The Reliability Bell Curve: What Does “More Reliability” Actually Mean?

In a recent post about up front spending, quite a few people mentioned their anecdotes about buying a cheap washing machine and having it last for many years – and then used that as a justification to ignore reliability data when making a purchase and instead go for the cheap item.

Let’s explore that idea a bit. Take a look at this picture:

graph 1

This represents the reliability of a hypothetical low-end washing machine over time. It has an average lifespan of ten years – and most of those models fail at around the ten year mark.

But there are exceptions to that general trend, of course. Some of them fail quite quickly – and this, being a cheap machine, doesn’t have good warranty support. Others last for a long time, even up to twenty years or more.

When you buy a machine, it’s going to wind up at some point on this curve. Most likely, it’ll be one of the ones in that big hump in the middle, lasting ten years or so like the average machine. However, there’s a chance that it might junk out in three years, or it might last you for twenty years.

Now, what about a more reliable machine? If you chose to spend a bit more and buy a more reliable machine, you’d get a curve that looks more like this:

graph 2

It comes with a good warranty, so there are no failures within the first five years. Overall, the average failure comes in at about the fourteen year mark – and some will last for twenty five years or more.

Obviously, when you look at Consumer Reports or other such research, this second washing machine would have a higher reliability grade than the first one. People (like me) are willing to pay a bit more to get a reliability curve that looks like this one than like the first one.

Let’s overlap them.

graph 3

Notice the three colored spaces? They each tell a different story, and their relative sizes are really important.

The blue area shows how likely it is that the cheap machine will fail before the expensive machine does.

The pink area shows something very similar: how likely it is that the expensive machine will last longer than the cheap machine.

To put it simply, you can combine the blue and pink areas – combined, they show the likelihood that the more expensive and reliable machine will outlast the cheaper and less reliable machine. There’s about a 70% chance of that.

Of course, there’s that orange area. That represents the chance that the cheap machine will actually last longer than the expensive machine. In this example, there’s roughly a 30% chance of that happening.

So what’s the story here? Paying more for reliability means that the orange area is your worst case scenario; buying the cheap machine means that the orange area is your best case scenario.

Let’s put it in terms of comments about low-end washing machines. Yes, you’ll hear occasionally from an individual that bought a $200 machine in 1987 that’s still running today. That person lucked out – they bought a cheap machine, but it happened to wind up in that orange area.

However, most of the time, the machine will be a blue one. It will fail before a higher-reliability machine ever will.

Similarly, someone might complain about the “supposedly highly reliable” machine they bought in 2001 that’s already failed. Again, it’s an orange case. Most of the time, that machine will be a pink one.

Here’s the real truth: the exceptional cases mean very little. If you hear about one or two cases, pure chance might give you a few tales straight out of the orange area – exceptional cases, whether or not it’s a good exceptional case or a bad one.

What actually matters is a lot of cases combined together. When you get a lot of cases together, you can get a real picture of the reliability of the machine. The whole curve fills out, with the poor exceptional machines, the great exceptional machines, and the average machines.

Whenever you go to make a major purchase, you’re placing a bet on reliability. You don’t know exactly how it’s going to turn out. Good reliability data simply means that the machine is likely to be more reliable than a machine with poor reliability data.

Given that a more reliable machine means less time investment (you don’t have to deal with time lost to a broken machine, nor do you have to deal with repairmen) and less money investment (the cost of repairs plus the cost of having to buy a replacement sooner than expected), paying extra for reliability is a bet I’m willing to take. I’m quite willing to pay a 50% premium to buy the machine with the pink curve than the one with the blue curve.

If you enjoyed reading this, sign up for free updates!

Loading Disqus Comments ...
Loading Facebook Comments ...

81 thoughts on “The Reliability Bell Curve: What Does “More Reliability” Actually Mean?

  1. Andi says:

    Three comments:

    1. A $200 machine in 1987 was probably not a cheap machine.

    2. Expensive and reliable are not synonymous.

    3. When comparing machines, are you paying for reliability or 30 extra wash cycles and gizmos?

    Bottom line, it pays to do your research. In our experience, we’ve not only read up in Consumer Reports but also visited with the local repairman – what types of machines does he do a lot of work on and which ones are hard to get parts for.

  2. Amy says:

    @Andi Good thought on visiting the repairman! We think about it for cars, why not other every day items?

  3. Jimbo says:

    Time value of money man, do you not get it??? The extra money you spend upfront which gives its rewards to you over a period of time involves opportunity costs – you could have invested the money and got a better return through an investment than spending an extra $500 on a washing machine whose payout is over a 10+ year horizon. Cannot believe you are repeatedly so myopic.

  4. My Journey says:

    Trent,

    I think there is something to be said about rolling the dice to hope you end up in the upper region of a cheap product. This obviously depends on the cost and risk, but it is something I’d love to see you comment on.

  5. Baker @ ManVsDebt says:

    I really enjoy the amount of specific depth in this post. My only issue is that you summarize that you would pay a 50% increase. According to your example, the washing machine lasts 15 years compared to 10, so that’s 50% more reliability.

    I think most people would be willing to pay 50% more for 50% more reliability given that they will save time. I think a much more interesting question is would you be willing to spend 60% more in this same example?

    What about 65%… 70%… Where do you draw the line on these kind of things. Obviously the time/stress saved is worth X. X is just different for each person.

    Either way, I really enjoyed the post! I encourage you to do posts like this more often.

  6. Jonathan says:

    I’m confused about the bell curve – are you saying that after an appliance is ten years old, it is increasingly LESS likely to fail each year that goes by? That is what the second half of the curve implies.

  7. C.J. Holmes says:

    I just recently purchased a dryer from Lowe’s for about $350. This was one of the cheapest models offered and would probably fall under the “potential for” 10 year reliablility curve. I would have had to double if not triple that price to get a dryer that would fall into that “potential for” 20 year reliability. So, to gamble the reliability being 10 or 20 years, I would rather spend the lower amount and find out that I bought a lemon, than to spend twice that amount and find out I bought a lemon. Those who can afford the gamble, I’m sure will be fine, but for those who are not financially able, I would suggest the lower priced dryer, if buying new.

  8. George says:

    His “chance of failure” line should really read “number of failed units”.

  9. C.J. Holmes says:

    I just recently purchased a dryer from Lowe’s for about $350. This was
    one of the cheapest models offered and would probably fall under the
    “potential for” 10 year reliability curve. I would have had to double
    if not triple that price to get a dryer that would fall into that
    “potential for” 20 year reliability. So, to gamble the reliability
    being 10 or 20 years, I would rather spend the lower amount and find out
    that I bought a lemon, than to spend twice that amount and find out I
    bought a lemon. Those who can afford the gamble, I’m sure will be fine,
    but for those who are not financially able, I would suggest the lower
    priced dryer, if buying new.

  10. Jonathan Vaudreuil says:

    @Jonathan (love the name) – the bell curve is showing when the washing machines will stop working, so as time goes on there are less machines left which can break, since more than half have already broken. It is not the likelihood of breaking, it is the number of cases where a machine breaks.

    I’d like to second getting a machine fixed. My roommates and I decided to try and fix our dryer last year. I went online to a forum, posted the problem, got feedback, contacted a part supplier, he agreed I had a good chance of fixing it with a new part, and we bought the part and fixed it. It saved us money ($75 part vs $400 for a cheap dryer) and I learned about fixing a dryer.

  11. Jonathan says:

    @george that makes sense.

  12. Michael says:

    Like Jimbo said, this post is a detailed response to the strawman. Our real concern was present value.

  13. Joe says:

    Plus this doesn’t even include the added efficiency that the more reliable machine will more likely give you.

  14. Trent says:

    “I think there is something to be said about rolling the dice to hope you end up in the upper region of a cheap product. This obviously depends on the cost and risk, but it is something I’d love to see you comment on.”

    I’m going to get to that. This post is somewhat setting the foundation for that discussion. I originally had one very long post, but decided it really needed to be split into two or three pieces.

  15. lurker carl says:

    Graphing chance of failure as a bell curve is incorrect. As the machines age, the failure chance approaches 100%. At some point, practically every machine will have failed, that chance of failure does not drop back down to zero over time. Perhaps labeling the graphs as percentage of machines failing over time would be more accurate.

    Consumer Reports often states the best performing and longest lasting products are seldom the high end or pricier models. Often the less expensive, less complex machines perform best and last longest. Also, energy efficient washing machines may use less energy per hour but must run longer cycle times to get clothes clean. If an ordinary machine runs for 30 minutes and the efficient machine runs for 50 minutes, using 20% less energy per minute will not use less energy per load. All things must be considered when evaluating these products.

  16. BJD says:

    The math is off somewhere …’I’m quite willing to pay a 50% premium ‘ but in the other post he say ‘I’ll buy the $600 one and leave the $300 one at the store.’

    That’s a 100% premium; the more reliable machine is twice the price of the other machine. If the more reliable machine was at a 50% premim ($450) I think there would have been much less discussion because that $450 machine is clearly the better deal over the life of the machine.

  17. I agree with Jimbo. If you buy up front, you lose the ability to invest what you didn’t spend up front. For example, suppose I was choosing between a $500 machine and a $1000 machine. The $500 machine supposedly lasts 10 years and the $1000 machine lasts 20 years. If I buy the $500 machine, I can invest the other $500 I would have spent for the next ten years. In that time, I would make (at 6% interest, which is what my rewards checking account pays me now) $395.42 in interest. At that time, I would pay another $500 for a new machine. Those two machines would last me a combined total of 20 years. Also, for the second ten years, I would still make interest on the $395.42 I had earned the first ten years. This would bring in about an additional $300. On the other hand, if I bought the $1000 machine up front, I would not be able to invest anything. Therefore, by purchasing two cheaper machines, I am about $700 better off. I certainly won’t be paying up front.

  18. Jerry Tsai says:

    Hi Trent,

    I follow your blog and enjoy your posts. While the intent of your post and the general conclusions you draw on the whole make sense, your choice of a normal distribution to depict reliability is not a good idea. “Time to event” is not typically modeled this way by statisticians and do not reflect actual experience.

    Typically, statisticians use distributions like exponential or Weibull distributions to illustrate what you depicted. In those distributions, probability of failure is typically strung out much differently than a normal distribution.

    Also, you might want to instead to depict on the y-axis the cumulative risk of failure. That way as time elapses the aggregate probability of failure increases and “makes sense” to the viewer.

  19. Pierre says:

    @IndependantBeginnings

    You’ll have also to calculate that a 500$ machine will probably more in the future due to inflation. If the replacement cost is 850$ in 10 years, then you’ll have lost the interest you made on your first 500$…

  20. Indymoney says:

    Trent, I really like the way you research on things. My question is, If you are debt free and you have some money on hand, why do you need to go and buy the machine with premiums? (You can save some dollars if you pay full cash and also you’ll get some good deals)

  21. Chris says:

    I installed 4 CFL “green” lightbulbs that are supposed to last for 5 years.. A month later 2 (50%) have burned out…

    I feel burned..

  22. Brent says:

    Yeah, but if you are going to be that anal retentive about making an extra 50 bucks off the interest you have to factor in all the extra down time, delivery costs, gas money to get the thing to there. Factor in inflaction and figure that that 50 bucks has less buying power in the future. Factor in that the $500 dollar machine now costs $550. Oh well.

  23. GlennH says:

    The failure analysis of a device follows what is known as the “Bathtub Curve.” Failures usually follow this pattern: if a device is going to fail, it will fail right away or within the first few weeks of use, called burn-in time. Then, there is a long period of stability, followed by an increasing failure of components due to wear. Reference: http://www.weibull.com/hotwire/issue21/hottopics21.htm

    One issue about consumer items is that the “guts” stay the same over several different models. The higher-priced model only differs in appearance and maybe a more complicated (and more failure-prone) controller. To use an example from the automotive industry, the Chevy, Buick, Oldsmobile, Pontiac, and Cadillac all used essentially the same engine with different trim and option packages. Appliances are the same way.
    I have repaired our washer and dryer numerous times over the past 14 years, since they were bought new. They are the Maytag brand, and I was quite happy with them. Our family runs them at an extreme rate, so I’m not surprised. The washer has needed inlet valves, belts, and hoses. The dryer wears out ignitors (it’s a gas dryer), valve coils, glides, and drive belts. Parts are available from Marcone Appliance Parts. They offer a lifetime replacement warranty on repair parts that has come in handy over the years. I’ve saved hundreds on just eliminating the service call fees alone. My background: Mechanical Engineer, working in the aerospace industry.

  24. Noko says:

    Trent, your choice of graphs here is bewildering. Graphing as a bell curve means that you’re graphing the chance that a unit _will fail_ in a particular year, as opposed to _will have failed_ by a particular year, which renders the entire enterprise frustratingly unintuitive. If you had chosen to graph the chance of a machine having failed by a given date, and thus ended up with two lines, each approaching 100% as time goes on, it would be much simpler to digest.

  25. Matt says:

    Something else to consider is that a washing machine failure(or other appliance failure) can have other effects than simply requiring replacement. It could cause a water leak or start a fire, not simply stop working.

  26. jc says:

    I have to strongly echo Carl’s comment. There is NO WAY that a bell curve even remotely models failure of a device. This may seem harmless to you, but Taleb’s “Black Swan” shows how (among other things) the current financial/economic crisis is the result of inappropriately using bell curves willy-nilly.

    I understand you want to make this comprehensible for a non-technical audience, but using the wrong curve re-inforces problematic statistical stereotypes.

    carl is right that your curves might better represent % of machines that have failed, but they’re still dramatically the wrong shape, even hypothetically.

  27. kitty says:

    @Jimbo, Independent Beginnings – you are ignoring inflation. If you bought your first machine for $500, it is not at all clear that after it fails you’ll be able to buy a second one for $500. In the meantime there is a good chance the prices have risen and you may have to spend $1000 for a cheap machine.
    On the other hand, with some items the prices go down e.g. most electronic toys, but I don’t believe this is the case with washing machines.

    Also, it is not necessarily the case of 10 years vs 20 years, it could be the case of 5 years vs 20 years depending on how lucky you are, so you may need to replace your cheap machine several times.

    I agree with Lurker Carl that expensive doesn’t always correlate with quality, so it pays to do research and determine the best combination of price and quality. There is also a hard ceiling on price at how much you can afford.

  28. Dave says:

    re comment 12, trent is graphing a probability distribution function, you are referring to a cumulative distribution function. EG, this graph shows how many machines fail at 10 years, how many fail at 11 etc, not the total number of failures from 1-10 or 1-11, this is the CDF, which is the integral of the curve shown above. in reality a bell curve is a gross oversimplification of the problem, and even given that, the curve shows ZERO failures in the first 5 years for the reliable machine, which is essentially impossible. but this isn’t a graduate statistics course, so i think we can let all this slide. trent is just trying to make a point about statistical outliers and that you can’t count on buying one at sears.

  29. lanhsin says:

    Your analysis also leaves out the factor of future technological improvements. A modern washer is more energy efficient than one from a decade ago. Depending on how much you believe the technology for a given appliance will change, you may not want to invest in the most reliable model if it happens to also be the most expensive because the “lifespan” of the appliance may determined by obsolescence rather than function.

  30. @kitty
    I agree that if the difference is 5 years vs. 20 years, it may be a different story. I am just saying that it is not always better to choose higher reliability with a higher price. Sure, each case should be considered individually.

    Also, if you do consider inflation, in my example it would still make more sense to buy cheaper. Between 1998 and 2008, inflation caused something that cost $500 in 1998 to cost $657.60 in 2008. This price is higher by $157.60. However, I would have made $395.42 in interest during that time, so I still would have made more in interest than I would have paid in inflation (I used an inflation calculator at http://www.westegg.com/inflation/infl.cgi to arrive at the inflation number).

  31. Heidi says:

    In other words, anecdotal evidence is not statistically significant.

  32. Sarah says:

    @ Heidi. Your comment made me laugh.

  33. JW says:

    I understand and agree with the point of your post Trent, but as a statistician I have to ask where you found that your random variable is normally distributed? That may be true but is a very strong assumption to be made. There are many tests to determine normality and it is not something anyone should just assume!

    I am certainly not saying that I know what the distribution should be, but if you read it is normal somewhere I would be interested to see the article/paper. Are you trying to use a sampling distribution of some kind?

    And Dave (#19) – what you are talking about certainly isnt GRADUATE level stats. More like 300 level stats :) – just had to say it heh. We did more measure theory in grad school. I have to diagree with you about letting poor math and bad statistics “just slide”. I expect anyone who chooses to use statistical terms to at least do so correctly.
    (of course this is why I can’t watch the news!)

  34. Sharon says:

    To build on Dave’s comment (or simplify it) – the Y axis is number of washing machines. The curve is highest at the middle, showing that most washing machines fail around the average failure time (e.g., for the low-end washer, 10 years). The curve is lowest at ~ 1 and 18 years, showing that very few low-end washers last for such short or long periods of time, respectively. Out of all possible washers (which are all – theoretically – plotted on the graph, according to their lifespans), you buy only one. The one you buy falls somewhere along the curve – but you don’t know where along the X axis until it actually fails.

    Another point: the graph doesn’t have to be a normal curve. It could be skewed to the right or left, or it could be flatter (indicating greater variation in the failure rates) or more narrow (indicating less variation in failure rates).

  35. Johanna says:

    Yeah, there are some issues with the exact shapes of the curves, but they’re basically right (as Dave already explained) and close enough to get the point across: Rules have exceptions, so a couple of anecdotal exceptions aren’t enough to disprove a rule.

    But what I am wondering is, where was this post back when we were discussing advantages, backgrounds, and discrimination as they relate to a person’s chances of success? Stories about individuals who have done well for themselves in spite of seemingly long odds are inspiring and instructive, but they don’t actually prove that certain people aren’t at a disadvantage due to race, sex, background, or other characteristics. It seems to me that that was a concept some people were having trouble with at the time.

  36. Kyle says:

    The post has the right idea, but I think the conclusion is somewhat wrong, especially ironically when you consider the post on confirmation bias recently.

    You *should* be considering reliability vs. price. But it’s not always a given that the more expensive item will be a better deal in that sense. I think Trent has a tendency to use confirmation bias to come up with sets of numbers that favor the expensive, longer-lasting product.

  37. lurker carl says:

    The actual shape of the curve is not relevant because there isn’t any data, it’s only an artistic representation of the concept.

    My comment was to point out the Y axis on the graphs are not labeled correctly. Chance of product failure over time is not what the bell curves represent. They represent the quantity of product failing as time progresses.

  38. Chad says:

    First-time commenter; this blog has been an inspiration to me, so I don’t like to see incorrect information on it.

    Graphing chance of failure as a bell curve is probably incorrect, but for the reason in comment #17, not that in comment #12. The curve will most likely be very low at t = 20 years, for the simple reason that most appliances won’t last that long. The correct picture of “how likely is it that my washing machine will fail by time X” is the _area_ under the curve between t=0 and t=X, and that does increase as X increases.

    It’s wrong because, as jc pointed out, most distributions aren’t bell curves. I have no idea what the failure distribution for washing machines is, but many devices don’t follow this model. Many computer components, such as hard drives, have failure distributions with peaks at t = a few months and t = a couple of years, and a very low distribution in between; it’s referred to as a “bathtub curve”.

    Second, your overlap illustration doesn’t show what you want it to show. For example, your descriptions of the blue and pink regions are logically identical; they can’t both be right, and in fact they’re both wrong.

    Sadly, it’s impossible to show what you want to show on a two-dimensional graph. The lifetimes of the cheap machine and the expensive machine are independent variables; you can’t put them on the same axis. You need two dimensions for the two lifetimes, plus a third for the probability density: how likely is it that El Cheapo will fail at time T *and* that Super Washer will fail at time S? That distribution will show up as a surface, and the quantities you want to compare show up as certain volumes under that surface.

    Your broader point is correct — there is some chance that the cheaper washer will outlast the more expensive washer, but we shouldn’t pay much attention to that. Actually, it’s nowhere near as likely as your graph makes it appear. The real question, as you note, is how much effective use we expect out of the appliance for each dollar we put into it.

  39. Joseph Tanner says:

    Personally, I get all my appliances at the scratch ‘n dent store. I can get the model with all the gizmos and doohickies for less than the low-end model in a regular store.

    I agree with the above poster who said that the low-end models are often more reliable. When it comes to washers and dryers, they’re around a lot of moisture. The ones with simple knobs and buttons will generally last longer that ones with any kind of circuit board. Replacing a circuit board on one, especially if it’s an older unit, can cost more than a brand-new low-end machine.

  40. susan says:

    When we bought our expensive washer/dryer a few years back, I specifically ordered the one without the fancy electronic controls and other bells and whistles (for about $150 less). The salesman seemed shocked that I wouldn’t want all the fancy features – but the more fancy features, the more things that can break. Besides, it isn’t really THAT hard to just turn a dial to select what wash cycle you want. :)

  41. Stephan F- says:

    Jimbo, Let’s see how well that would have worked for us.

    Dow
    May 1989 2,500
    May 1999 11,000
    May 2009 8385 (a few minutes ago)

    So if you did invest say $500 in 1989 you would have made just over 3x your money. or about $1,600. Not bad. However, inflation took away about half that gain. So you’d end up at about $800 in value. But that is still a really good washing machine today.

    I think the effect of investing in 1999 is obvious, isn’t it?

    How about gold? May 1989 $235 May 1999 $269 2009 $922 That would have been great but honestly who saw that one coming?

    Heidi: Yes, the plural of anecdote is not data :) but is often an indicator of worthwhile investigation. Remember rogue waves.

  42. Kyle says:

    “My comment was to point out the Y axis on the graphs are not labeled correctly. Chance of product failure over time is not what the bell curves represent. They represent the quantity of product failing as time progresses.”

    There is a different between chance of product failure over 25 years and chance of product failure in each of the next 25 years from the present.

  43. Rob Bennett says:

    I like this blog entry a lot. Great analysis.

    One argument against paying extra for a machine that will last longer is that improved machines may come out and you will need to wait longer to move up to the machine with the improvements. Paying less up front leaves you with enhanced flexibility because you have less at risk when considering a switch.

    Rob

  44. Jim says:

    I think Lurker Carl is right. The Y axis label seems wrong.

    The chance of failure over time looks more like a bathtub curve: http://en.wikipedia.org/wiki/Bathtub_curve

    I think what Trent is trying to show is the distribution of years until failure.

  45. Mike S says:

    Ten years ago when I bought my condo, I went shopping for a new washer and dryer. I settled upon Roper (a Whirlpool-made brand) because “Consumer Reports” rated it the highest in reliability among all makes.
    It was one of my better decisions. Not only has the pair performed well for my laundry needs, neither has required as much as a service call. And I second Susan’s comments–my Ropers don’t have all the bells and frills of more expensive models, but it’s no hardship to turn a few dials and decide water levels, temperatures and cycles.
    Another example of why inexpensive doesn’t always mean cheap.

  46. Jimbo says:

    Stephan F – I never said that the investment has to be in the stock market, you could have put the $500 in TIPS or a bond fund and received a decent return.

    In general, it bothers me that Trent does all this nifty “calculations” that “prove” his point while overlooking other variables that detract from his points and make his conclusions and recommendations wrong.

    And Trent, I know that no one is obligating me to read your posts and that I’m free not to. I think it’s good to have some voices here who are not fawning over you and actually thinking about the arguments presented.

  47. Jim says:

    OK,.. looking again I see the x-axis is “years in use”. So the graph is showing the distribution of lifespan. It makes sense to me now. I’m just used to seeing failures over time shown differently than this, i.e. the bathtub.

  48. Congratulations Trent!

    You just explained some of the basics about Gaussian probability distributions more clearly and more quickly than my engineering probability professor could in several hours of lecture.

  49. Bill says:

    The use of a bell curve in this situation (either measuring the age of a machine at failure or measuring the chance of failure at a particular time) is not quite appropriate. Reliability and quality engineers generally use Poisson or Weibull curves. These curves can have many shapes, but in failure graphs what you’ll see is a front-loading of failure for poor-quality products. That is, take the bell curve shown and “skew” it to the left such that in cheap products, there is a quicker ramping up of failures, an earlier peak, then a longer, fatter tail that does not mirror the ramp-up. Higher quality products have the peak later in time, or perhaps a skew to the right. This makes sense because in complex systems made of cheap components, any one of them failing causes a system failure. This means that there are more opportunities for failure at an earlier time than products whose components are generally more reliable. After the initial “fall apart” timeframe of cheap products, they do better because we have basically demonstrated that we don’t have bad components. So, sometimes if it is cheap enough, and the “fall apart” timeframe is short, we can quickly weed out bad products by getting a bunch of them and end up with a few high quality, cheap products (I do this with dress shoes). When the “cheap” product isn’t that much cheaper, then the risk isn’t worth the weeding out process (e.g. Craftsman vs. Chinese hand tools).

    It’s great to see everyday problems analyzed quantitatively!

  50. Joey says:

    Here’s to another post defending your positions and mischaracterising those of your dissenters, eh Trent?

  51. lurker carl says:

    Failure has not been defined here.

    There are many parts within a product that can cause failure yet still be easily repairable. Most folks toss out the whole thing and buy a new one.

    Many products that are deemed unreliable as a model line often have only one or two internal parts that cause the breakdown, like switches and sensors. The specific parts with the high failure rates are usually to blame for the poor reliability ratings. Swapping out the dud parts for better quality replacements often resolves the reliability problem.

  52. k2000k says:

    Trent I love your blog, but I am having a hard time understanding this statement:

    “It comes with a good warranty, so there are no failures within the first five years.”

    That does not make sense, a warranty is no indication that a product won’t fail, it is simply insurance in case it does. It almost sounds like that since the machine is covered by the warranty and if it fails within that warranty period the customer gets a replacement so it doesn’t statistically count.

  53. Ben says:

    Gosh im glad Jimbo is here to make us think about trents posts. Without him, I would be taking the words on this blog as scripture. Not Jimbo, he takes hours upon hours just to tear down every point that trent makes, he truly is doing God’s work!

    Seriously though, I too am a seeker of truth and I think that healthy debate is a fantastic way to grow and learn. I do not agree with everything that Trent writes but I come here because there is always something that I can take away from it. On the other hand you come here to get on your soapbox and make yourself feel important.

    The real kicker is that I have let you spread your negativity to me. I have just had a epiphany that a tiny part of me reads the comments just to see what ridiculous comment you have made. Realizing that fact has put a real damper on my day. So thanks again for brightening all of our lives Jimbo.

    @Trent- I have been a reader for six months now and I really enjoy this blog, sorry my first post was so negative.

  54. jc says:

    there’s a lot of hand-waving this post, dressed up in graphs & percentages. even a machine reliable enough to get CR’s seal of approval may start failing at a HIGHER rate later in its life-cycle (after CR stops paying attention), especially if it comes with lots of extra features & buttons.

    without a full picture of the failure rate (which manufacturers jealously guard), you’ll never know whether a more expensive machine is worth it. the price is higher partly to try to convince you that it’s more reliable!

  55. The chance of a thing (washing machine, water pump, tire, dog, light bulb, etc, etc) is high early in its life (due to manufacturing defect, uneven wear, etc), low through the middle of its life, and high again when it is completely worn out at the end of its life. The shape of failures on the y axis to time on the x axis looks like a bath tub. I have a post on it on my blog as an argument towards buying something gently used vs buying new.

    http://cubiclewall.blogspot.com/2009/04/buying-used-and-extended-warranties.html

  56. TC says:

    If the bell shape is PDF, then the interpretation of three underlying areas and their meaning seems incorrect. Consider two nearly equally reliable machines, and their curves nearly overlapping with mean (peak) at 10 yrs and 10.01 yrs. In this case, it’s obvious that the orange area will occupy 99% or more of the entire area. But can you say, the slightly less reliable machine has 99% chance to outlast the slightly more reliable machine?

    In fact, if the two machines have exactly the same curve, the chance of one outlasting the other is equal to 0.5 (for normal distribution).

  57. I am not sure why my first comment isn’t going through, so here it is again.

    @kitty
    I agree that if the difference is 5 years vs. 20 years, it may be a different story. I am just saying that it is not always better to choose higher reliability with a higher price. Sure, each case should be considered individually.

    Also, if you do consider inflation, in my example it would still make more sense to buy cheaper. Between 1998 and 2008, inflation caused something that cost $500 in 1998 to cost $657.60 in 2008. This price is higher by $157.60. However, I would have made $395.42 in interest during that time, so I still would have made more in interest than I would have paid in inflation (I used an inflation calculator at westegg to arrive at the inflation number).

  58. Jimbo says:

    Even though you’re evidently a troll Ben, I’ll bite. Try and be sarcastic next time though, you failed this time.
    Have a good day!

  59. tightwadfan says:

    These lifespans were accurate in the past but they are no longer valid. Nowadays even the expensive machines have a lifespan of 10 years. And the cheap machines nowadays are not like the cheap machines of the past that might have lasted for 20 years. The cheap machines nowadays are junk.

    I would make the choice based on user experience (cleaner clothes, more features, fewer breakdowns, more energy efficient) rather than longevity. To each his own, having done a bit of appliance shopping the last 3 years I think there’s no ideal method. I miss the days when you could buy a reliable, mid-priced appliance from Sears and be done with it.

  60. Joe says:

    I agree with Jim’s comment… I’m more familiar with the “Bathtub” curve, where some appliances die in ‘infant mortality’…. essentially the customer is the QA. Those appliances that survive the early years breakdowns due to faulty parts or improper assembly, then they live on relatively trouble free for a number of years, (hopefully quality shows up here), and then the break down curve starts to climb as parts reach the end of their natural lifespan. Some items aren’t worth paying a lot more for, unless the features warrant it, as technology seems to be the big obsolescence factor these days in a lot of electrical gadgets.

    If I’m buying a shovel, I’ll buy a mid to upper mid level one, as I know it’s going to take a beating. If I’m buying anything with a plug on it, I’m figuring it will beg replacement in only a few years, but that’s just me. ;-)

  61. Carrie says:

    Why not just scour Craigslist and local ads to buy one cheap. We bought a previously owned dryer when we moved into our first home 7 years ago for $50. It lasted for 4 1/2 years. We went searching again and quite easily found a washer and dryer set for $175 and only 3 years old. The couple was moving and needed to get rid of them. They were energy star approved and have everything we need to wash and dry our clothes. Bonus: we sold our perfectly working older washing machine for $100. (It was inherited from my husband’s grandmother) Washers and Dryers are usually behind closed doors so they don’t have to look perfect or match! Unless we hit an unlucky streak where several fail in a short amount of time then we will never buy new again.

  62. Trent says:

    Jimbo, you are the troll and you are actively driving away commenters. I tolerate your trolling simply because, on the whole, you tend to spur more discussion, but as time goes on, you’re slipping more and more down the troll path.

    Today, two readers emailed me and pointed to you specifically as the reason that they do not comment on here any more – you bring such an air of negativity to the site that they no longer want to be a part of the discussion. That hurts the entire purpose of this site.

    I don’t mind criticism. I do mind mindless negativity and trolling that drives away people. I know from your comments that you know where this line is. If you keep crossing it, I’m going to start deleting comments for the sake of other readers who don’t want to participate in the negative environment you are trying to create here.

  63. This is a tough one– I would go and have gone with an informal survey of friends, family, and neighbors as part of my decision making process with appliances . . .

  64. Jimbo says:

    Well I’m sorry to be bringing “mindless negativity” to your pristine blog. You’ve succeeded, I won’t be commenting here or visiting this website anymore.

  65. Bill in Houston says:

    Actually, the Bell Curve does work when calculating failure. Remember that at the very end of the curve, 100% of the appliances have failed. Now his curves were simplified to drive home a point. In reality the ends (minima) should be elongated and tapered more. The left minimum should taper all the way to 0 no matter how much you pay. The difference is that the maximum point (the highest part) shifts rightward and the right minima shifts farther right as well.

    Paying more up front for higher quality means that you replace that item less often.

    Jumping back to the first post: While I do read Consumer Reports, I hadn’t thought to consult a repairman. Hang on while I slap myself in the forehead. My wife and I moved into a new house last fall. The previous owner left her appliances, of which we are using until either we remodel or they break beyond repair (that depends on finances, future children, my need for high class booze and expensive socks, et cetera)Excellent idea. Thanks :)

  66. Trent says:

    “Well I’m sorry to be bringing “mindless negativity” to your pristine blog. You’ve succeeded, I won’t be commenting here or visiting this website anymore.”

    I’ve sent your comment to the two readers who emailed me about your negativity making them uncomfortable. Addition by subtraction, I suppose.

    You’re welcome to comment whenever you like, but relentless personal attacks are not welcome here.

  67. Bill in Houston says:

    Oops, I goofed. I meant to say, the minima of the higher reliability curve should be elongated and tapered more.

  68. tentaculistic says:

    I appreciate the graphs, Trent, and the commenters who obviously got *way* more out of their stats classes than I did!

    I guess the take-away message I’m getting here (from Trent’s post and comments) is: do your product research, since price tag (high or low) does not guarantee quality level; however, the trend is toward better quality at a higher price. This is good input, since I am about to replace 3 appliances (pretty much everything *but* my washer!). Thanks Trent, and thanks folks.

  69. Chuck says:

    I would also like to see added in the equation the energy efficiency that a newer machine would have if you were to buy the less expensive machine and replace it after 5-10 years. By then the machine you purchase would most likely be more efficient and cost less to operate. You might come out ahead by buying the less expensive machine and replacing it later.

  70. dagny says:

    Engineers have a formal definition of reliabilty which has 3 parameters: mean time between failure, mean time to repair, and mean cost to repair. (“Mean” as used here is the average.)
    A washing machine should not go from working well to junk in a single step, as it often can be repaired. How often it needs repair, how long it is out of service for, and how much the repairs cost are the variables to consider. When it is older it may become irreparable if the needed parts are not available. Also at some point it may become so unreliable (fails often) that it becomes better to replace it.

  71. Sharon says:

    JC and Tightwadfan’s comments made me think of another (obvious?) point: I think the distribution of the failure curve is not known until the end of the lifespan of a particular model. And by that time, manufacturers will have begun producing different models with different reliabilities. So, I suspect that even manufacturers themselves don’t have complete reliability info by model (though if they did, they would certainly guard it closely, esp. for products of poor quality; this is the “market for lemon” problem from economics, in which the seller has info that the buyer doesn’t, right?)

    Re Consumer reports: I think the reliability info they report is based on reader reports of problems with products by brand, but not necessarily for particular models. (Also, CR’s ratings of brand reliability depend on who responds, and these respondents may not be representative more generally of the population of purchasers of a particular item.)

    Also, speaking of CR… a vacuum cleaner repairman/salesman suggeted to me that when products are highly rated by CR, their price goes up. So, you may actually pay more for a product that is well-rated by CR than for a comparable product. This guy was so persuasive that he ended up selling me a vacuum cleaner of a brand not rated at all by CR that was cheaper than the CR “best buy” I had been planning to purchase.

  72. jc says:

    Bill, you’re sorta right, but the curve is never going to be a “bell” or “normal” curve. there are, under certain circumstances, curves that look a lot LIKE a normal curve (especially a weibell distribution with a shape parameter = 3) but they’re still not the classic bell curve. i don’t believe that this is a distinction that should only be under the purview of those with a graduate education, though somehow it is…..

    i think this discussion should be fairly comprehensible, with a bit of patience: http://www.weibull.com/hotwire/issue14/relbasics14.htm

  73. Jeremy says:

    One thing to keep in mind:

    Scenario 1 — Buy a 10-year-lifespan washer at years 0 and 10.
    Scenario 2 — Buy a 20-year-lifespan washer at year 0.

    The average age of the washer in scenario 1 is 5 years. The average age of the washer in scenario 2 is 10 years.

    There is a good chance that the washer bought at year 10 in scenario 1 is more efficient and washes better than the washer in scenario 2.

    This is especially true for electronics. I would rather buy a $1000 TV now and a $1000 TV in five years than spend $2000 for a now-top-of-the-line TV and keep it for 10 years. Your time-averaged TV quality in the first scenario is almost certainly higher.

  74. Shauna Redmond says:

    I beg to differ being a former kitchen sales rep. The “cheap” machine often is simply the more expensive machine with a different face plate. They are made in the same factory on the same assembly line with the same parts and the only difference is: Drum roll please after they run 2500 GE top lines and apply the GE logo with the “prettier” front they run 5000 Estate machines and apply a less appealing face plate at the end. This is true in many scenarios. Not only do you pay more for the machine you are almost obligated to pay for the warranty on top of it to protect your investment. It is common knowledge in the industry that both machines are the same mechanically and you are simply buying a brand name.

  75. Shauna Redmond says:

    Ooops! I guided you wrong, Estate is Whirlpool not GE. But the facts remain the same. Same product different face plate.

  76. Shauna Redmond says:

    Ah Ha! I remember now Hotpoint is GE’s low end brand. Go to the Hotpoint Webpage and click the contact button and it will lead you to GE’s Website as a pop up. http://www.hotpoint.com/

  77. Sharon says:

    My mother is still using the Maytags they got in 1965. They certainly don’t make them like that any more! Dad had to do some minor repairs, but if mine die and Mom moves, I’m taking them!

  78. Katie says:

    Jimbo – If you bought a cheap machine, invested in this market, and had your machine break, you’d be kicking yourself for letting your money depreciate so quickly. Investments that beat inflation are never guaranteed.

  79. littlepitcher says:

    Shauna knows her appliances and can be relied upon. Best choice is the factory store and/or scratch and dent if you can afford that. The newer energy efficiency ratings make new a bargain when you factor in savings on electric and water bills.

    That said, I’ve always done well purchasing from moving sales, especially if you live in a military base area where folks transfer frequently. If you are short on cash, your employment is insecure, and your landlord supplies water and/or lights, get the used article and check websites for instructions on changing belts, hoses, and other ephemera.

  80. Eric says:

    Love the article Trent. Brings out the economist and love of graphs in me!

  81. kcdesi says:

    I full agree with lanhsin on this. Technology improvement is a big point to consider especially in electronics. I would rather buy a short life product for a cheaper price than a very long life product for a very large price!! That way I can get the latest technology product more frequently.

    KCDesi

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>