Reliability by brand cars 1 to 4 years old

[Diddums]

I wouldn’t read too much into it from an electric car situation, more the build quality of Tesla - weren’t they building Model 3s in the factory car park at one point (or was that an urban myth?)

No, you're correct. They used massive tent things which they erected in their car park.

https://arstechnica.com/cars/2018/0...emporary-assembly-tent-as-factory-2-0-awaits/


I also take these lists with a massive pinch of salt, far too many variables to get reliable stats.

 

[chaparral]

At lest I know now that I should be telling people to buy a Hyundai or Vauxhall or a fiat as there more reliable then a BMW, Audi or a Merc

 

[MichaelAwkward]

At lest I know now that I should be telling people to buy a Hyundai or Vauxhall or a fiat as there more reliable then a BMW, Audi or a Merc

Did you previously recommend German brands to people on the basis of reliability? It's been this way for decades. Even Mercedes famously reliable period pre-1990s wasn't actually that reliable in reality.

 

[MartinPrince]

A guy on my street was one of the first I saw with a Tesla; I haven't seen it for a little while now which might explain things.

 

[chaparral]

Did you previously recommend German brands to people on the basis of reliability? It's been this way for decades. Even Mercedes famously reliable period pre-1990s wasn't actually that reliable in reality.

No I normally keep away from them German brands due to them normally having high running costs (high parts & labour costs)

But even the cheaper brands parts & labour costs seem to be getting very high these days when going through the main dealers..

 

[PCM2]

That's not how statistics work though? If there's only a 1000 Tesla owners in the UK compared to a 100,000 Skoda owners, only 28 Tesla responses is perfectly valid if only 2800 Skoda owners responded also. It might not be 100% reflective of reality, but it doesn't invalidate the results so long as the overall sample is normally distributed.

In fact it's likely that much less than 0.3% of owners of other cars responded to this survey, because people just don't respond in general, to any survey.

Actually, that's exactly how statistics work. You can sugar-coat it any way you want. It's not a statistically significant sample. The margin or error in responses when you've only got 18 samples (28 samples - my mistake) is massive. No scientific study would get away with such a small sample size, nor would it be accepted in the first place for any sort of credible publication. It simply isn't a statistically relevant sample size to use.

 

[chaparral]

Actually, that's exactly how statistics work. You can sugar-coat it any way you want. It's not a statistically significant sample. The margin or error in responses when you've only got 18 samples is massive. No scientific study would get away with such a small sample size, nor would it be accepted in the first place for any sort of credible publication. It simply isn't a statistically relevant sample size to use.

The other thing that can really effect this is that you normally only hear from the people that had faults with something

 

[Kenai]

Actually, that's exactly how statistics work. You can sugar-coat it any way you want. It's not a statistically significant sample. It's exactly how statistics work - the margin or error in responses when you've only got 18 samples is massive. No scientific study would get away with such a small sample size, nor would it be accepted in the first place for any sort of credible publication.

This is not only applicable to the Tesla though.

A quick Google suggests 7.5m cars were registered in 2016,17,18. A sample of 28,000 is 0.37%.

If you throw away the Tesla results, you throw away all the results, the response rate was consistent.

 

[PCM2]

The other thing that can really effect this is that you normally only hear from the people that had faults with something

Yes, exactly. And with that bias in play from these self-selected results you simply can't draw anything meaningful from 28 responses. 28 responses! It isn't just about a percentage of the overall sample (i.e. how many cars of a given type are on the road) but about having a good absolute sample size. Which isn't the case for Tesla here.

Anyhow, this has been discussed to death as this is not new news. Guess there's nothing useful to be added to it in this thread.

 

[MichaelAwkward]

Yes, exactly. And with that bias in play from these self-selected results you simply can't draw anything meaningful from 18 responses. 18 responses! It isn't just about a percentage of the overall sample (i.e. how many cars of a given type are on the road) but about having a good absolute sample size. Which isn't the case for Tesla here.

Those 18 responses (or 28?) aren't the sample size though. They're just a subset. The sample was 18,000. Any sample over 30 responses tends to be normally distributed as per CLT. If the question was 'How reliable are Teslas up to 4 years old?', you'd have a point - but the question was how reliable are cars of all different brands up to 4 years old. That Teslas only made up 18 or 28 of those responses doesn't invalidate the survey.

 

[MichaelAwkward]

The other thing that can really effect this is that you normally only hear from the people that had faults with something

Doesn't matter in this case if it applies to all brands. It'd only be a problem if you only heard from people with faults for one brand and not another, which is unlikely.

 

[PCM2]

Those 18 responses (or 28?) aren't the sample size though. They're just a subset. The sample was 28,000. Any sample over 30 responses tends to be normally distributed as per CLT. If the question was 'How reliable are Teslas up to 4 years old?', you'd have a point - but the question was how reliable are cars of all different brands up to 4 years old. That Teslas only made up 18 or 28 of those responses doesn't invalidate the survey.

I see what you're saying, but it doesn't make this survey at all convincing. You also have to consider what exactly they're measuring - "reliability" is a very broad term. The Tesla Model S has common issues with their fancy electronic door handles for example. Doesn't apply to the Model 3 and a rear door handle not working as normal isn't the same as a car being completely unusable!

 

[MichaelAwkward]

I see what you're saying, but it doesn't make this survey at all convincing. You also have to consider what exactly they're measuring - "reliability" is a very broad term. Model S have common issues with their fancy electronic door handles for example. Doesn't apply to the Model 3 and a rear door handle not working as normal isn't the same as a car being completely unusable!

Depends on their methodology, but if there was no weighting in play, it wouldn't affect the results.

i.e. you might have 50 total Tesla owners respond, 28 of them own Model S, 18 had problems with the doors. Therefore overall they're 64% reliable. Then add in 22 Model 3 owners, 11 of whom had problems with... steering or whatever. Overall reliability is now 42%, that is 42% of Tesla owners who responded had no problems.

It'd make sense if they weighted the results according to the time the car was then with the dealer or a 1-5 rating of how inconvenienced the owner was or something, but without seeing their methodology I've no idea. Personally I think a rear door handle not working is just as bad as a car being unusable. It's a pretty fundamental thing, being able to open doors. I think any issue that requires a trip back to the dealer is significant.

 

[PCM2]

Personally I think a rear door handle not working is just as bad as a car being unusable. It's a pretty fundamental thing, being able to open doors. I think any issue that requires a trip back to the dealer is significant.

Not if you rarely have rear passengers or can get around the issue by simply opening the door on the other side instead. Or opening the affected door from inside (after opening another door). One outside door handle not "presenting itself" would be the issue I'm referring to here. It certainly is annoying, but doesn't stop you using the car nor even the door itself.

I don't know exactly what counts as a reliability issue on this survey either, but these sorts of statistics just annoy me in general. If you read more into them and ask questions about what the data it actually telling you, it becomes clear it's quite uninformative really.

 

[chaparral]

Personally I think a rear door handle not working is just as bad as a car being unusable. It's a pretty fundamental thing, being able to open doors. I think any issue that requires a trip back to the dealer is significant.

Also if I remember right it meant to cost around £1000 to fix there door handle problem

 

[Nasher]

Wow what happened to Honda and Nissan. They have seemingly slipped right down to German and French levels...

Absolutely! But it’s a funny old world and I don’t think the German reputation for engineering prowess/efficiency and what not is one that’s reflected in reality.

Definitely not. Not with cars anyway.

 

[Terminal_Boy]

Interesting that Dacia (60%-70% Renault built by Romanians) is considered more reliable than Renault (100% Renault built by the French).

 

[MichaelAwkward]

Interesting that Dacia (60%-70% Renault built by Romanians) is considered more reliable than Renault (100% Renault built by the French).

Probably because their cars are, on average, simpler designs than Renault even with component/platform sharing.

 

[Freakbro]

As well as Honda and Nissan, surprised to see Volvo low down there too, I always thought they had the reputation for reliability

 

[chaparral]

Am surprised big time by seeing Subaru so high up as everyone I seem to known or seen that owned a scooby had there head under the bonnet most weekends fixing it
(fair enough there cars were far older then 4 years old)