements about its reliability. Both the questions and its statement are at the end of the article.
Reliability: really a liability?
Since Consumer Reports dropped the Tesla Model S from its list of recommended buys due to a "worse-than-average overall problem rate", the barbs between the company's supporters and detractors have only sharpened.
At issue is the significance, if any, of Consumer Reports' findings. Given that Mercedes-Benz--another highly successful luxury brand--also gets a worse-than-average rating, does it really matter? And should one also consider TrueDelta's more pessimistic assessment?
Do a drive experience rated at "103 out of 100" points and 97-percent customer satisfaction mean Tesla has little to worry about? Or do lengthening waiting times at service centers foreshadow a fall in customer loyalty and brand prestige?
Without the proper context, it all reduces to an electric-vehicle version of the fable about the blind men and the elephant.
But there's a trove of statistics that can contribute to the conversation. And a Weibull analysis of that data suggests that two-thirds of early (2012 and 2013) Model S cars can expect a drivetrain failure within 60,000 miles.
Clearly, Tesla's eight-year warranty coverage on the drivetrain protects new and used buyers. And the company has said several times that it has made great leaps in quality as it gains experience in building the car.
Tesla's November update said the company has cut its failure rates by half, while CEO Elon Musk has expressed strong confidence in the improved drivetrains it has been shipping in recent months.
But the data set used to analyze drivetrain reliability includes about 10 times as many early (2012 and 2013) cars as recent ones from the 2014 and 2015 model years.
If enough recent buyers add their information, we may be able to improve the analysis and get statistical backup for the trend of improved reliability Tesla has mentioned.
Plug-In America's big (enough) data
In the survey, respondents provided a variety of information on their Model S ownership experience, including total miles driven, whether they've had a motor swap (drivetrain replacement), and if so, what the odometer reading was at the time.
In October, when this analysis was run, the data set had 370 respondents. Not having any reliability software handy, I asked a reliability engineer to crunch the numbers--and received the startling reply that the "characteristic life" of the drive train was about 50,000 miles.
In a reliability context, the so-called characteristic life is the age at which 63.2 percent of parts are expected to fail. This represents a survival rate of 36.8 percent, or 0.368. That value is the inverse of e, the base of the natural logarithm. In math terms, 1/e = 1/2.718 = 0.368.
The reliability engineer was careful to emphasize that the results were only valid if the data was correct, had no selection bias, and was random. While selection bias was definitely in play (only those customers who knew about the survey could choose to fill it out) the sample covered more than 1 percent of the total population of 2012 and 2013 Teslas (327 respondents for about 25,000 vehicles).
So selection bias and randomness were probably acceptably low. But on a line-by-line review of the Excel data file, I noticed a few transcription errors which indeed compromised the result.
When cleaning the data set, to give Tesla the benefit of the doubt--where respondents had reported a motor swap but didn't specify when it happened--I used the final odometer reading. (This would push up the characteristic life slightly.)
I repeated the analysis using free reliability software found online (I used Reliasoft, and there are other packages). For 2012 and 2013 Teslas, the characteristic life rose to about 57,000 miles. (The probability function predicted that 66 percent of drive trains would fail within 60,000 miles, giving the title of this article.)
By and large, calculations are only as credible as they are transparent, so the raw data and calculation method have been outlined at www.tinyurl.com/TeslaWeibull, along with the author's contact information.
This should be enough for readers to duplicate the drive train analysis, and assess the Model S battery pack and onboard charger as well, the latter of which apparently has a characteristic life of about 1,000,000 miles.
Only 43 people in the data set owned 2014 and 2015 Teslas. And while this verges on a big enough sample--a statistical rule of thumb is that 30 to 50 random samples allow you to make accurate inferences about large populations--only a handful had even reached 20,000 miles.
That means that an analysis on that data would have been invalid. Far from measuring Tesla's success at mitigating the "wear-out" failure mechanism(s), it would reflect quirks in infant mortality and/or random mid-life failures.
There should now be enough 2014 Model Ses with enough road miles that we can refine our drivetrain reliability trend analysis.
That will move us from discussing the yes/no question of whether or not the Model S is reliable to using crowdsourced data to evaluate its past reliability rate, and the pace of improvement in newer vehicles.
The calculations above are generally called a Weibull analysis, after the eponymous probability distribution. Its utility comes from the fact that it measures survival rates over time, which means inferences can be drawn from parts which haven't yet failed (called "suspends" or "suspended tests").
This is very important for long-lived parts, where it can take an impractically long time to run all samples to failure. (Calculations on early Teslas consisted of 77 failures and 250 suspends.)
In short, the Weibull analysis docks you points for items that have failed, and gives you credit for items that haven't.
Weibull distributions are power functions, meaning that they can be highly curved when plotted on a linear scale. For this reason, they're often plotted on a logarithmic scale, where they form nice straight lines.
Both linear and logarithmic charts have been included here. Failed tests are shown in blue, and suspended tests in red (along the bottom axis). In the bottom left, three values are given. Beta is the shape parameter.
Values above 1.0 represent failure rates that increase over time, indicating a component that wears out over time. Eta is the characteristic life noted above, at which point 63.2% of components are expected to fail. Finally, Rho is the correlation coefficient, which should ideally be close to 1.
No Tesla answers to questions
Before publishing this article, Green Car Reports reached out to Tesla with a list of specific questions to help us put this analysis in context.
Specifically, we asked:
- How many motors has Tesla repaired or replaced in 2012 and 2013 Model S carsto date?
- What percentage of total cars does that represent?
- What does the company's own reliability data indicate about the percentage of cars that will require motor replacements over the life of the vehicles?
- Musk said in November that reliability had doubled; how confident is Tesla that it has now fully resolved the reliability issues with earlier motors?
- What is the company's projection for failure rate in the motor(s) of current production cars?
- What would Tesla Motors say to owners (and future used-car buyers) of the early cars regarding the reliability of the motors?
Tesla declined to answer any of these specific questions. A company spokesperson sent the following statement:
Close communication with our customers enables Tesla to receive input, proactively address issues, and quickly fix problems. Over-the-air software updates allow Tesla to diagnose and fix most bugs without the need to come in for service. In instances when hardware needs to be fixed, we strive to make it painless.
Despite the problems, our data show that Tesla owner satisfaction is still very high: Ninety-seven percent of owners said they would definitely buy their car again. It appears that Tesla has been responsive to replacing faulty motors, differentials, brakes, and infotainment systems, all with a minimum of fuss to owners.
And Tesla’s attention to customer service has been effective. Almost every survey respondent made note of Tesla’s rapid response and repair time, despite the lack of a traditional dealer service network. For its early adopters, Tesla has made a practice of overdelivering on service problems under the factory warranty.