By on March 14, 2014

Warranty claims paid

From time to time someone comes to me with a great idea: instead of surveying car owners to get TrueDelta’s reliability stats, why not use warranty claims data? The reason why not: manufacturers consider such data to be highly proprietary. So when I heard that the auto industry’s “first OEM warranty and recall study” was going to be presented at a Society of Automotive Analysts meeting, I was intrigued. Had someone gotten their hands on this data? What were they able to learn from it?

Warranty claims comparison

It turns out that Stout Risius Ross, the financial consulting firm that conducted the study, didn’t have access to any proprietary data. Instead, they used recall filings and the manufacturer-level data available in financial statements. Recall filings include the number of cars affected. Annual costs of warranty claims must be disclosed in financial reports. These data tend to be messy, as different companies include different things in these costs; there’s no precise universal standard. Also, multiple model years are often lumped together, and changes in both warranty claims and recalls lag changes in the number of cars sold. Keep these limitations in mind when viewing the “warranty claims as a percentage of revenue” comparison.

Recall trends

The presenters downplayed the relationship between recalls and quality, as the latter includes things gone right as well as things gone wrong. So what’s the point of studying all of this warranty and recall data? The usual point for a business audience: money. Warranty claims and recalls cost manufacturers billions of dollars each year. Finance needs to be able to predict these costs so they can set aside appropriately sized contingency funds.

Engine vs. non recalls

Of even greater interest to the people in the room: who pays. OEMs want suppliers to pick up half or more of the cost of warranty claims and recalls. So far suppliers have been picking up only a small fraction of these costs, but OEMs have been getting more aggressive. Suppliers are smaller, sometimes much smaller companies. A big recall could bankrupt them—if they had to pay for it, and they were not insured.

Claims percent revenue

To help them not pay for it, they can hire good lawyers. One was on the panel. To the many suppliers in the room he suggested “if you’re not at the table, you’re part of the meal.” To avoid getting stuck with the costs of recalls and warranty claims, suppliers must first push to be included in any discussions the OEMs have regarding who pays. Otherwise they’ll just receive an invoice.

Despite the availability of excellent legal assistance, a supplier might end up having to pay out millions for a recall or warranty claims anyway. To protect themselves, they can buy insurance. The price of this insurance will depend on calculations of risk. Currently, no one outside the OEMs has good data on the costs of recalls and warranty claims. To fill this need, we have this study.

Two trends of greater interest to car owners did come up in the presentation. The first is fairly well known: warranty costs have been declining as cars’ reliability (a narrower term than quality) improves. The second trend is much less well-known: since the mid-1990s the number of cars recalled each year has been surging. The presenters weren’t able to get too far into the reasons for this. The addition of airbags has been one driver. When cars didn’t have them, they couldn’t be recalled. In recent years airbags have been one of the most frequently recalled systems. Engine-related recalls have grown much less quickly than non-engine-related ones. Also, the NHTSA has gotten more aggressive. The lawyer on the panel offered protection from big government—it’s important to push back.

A third driver of the increasing number of cars recalled wasn’t part of the presentation. To increase economies of scale, OEMs have been striving to communize as many parts as possible across multiple products. As a result, when a critical part does fail, a much larger number of cars can be affected. The number of cars recalled has grown faster than the number of recalls.

This is just the initial phase of a planned annual study. Perhaps they’ll be able to get more detailed data in the future? Unfortunately, this doesn’t seem likely. The people in the room (OEMs, suppliers, business consultants) generally agreed that the companies should not have to disclose any more detail; model-level information should be proprietary, and the requirements suck up an excessive number of person-hours already.

Get the latest TTAC e-Newsletter!

71 Comments on “Guest Post: Society Of Automotive Analysts Reliability Study...”


  • avatar

    > To increase economies of scale, OEMs have been striving to communize as many parts as possible across multiple products. As a result, when a critical part does fail, a much larger number of cars can be affected.

    This doesn’t make statistical sense. The overall failures rate doesn’t really increase since the more common part is just as likely to not fail (ie zero recalled) as each of less common ones.

    If anything, because it’s mostly less common cars using more common parts, aggregate reliability should improve.

    • 0 avatar
      Pch101

      It makes perfect sense.

      Reliability is essentially a failure rate, i.e. a percentage that a given part breaks or malfunctions.

      Recalls are undertaken to repair or replace items that could jeopardize safety in the event that they fail. In the vast majority of cases, there were few actual failures and the customers would have not known that there was a problem prior to the recall notice. The recall is undertaken because of the possible consequences of failure, not necessarily because there have been a high rate of failures.

      Efficient parts sharing should lead to larger individual recalls, due to the proliferation of the parts. Again, most of the parts that are recalled gave no indication to the vehicle owners that they needed to be replaced.

      • 0 avatar

        I didn’t think of it myself, but this is absolutely the case. Recalls aren’t based on the percentage of cars affected, but on the number of people likely to be affected, and how severely. With commonization even a very low chance of any one person getting harmed can lead to a recall.

      • 0 avatar

        > The recall is undertaken because of the possible consequences of failure, not necessarily because there have been a high rate of failures.

        This is a purely mathematical argument where “failure” and “potential to fail” are entirely fungible. Iow, it makes no difference in the equation if 1% of parts fail, or 1% of parts are determined might be a safety hazard.

        Michael should know the math and acknowledges the case below.

        • 0 avatar
          Big Al from Oz

          There are better ways to describe ‘mathematical argument’;

          1. risk assess, and

          2. probability.

          I will start to correct your grammar. You seem to be limited in the use of the English language.

          • 0 avatar

            > Do you understand risk assessing? Not what you can google, but what is taught

            Oh God no, Big Al thinks he knows math too. What have we done to deserve this.

        • 0 avatar
          Big Al from Oz

          @agenthex
          Do you understand risk assessing? Not what you can google, but what is taught.

          Do you understand what elimination is concerning risk assessing?

          Do you understand what acceptable risk is?

          How does this tie into engineering?

          How do you define a hazard?

          You’d better go to college and/or uni to learn.

          Judging by your use of language you only made it to the 10th grade.

        • 0 avatar
          Pch101

          “‘failure’ and ‘potential to fail’ are entirely fungible.”

          Except that they’re not. Reliability is a layman’s term for “doesn’t break much.” It’s an implied percentage.

          Recalls often target items that have never or have rarely broken, or that have otherwise provided any indication that they would break. Hence, the notion of a reliable car being recalled is not at all contradictory.

          • 0 avatar

            > Except that they’re not. Reliability is a layman’s term for “doesn’t break much.” It’s an implied percentage.

            The simple stat equations themselves are agnostic to particulars, and a percentile is in fact by definition unitless.

            The percentage here simply means % likelihood of recall. For example, any given ignition cylinder model (delphi, ford, etc) has x% chance of being recalled for something.

          • 0 avatar
            Pch101

            “The percentage here simply means % likelihood of recall.”

            The point made above was that the size of an individual recall will be determined by the degree of parts or design proliferation, and isn’t a reliability measure per se.

            TMC has large recalls because it is very good at parts sharing; a bad part will end up in a large number of vehicles. GM has more but smaller recalls — more unique parts to fail, but fewer of them in use at any given time.

          • 0 avatar

            > The point made above was that the size of an individual recall will be determined driven by the degree of parts or design proliferation, and isn’t a reliability measure per se.

            Srsly? I explain the frankly trivial math here:

            http://www.thetruthaboutcars.com/2014/03/guest-post-society-of-automotive-analysts-reliability-study/#comment-2955082

            > TMC has large recalls because it is very good at parts sharing; a bad part will end up in a large number of vehicles. GM has more but smaller recalls — more unique parts to fail, but fewer of them in use at any given time.

            Yes, but the *aggregate effect* of both operation models is mathematically identical. All these *statistical events* (whether recall or failure or Big Al grasping the point) are completely independent, thus incredibly straightforward to calculate or intuit.

          • 0 avatar
            Power6

            Is that verifiably true or you are just making stuff up? Just casual observation and personal knowledge doesn’t seem to me TMC is using any more or less parts aharing than any other auto maker. GM was prolific at this traditionally. We are in the midst of a massive ignition switch recall no?

            TMC was sharing plenty of parts back when they didn’t have ANY recalls ever.

          • 0 avatar
            Pch101

            “Just casual observation and personal knowledge doesn’t seem to me TMC is using any more or less parts aharing than any other auto maker.”

            You could just look the data. Recalls are a matter of public record. You can see how many recalls that there were, what the recalls entailed and how many vehicles are affected. The numbers speak for themselves.

          • 0 avatar
            Pch101

            “Yes, but the *aggregate effect* of both operation models is mathematically identical.”

            They aren’t, as I explained above.

          • 0 avatar

            > They aren’t, as I explained above.

            If that’s the case can you point to the flaw in my math so we have something quantitative to go on?

            I understand they’re “different” just as the probability of Big Al replying on-point is “different”, but we can use the exact same stat concept to calculate the average expected probably of Big Al replying on point with one big post vs many small ones. Given same likelihood of for each post, and the fact that any arbitrary distribution of small posts will sum to 1 (unity chance of the assumed one big post), the result will always be the same.

            Some of the likelihoods will not be identical, but generally in the case of recalls they will favor more popular better tested part models.

          • 0 avatar
            Power6

            “You could just look the data. Recalls are a matter of public record. You can see how many recalls that there were, what the recalls entailed and how many vehicles are affected. The numbers speak for themselves.”

            That does not indicate anything at all. If there is no design flaw there is no recall nothing to do with parts sharing. I guess that answers my question though you have no clue just making stuff up. I’m sure tons of throttle pedals are shared in GM cars, but we wouldn’t know since they haven’t had a recall on those. We do know they put more motor mounts into 5 years worth of cars in the 60s than Toyota did pedals in 6 years of cars more recently.

          • 0 avatar
            Pch101

            I’ve explained it above. I’m not sure how else to explain it.

            Reliability is a breakage rate (more or less). In contrast, recalls are often determined by the severity of the worst-case scenario, with little regard for the likelihood of the worse-case scenario, i.e. the breakage rate.

            In the case of the GM ignition recall, it is motivated by the fact that it killed people, not by the failure rate of the ignition. The failure rate itself was low, but the consequences were anything but, hence the drama.

            If the ignition failures weren’t related to something else of importance (in this case, airbag deployment), then I doubt that you’d hear many complaints about the ignitions being unreliable. It’s the body count that creates the problem.

          • 0 avatar

            > They aren’t, as I explained above.

            It’s also possible my claim that this is abstract math which has nothing to do failure or anything isn’t clear.

            Consider for example 2+2=4. Two and two of what? It doesn’t really matter. It could be failures or recalls or Big Al posts. I probably shouldn’t have referred to it as “failures”, but it’s only a euphemism.

            The only thing that matters here is that recalls in the multi-part model are independent events (ie not causally related), and the total parts in each sum up to the same number (1mil toyota switches = 1mil total gm switches).

          • 0 avatar
            Power6

            pch I guess you got on to something else there are a couple threads there we were talking about your assertion TMC shares more parts than other makers. Not sure what you are on about. Damn its too confusing here with the replies.

          • 0 avatar

            > Reliability is a breakage rate (more or less). In contrast, recalls are often determined by the severity of the worst-case scenario, with little regard for the likelihood of the worse-case scenario, i.e. the breakage rate.

            Again to be crystal clear, it’s just unfortunate I happened to use “% failure” which brought this on.

            These are just naked ratios in the math. So forget I ever said that and consider the math reasoning again. If you don’t believe me, just google stats equations and rest assured they are not just for failure rates.

          • 0 avatar

            > pch I guess you got on to something else there are a couple threads there we were talking about your assertion TMC shares more parts than other makers. Not sure what you are on about.

            PCH is probably correct on the greater variety of parts for GM if the recalls show smaller recall numbers. This is a simple numerical argument. More parts sharing will imply greater recall sizes.

            However when we add statistics it shows that the aggregate impact is the same in the long run.

          • 0 avatar
            Pch101

            Again, I don’t understand the lack of comprehension here.

            Recalls are often initiated when little or nothing has gone wrong. But when the potentially negative outcome poses a safety issue, then there will be (or at least should be) a recall, even if there was no problem that was apparent to the drivers or the owners.

            In the case of the GM issue, it makes no difference at all that 99+% of those owners experienced no problem. In that sense, the part was “reliable.” But that “reliability” doesn’t change the need for a recall.

          • 0 avatar

            > Again, I don’t understand the lack of comprehension here.

            Unfortunately you’re pulling a Big Al here. I *completely and totally* agree with everything in your posts. Totally and completely agree that recalls are nothing like part failures in their specifics. But that’s not the point.

            It’s more like:

            A: Two apples and two apples make 4 apples
            P: But these are oranges
            A: Ok, yeah, but the math is the same
            P: But you’re comparing apples and oranges
            A: Technically yes, but the units in 2+2=4 are agnostic
            P: I’ve explained it above. I’m not sure how else to explain it. Apples ≠ Oranges
            A: The 2′s and 4′s don’t have units so you can substitute anything you like. The math doesn’t care if if it’s apples or oranges or failures or recalls (ok, technically, fruits are not independent events so the stats does care).
            P: Again, I don’t understand the lack of comprehension here.

            and so on.

          • 0 avatar
            Pch101

            “The math doesn’t care if if it’s apples or oranges or failures or recalls”

            The math is impacted because a recall will remove all of the parts from the system, while an unreliable part that is not subject to recall will be allowed to remain in use.

            For example, an “unreliable” part may have a real-world failure rate of 20%, and only those 20% will be replaced as needed. In contrast, a recalled part may have a failure rate of well below 1%, yet 100% of them are replaced because the cost of that failure is unacceptably high, i.e. dead people.

            A campaign that seeks to purge all parts from the system, failed or not, is not comparable to hoping for the best and replacing some of the parts ad hoc. The math for these things is not at all the same.

          • 0 avatar

            > The math for these things is not at all the same.

            I’ve been trying to avoid it because it’s tedious, but let’s examine this from first principles to demonstrate one way or the other:

            A chance of a probabilistic event is by definition the ratio of it happening. For a quarter coming up heads, it’s 0.5. The expected cost of a probabilistic event is this * cost. For a game where I give you the quarter if it comes up heads, it’s 0.5 * 0.25 = 0.125. This is our 2+2, note it doesn’t involved car parts of any sort. It’s also how insurance and whatnot is calculated so I expect someone with a finance background can grasp it.

            One way of using this involving parts is probabilistic event of failure. 1% chance of failure on a $100 part expects $1 cost. 1% chance of failure on 10k of these ($1mil) expects $10k cost. (this is a setup for later)

            Another way of using this math involving parts is probabilistic event of a recall of a collection of them we “part models”. 1% chance of recall on *part model* that will cost $1mil to recall is $10k expected. This is the specific case here, but note the multiplication process is the same.

            We’re not done yet because the goal is to prove equivalent costs. In order to do this we can expand “cost” here into unit costs and #units. Cost here is by definition unitCost * #units. Note how this was already done for our first way above, and we can similarly do this for recalls as a $1mil recall divides into #cars recalled and cost per car.

            This is the point we’re at the math reply to Mike: http://www.thetruthaboutcars.com/2014/03/guest-post-society-of-automotive-analysts-reliability-study/#comment-2955082

            Assume our large common part recall is that $1mil over 10k cars or $100 per car. Assume our non-shared paradigm uses 1000 part models, therefore recalls are 1000 times smaller so 10 cars for each part model with same $100 per car or $1k recall cost. Since to be fair to both paradigms we assumed same probability/susceptibility of a part model to recall (ie 1%), and obviously same 10k cars total: (10k cars * 100/car * 1%) = 10k expected cost, or sum of 1000 models of (10 cars * 100/car * 1%) = 1000* $10 = 10k expected cost.

            Note in reality all these numeric assumptions of parity across paradigms favored the non-sharing. More common parts are less likely to be recalled, and per units cost of a large recall is likely to be smaller than many small recalls (I also point this out in the linked post). Therefore we should only expect the large recall model to cost less.

            QED.

          • 0 avatar

            ^ some of the math at the end was unclear, so I rewrote the last part.

          • 0 avatar
            Pch101

            The issue that was raised in the article was the number of vehicles impacted by any one given individual recall. The point made was that parts sharing makes for bigger individual recalls when they do occur.

            That’s a logical argument, and you haven’t refuted it. If you’re trying to claim that those individual recalls may not necessarily lead to a higher aggregate number of vehicle recalls, then I won’t disagree, but that wasn’t the point being raised in the article

          • 0 avatar

            > The issue that was raised in the article was the number of vehicles impacted by any one given individual recall. The point made was that parts sharing makes for bigger individual recalls when they do occur.

            No, look up the sentence which motivated this claim:

            > “A third driver of the **increasing number of cars recalled** wasn’t part of the presentation. To increase economies of scale, …”

            The math is intended to show this isn’t true, and generally speak the opposite of correct.

          • 0 avatar
            Pch101

            From the article:

            “To increase economies of scale, OEMs have been striving to communize as many parts as possible across multiple products. As a result, when a critical part does fail, a much larger number of cars can be affected.”

            Also note the next comment: “The number of cars recalled has grown faster than the number of recalls.”

            All of that makes sense. I’ve run out of ways of explaining how utterly logical that is.

          • 0 avatar

            > “The number of cars recalled has grown faster than the number of recalls.”

            This implies that part sharing is becoming more common; this point of yours is correct. However as we can see from the math, part sharing doesn’t lead to “increasing number of cars recalled” as Mike suggested, it only makes each recall larger (which is more than balanced out by their lesser frequency).

            The question is then why has the frequency not dropped (assuming total = frequency * units), and this is likely due to the other factors discussed in the article. I suspect it’s due to higher safety standards in general. Back in the day we’re willing to live with a few extraneous deaths here and there, now everyone freaks out over 10 in a pool of a million. IOW, that percentile in the math is moving with respect to time.

          • 0 avatar
            Pch101

            “However as we can see from the math, part sharing doesn’t lead to ‘increasing number of cars recalled’ as Mike suggested”

            He didn’t say that. What he said was that each individual recall that does occur involves more cars.

            You’re misintrepreting his point, then arguing based upon your misinterpretation. The statistical issue here isn’t the likelihood of recalls taking place, but the likely average size of individual recalls when they do occur.

          • 0 avatar

            > He didn’t say that. What he said was that each individual recall that does incur involves more cars.

            Is that how you interpret this?:

            > “A third driver of the **increasing number of cars recalled** wasn’t part of the presentation.”

            Note the first and second driver both also concern “since the mid-1990s the *number of cars recalled each year* has been surging”, as the third driver logically should.

            So either his math is wrong (my take) or the entire premise of his reasoning is wrong (your take).

          • 0 avatar
            Pch101

            He also pointed out that the total number of vehicles recalled has been increasing since the mid-90s, which he attributed largely to airbags. In other words, the increase in the quantity of safety equipment produces more items that are likely to be subject to government-mandated (or inspired) safety recalls.

            But that is a somewhat different point from the impact of parts sharing on individual recall size.

          • 0 avatar

            No, you simply read the article wrong. *All 3 drivers* are motivations for the second *trend* noted:

            > The second trend is much less well-known: since the mid-1990s the number of cars recalled each year has been surging.

            Driver 1:

            > The addition of airbags has been one driver.

            Driver 2:

            > Also, the NHTSA has gotten more aggressive.

            Driver 3:

            > A third driver of the increasing number of cars recalled wasn’t part of the presentation. To increase economies of scale,

            He simply made a bad math argument (vs a bad structural argument) by assuming larger recalls leads to more total recalled cars without considering dependency on frequency.

          • 0 avatar

            Christ, he noted it himself in the reply just below:

            > I get what you’re saying. If x% of parts have a fault that results in a recall, then if there are fewer parts in total then recalls should be less common, and the net result should be a wash.

          • 0 avatar
            Pch101

            I had an easy time understanding the article, and the points within it are sound.

            -There are more recalls than in prior decades, for several reasons.

            -Individual recalls involve more vehicles per recall, due to parts sharing

            To anyone who follows this stuff, this isn’t controversial.

            I’ve pointed out on various threads that Obama’s NHTSA is more aggressive than Bush 43′s NHTSA. This article is consistent with that.

            Since recalls involve safety, the increased use of safety equipment produces more items that could be potentially be recalled. Also not a surprise to anyone who has followed the evolution of vehicle design.

            And parts sharing necessarily leads to larger individual recalls, since the bad parts are proliferated more widely.

            This is an example of what he is talking about: http://www.autoblog.com/2013/04/11/toyota-honda-nissan-mazda-recall-3-4-million-for-faulty-airbags/ The Takata airbag recall didn’t just impact those 3.4 million vehicles made by TMC, HMC, Mazda and Nissan, but it also was later determined that it hit 220,000 BMWs and some GM cars.

            It crossed international boundaries and impacted completely separate automakers. This degree of parts sharing is unprecedented, but the need for scale economies drives it. (Automakers would prefer to outsource the airbag to a specialist rather than develop and build the entire thing in house.)

          • 0 avatar

            > I had an easy time understanding the article, and the points within it are sound.

            Sure, I guess that’s why Mike himself agrees the third driver for the second trend wasn’t sound.

            > And parts sharing necessarily leads to larger individual recalls, since the bad parts are proliferated more widely.

            Otherwise it’s an irrelevant point since it doesn’t actually cost more (esp if insured against). Eg. for injury protection if you can reduce the risk (ie frequency) 10 fold, but the cost of each incident doubles or even 5x, you still come out ahead.

            Everything else is fine, though this is worth mentioning:

            > I’ve pointed out on various threads that Obama’s NHTSA is more aggressive than Bush 43′s NHTSA. This article is consistent with that.

            This is the % in the simple equations above. Move the threshold of a recall and likelihood increase (eg 1% to 1.5%).

          • 0 avatar
            Pch101

            “Otherwise it’s an irrelevant point since it doesn’t actually cost more”

            It’s relevant because large recalls create bad PR. A lot of people confuse recalls with reliability, and assume that mega-recalls are indicative of cars becoming less reliable.

            That misinterpretation of the facts can lead to lower sales volumes and lower revenues, particularly for those who promote themselves based upon quality.

            I’ve already covered how reliability and recalls aren’t the same. The article also notes that reliability has been improving over time.

          • 0 avatar

            > It’s relevant because large recalls create bad PR. A lot of people confuse recalls with reliability, and assume that mega-recalls are indicative of cars becoming less reliable.

            Anything can be made into bad PR. Apparently ~10 deaths can magically transform into 300 overnight with the right spin. Toyota specifically was responsible for approx 0 deaths from UI defects and look where that got them.

            Given this is evidently a completely tangential process from recall costs & other realities which can be logically reasoned about, there’s not much of any point to ponder it as if it were.

      • 0 avatar

        Btw, pch it turns you were right, this guy is even worse than ruggles.

        • 0 avatar
          Big Al from Oz

          @agenthex
          So, you don’t understand how to correlate data and trend it?

          So what is the definition delta lieu, what is it used for? How do you determine and model failure?

          What do you know the difference between the run in phase to the failure phase of a mechanical or electrical component?

          What is spectro analysis? How is it used?

          • 0 avatar

            > So, you don’t understand how to correlate data and trend it?

            I wonder how clueless someone to be to believe this has anything to do with what’s going on here. I’d ask Farnsworth for his clue-o-meter, but unfortunately that one stops at zero.

          • 0 avatar
            GeneralMalaise

            Hex is simply using Obamath… a trillion here, a trillion there, pretty soon we’re talking real numbers…

          • 0 avatar

            > Hex is simply using Obamath… a trillion here, a trillion there, pretty soon we’re talking real numbers…

            So these people apparently use a different math, like 2+2=22. No wonder it’s always wrong.

        • 0 avatar
          Big Al from Oz

          @agenthex
          You ‘opened’ you mouth at the wrong time to the wrong person.

          Like I stated you think you are clever than all on this site.

          You a nothing, but a troll.

        • 0 avatar
          Big Al from Oz

          @agenthex
          At least I have quals. I’m not some charlatan with a tenth grade education.

        • 0 avatar
          GeneralMalaise

          This guy doesn’t get it.

    • 0 avatar

      Thanks for not pointing out that the word should be “commonize.” I’ll see if Derek can fix that…

      I get what you’re saying. If x% of parts have a fault that results in a recall, then if there are fewer parts in total then recalls should be less common, and the net result should be a wash. Unless you also assume that fewer parts means that each part can be more thoroughly tested, in which case commonization should reduce the number of cars recalled.

      But I suspect there’s another variable in play that keeps the number of parts recalled at least even. Couple this with commonization, and you get a larger number of cars recalled.

      Then there’s a factor I didn’t think of until after writing this article (and which the presenters don’t seem to have considered, either). Another new factor since the early 1990s (when recalls began to increase) is the Internet. Car owners who notice a problem can now easily report it to the NHTSA. They can also share experiences of problems in forums. When NHTSA receives enough complaints about a particular problem, they ask the manufacturer to look into it. Many recalls begin this way.

      • 0 avatar

        > If x% of parts have a fault that results in a recall, then if there are fewer parts in total then recalls should be less common, and the net result should be a wash.

        That’s a creative way to look at it, but the more straightforward explanation is that assuming equal cost & prob of failure for parts the expected cost for 1 part X is (cost * prob * countX), whereas expected cost for arbitrary number of parts is sum of (cost * prob * countOfEach). “countOfEach” obviously sums to countX here.

        The prob and cost of a shared part is likely lower than prob and cost of each. The only other major diff would be higher variability because the 1 recall would be larger than some number of smaller ones. This may affect insurance if the pool cannot be spread around.

        • 0 avatar
          Big Al from Oz

          @agenthex
          A simple phrase would have described much of what you posted.

          “Increased probability of failure due to an increase in risk, caused by increased complexity/technology and component numbers”.

          Not hard.

          • 0 avatar

            > A simple phrase would have described much of what you posted.

            Toyota has kill more people in UI incidents than the number of thoughts Big Al has ever grasped.

          • 0 avatar
            Big Al from Oz

            @agenthex
            Boy, you are quite a piece.

            Always with the personal asides.

            That makes me feel better. If you can only attempt to belittle others then you really don’t have much.

          • 0 avatar

            > Always with the personal asides.

            It’s always funniest when those being mocked can’t grasp that it’s funny because it’s true.

            I better go correlate some data and trend it, LOL.

          • 0 avatar
            GeneralMalaise

            The guy amuses himself, Big Al. Probably better than self-abuse… no?

  • avatar
    mfgreen40

    Are cars recalled only for safety related reasons?

    • 0 avatar
      Pch101

      Yes.

      • 0 avatar

        Not entirely the case. The NHTSA will compel (or threaten to compel) recalls only for safety reasons. But manufacturers can conduct a “campaign” (they avoid “recall”) for any reason they want to. Some entirely voluntary campaigns are conducted for “customer satisfaction.”

        There are also different levels of campaigns:

        1. Send letters telling people to get the car fixed immediately, and follow-up aggressively.

        2. Send letters telling people to get the car fixed.

        3. Send letters telling people that if a part fails, they’re covered up to X miles.

        4. Inform dealers to check X and fix it as needed if and when the car is brought in for service.

        5. Extend the warranty on x part, with no other notification.

        And so on…

        • 0 avatar
          Pch101

          “The NHTSA will compel (or threaten to compel) recalls only for safety reasons.”

          That’s what I was referring. ODI doesn’t care about reliability per se, but only about safety. (This is yet another reason that people shouldn’t confuse recalls with reliability.)

          I don’t believe that those voluntary repair campaigns that aren’t safety-related are considered to be “recalls.” They may come in the form of TSB’s, but TSBs aren’t recalls.

          But I should have noted that EPA can issue its own recalls for emissions equipment, so those aren’t necessarily safety-related, either.

  • avatar
    Dimwit

    While interesting it’s just mostly conjecture. The data has to be far more granular to be of any use. In each recall, is the problem design, engineering, production, Q.C. or other? Is each recall the same. I would submit no. They would have to be split between definite identifiable problem and political interference. The Toyota run on debacle high on everyone’s mind, I’m sure.

    The problem with gathering data like this is the results. It’s probably better that a lot of this doesn’t get examined. It’d be a real bitch if the government got sued for anticompetitive behaviour for using recalls as a whip for some companies.

  • avatar
    jim brewer

    There is pretty decent public data for Ford because of their ESP factory warranty. It gets a paragraph or two in the 10-K filings.

    They offer four levels of coverage from powertrain only, to the full monty, with varying years and mileages up to 7 yrs and 125K. Moreover, since they allow internet discounting of the product, an enterprising person with a knack for statistics could figure out where the trends, the costs and the trouble areas are pretty rapidly.

    P.S. I like that “If you aren’t at the table you are the meal” line. I’ll have to remember it.

  • avatar

    If you figure it is all odds….They make each part as inexpensively as possible. Some times they get away with it…some times they don’t. I once had a SAAB recalled because the temperature knob broke off. It broke off because the one half cent piece of plastic behind it that wouldn’t pass muster at the Testors Model factory would break and jam. Someone cheaped out….a touch too much.

    While not mission critical, there are endless examples of this in all car makers. Coils go. Torque Converters don’t (thanks Acura !) In cold weather, mirrors don’t work. Cat Converters don’t light off. Recalls are justified, and I’d rather they fix it.

    The best you will do are parts good for about 150k. A car maker only cares about the warranty period…or that which the first owner cares about. Second owner ? They replace the valve drive belt. (in Saab’s case, they eliminated a 50 cent cut in the floor pan, and made you drop the gas tank to change a fuel pump, with two hours of labor.)

    • 0 avatar
      thornmark

      I remember when Iacocca was lamenting the state of the auto industry, poor car sales et al. He was complaining that cars lasted TOO LONG, or at least certain parts.

      His solution was to make sure parts didn’t last too long – i.e., exhaust systems should not last 8 years when other parts failed much earlier. Take quality (&cost) out. At least w/ most of his cheezy designs – not the Cordoba – he was doing US a favor.

      THe Japanese had a different solution which was maybe more inefficient, mandating part replacements that made keeping an auto uneconomic, at least in comparison to replacing the unit. In a land of savers, that forced consumption and kept the plants humming.

      • 0 avatar
        Big Al from Oz

        Robert Lutz took over Chrysler and found out that Lee’s ideal’s were putting the company under.

        I bought a 95 XJ Cherokee Sport and found out Chrysler parts were designed for a 6 000 mile life cycle.

  • avatar
    SCE to AUX

    Michael – glad to see your post here.

    The data in their report is annoying because some of it isn’t normalized to account for time, such as the number of recalls and their costs. These numbers should be presented as a fraction of the cars on US roads, not as a raw number.

  • avatar
    VoGo

    Interesting article. Good to see Michael back.

    My takeaway: VW’s claims as a % of revenue are 5X Honda’s. Five times. Wow.

    • 0 avatar

      As noted in the article, there are no set standards for what to include in warranty accruals, so these figures are loosely comparably at best.

      • 0 avatar
        CapVandal

        Great post. I was surprised at the the total dollar figures. Ouch.

        If VW is even 2x Honda on an apples to apples basis — That’s a big difference.

        It’s nice to see some facts, even if the data is more aggregated than we would like.

  • avatar
    TrailerTrash

    Perhaps there is another way…Technical Service Bulletin

    TSBs…is there a way to collect and list all reported and published TSBs released or sent out by manufacturers?
    So…although these are not life threatening, they do show a repair or issue important enough that a TSB had to be sent out.

    Are these available to use as data?


Back to TopLeave a Reply

You must be logged in to post a comment.

Subscribe without commenting

Recent Comments

New Car Research

Get a Free Dealer Quote

Staff

  • Contributing Writers

  • Jack Baruth, United States
  • Brendan McAleer, Canada
  • Marcelo De Vasconcellos, Brazil
  • Vojta Dobes, Czech Republic
  • Matthias Gasnier, Australia
  • W. Christian 'Mental' Ward, Abu Dhabi
  • Mark Stevenson, Canada
  • Cameron Aubernon, United States
  • J Emerson, United States