mkaresh

Not all surveys are conducted the same way. Yet many people instantly assume that they're all conducted badly.

This survey was designed from the start to avoid the potential bias assumed by a couple of people here. People don't simply sign up, instantly report their past experience, and that's it. Conduct a survey this way, and responses will include a disproportionate number of complaints.

With this survey, people sign up then report each quarter going forward whether or not they've had a repair. Emails are sent with direct links to the survey to maximize the response rate. Most responses come from people who've responded quarter after quarter. Assuming that they could not predict the future when they signed up, this approach yields similar results to real-world random samples (which tend to have much lower response rates).

This method has been very effective. For a surprising number of cars (if not out-of-warranty BMWs) nearly all owners responding report no repairs. This would not happen if people only responded when they had something to complain about.

The #1 limitation, by far, hasn't been the research process, but the number of people involved. With enough participants this survey would yield information of a very high quality.

Which is why I keep trying to get more owners involved.

Join TrueDelta

FormerE30Owner

Thank you for that explanation. Its seems that my analysis of your statistics is not quite accurate.

However, if a truly random sample can be obtained, it isn't necessary that the sample size be large.

Have you determined the minimum sample size for your data to be usefull? Is that why the 2005 year is not included?

salut, Bob P.

mkaresh

It's expensive to get a truly random sample, as you must buy the owner data. And then you'll be lucky if 20% of the people contacted respond, so non-response bias remains a real possibility.

We remove an asterisk for "small sample size" once we have data for 25 cars, but this is borderline. At 25 we still have some results that don't seem right. Double that, and this hardly ever happens.

It's also a matter of what you're trying to measure, and how precisely. We're focusing on the overall repair frequency, and precisely enough that meaningful differences should be detectable. With a sample size of 25 we're not accurate within 10 repair trips per 100 cars, but I wouldn't expect someone to buy one car over another to avoid one-tenth of a repair trip per year.

Another way to think of it: if you need data on more than 25 cars to measure something, it's not likely to be a meaningful difference when you're buying a single car.

The 2005 isn't in the list because we didn't receive enough responses for it. Looks like we're two short for even an asterisked stat. We have plenty of cars signed up--55--but the older cars get the lower the response rate tends to be. Many owners of older cars seem to think that no one is interested in the reliability of cars over 10 years old. this clearly isn't true, but is a common perception, with consequences.

mkaresh

So, how have these cars been doing lately?

Our reliability stats for the 5-Series now include owner experiences through March 31, 2015. (Others are nearly a year behind.)

In terms of repair trips per 100 cars per year ("low" is best):

2010 5-Series: 77, high
2009 5-Series: 84, high, small sample size
2008 528 / 550 / M5: 96, high
2008 535: 75, moderate
2007 5-Series: 103, high
2006 5-Series: 110, high
2004 5-Series: 87, moderate

Plus, the percentage of cars with no repairs in the past year and those that required 3+ trips to the repair shop:

2008 528 / 550 / M5: 25, 7
2006 5-Series: 37, 15

No stat for a model year? Then we very much need more participants for it. Same for those with small sample size notes. Next updates in August and November.

For repair descriptions, the stats for other cars, and to sign up to help improve this information:

BMW 5-Series reliability

mkaresh

Our reliability stats for the 5-Series now cover the year ending June 30, 2015. (Others are over a year behind, and so report how these cars were doing when a year younger.)

Repair trips per 100 cars:

2010 5-Series: 66, moderate
2009 5-Series: 53, moderate, small sample size
2008 528 / 550 / M5: 85, moderate
2008 535: 95, high
2007 5-Series: 98, high
2006 5-Series: 95, moderate
2005 5-Series: 101, moderate, very small sample size
2004 5-Series: 87, moderate, small sample size

If a model year's sample size is small, then we especially need more owners involved.

For the descriptions of all reported repairs, results for other models, and to sign up to help improve this information:

BMW 5-Series reliability ratings and comparisons

mkaresh

We've updated our reliability stats for the 5-Series to include owner experiences through September 30, 2015.

Repair frequencies, in terms of repair trips per 100 cars per year:

2008 528 / 550 / M5: 96, high
2008 535: 92, high
2007 5-Series: 93, high
2006 5-Series: 65, moderate
2005 5-Series: 95, moderate, very small sample size
2004 5-Series: 91, moderate, small sample size

We have two additional statistics, "Nada-odds" and "Lemon-odds", to indicate the percentage of cars with no repairs in the past year and those that required 3+ trips to the repair shop:

2008 528 / 550 / M5: 35, 4

We'll have further updates in February and in May. Additional participants are needed to provide precise stats for all model years.

For the details, including repair descriptions, and to sign up to help improve this information:

BMW 5-Series reliability ratings and comparisons

mkaresh

We've updated our reliability stats for the 5-Series to include owner experiences through the end of 2015, about eight months ahead of other sources. In terms of repair trips per 100 cars per year--lower is better:

2010 5-Series: 68, moderate
2009 5-Series: 55, moderate, very small sample size
2008 528 / 550 / M5: 90, high
2008 535: 110, high, small sample size
2007 5-Series: 82, moderate
2006 5-Series: 68, moderate
2005 5-Series: 75, moderate, very small sample size
2004 5-Series: 108, moderate, small sample size

We have two additional statistics, "Nada-odds" and "Lemon-odds", to indicate the percentage of cars with no repairs in the past year and those that required 3+ trips to the repair shop:

2008 528 / 550 / M5: 39, 7
2006 5-Series: 47, 4

We'll have further updates in May and in August. The more owners participate, the more comprehensive and precise these will be.

For the repairs behind these stats, reliability information on other models, and to sign up to help improve this information:

BMW 5-Series reliability ratings and comparisons

mkaresh

We've updated our reliability stats for the 5-Series to include owner experiences through June 30, 2016. Reliability information elsewhere is based on a survey conducted in April 2015--well over a year ago.

Repair frequencies, in terms of repair trips per 100 cars per year (lower is better):

2010 5-Series: 65, moderate
2009 5-Series: 36, low, very small sample size
2008 528 / 550 / M5: 136, high
2008 535: 87, moderate, small sample size
2007 5-Series: 96, high
2006 5-Series: 78, moderate
2005 5-Series: 48, low, very small sample size
2004 5-Series: 119, high, small sample size

We'll have further updates in November and next February. The more owners participate, the more comprehensive and precise these will be.

For the repairs behind these stats, reliability information on other models, and to sign up to help improve this information:

BMW 5-Series reliability ratings and comparisons