Technology / 

17 Jul 2017

Notes on laptop reliability (2009)

Link to PDF of original study: https://www.squaretrade.com/htm/pdf/SquareTrade_laptop_reliability_1109.pdf

I found this study particularly interesting; these were my thoughts on reading it, updated and revised in 2022 where noted:

• “1 in 3 Laptops fail over 3 years”
  • Well, no–one in three laptops covered by Squaretrade’s warranty failed within 3 years. Or, put differently, 1/3rd of people who maintained a Squaretrade warranty for 3 years ended up taking advantage of it.
  • We don’t know for sure what the reliability is like of laptops generally, although it’s almost certainly higher (i.e. lower failure rate) than Squaretrade’s cross-section of the laptop market.
• “Two-thirds of this failure (20.4%) came from hardware malfunctions, and one-third (10.6%) was reported as accidental damage.”
  • The hardware malfunction number feels high, but I can’t back that up with any actual data.
  • If there’s a higher deductible or out-of-pocket copay for accidental damage vs. hardware failure, that would go a way towards explaining it; people would have an incentive to miscategorize damaged laptops as malfunctioning (defective) ones.
• “ASUS and Toshiba were the most reliable manufacturers […]”
  • Despite some of the other flaws in the methodology (the sample not necessarily being representative of the average laptop user), I can’t see any obvious reason why this conclusion isn’t sound. Or at least worth considering.
  • ASUS and Toshiba definitely wouldn’t have been at the top of my guesses for “most reliable” laptops, although I’m not sure on second thought what brand would have been.
    • Maybe IBM/Lenovo? I’ve personally not had much trouble with “real” Thinkpads (aka T-series and R-series; none of the slimline Lenovo-designed stuff), and they seem to be designed for easy in-the-field repairs, which might cut down on the number of machines sent in as DOA to a warranty company. (I mean, if you could repair the machine easily, you’d want to do that, right? And not have to migrate all your data, etc.?)
    • Apple? No, probably not Apple; they seem to occasionally push into the boundaries of class-action lawsuit territory with hardware defects.
    • Dell? My gut feeling is that Dell produces middle-of-the-road machines across almost all dimensions, including reliability, and then compensates for this–if you’re the sort of person or business who cares–with same-day onsite service plans. Spendy, but probably less expensive than building a really bulletproof laptop would be?
    • Panasonic. Speaking of ‘bulletproof laptops’, I’m surprised they’re not at the top of the list. Aren’t the Toughbooks literally designed for people who want to be able to run their laptop over with a car once in a while?
• “Laptops have historically been among the most popular electronics gifts, and this year the netbook is especially attractive to shoppers operating on tight budgets looking for an inexpensive way to replace an aging computer.”
  • Ugh, ‘netbooks’. They’re just cheap laptops, people! There’s no difference between a higher-end ‘netbook’ and a low end ‘notebook’ computer.
  • 2022 Note: Ah, netbooks. Remember them? I’m not really sure why we were supposed to buy them, in retrospect. They made a bit of sense for kids and students: carry a cheap laptop to class, and no-harm/no-foul if you drop and kill it. But for an adult, why would you want an underpowered laptop, if you already have a good laptop? And most people had already started to move from desktop-as-main-computer to laptop-as-main-computer by 2009. Today (2022), most people seem to have settled on a (decent) laptop plus a smartphone, and maybe a tablet if you’re into tablet-y things. And netbooks pretty much don’t exist as a distinct category. Oh, well.
• Both accident and malfunction / defect rates appear to be linear over time (see Figure 1 in the PDF)
  • I wouldn’t have guessed this, and it’s one of the reasons I suspect that a lot of the claimed ‘malfunctions’ are actually accidental-damage claims being mischaracterized for one reason or another.
  • Defect rates for most types of hardware are higher immediately after purchase and during the first few months, then tend to decrease for months or years, and then go up again as components get closer to their MTBF and end of lifespan.
    • See the “bathtub curve”.
    • Total failures are generally thought of as the sum of three main categories of component failure: early failures, random failures, and wearout failures. Early failures are highest immediately after delivery, random failures occur at the same rate throughout the product’s life, and wearout failures become more and more likely over time.
    • Squaretrade’s data seems consistent with only being driven by random failures. (Or, you know, people dropping their laptops and then claiming it suddenly and mysteriously stopped working.)
  • Squaretrade notes in the text accompanying Figure 1 that there is some nonlinearity to the failure rate: “There is […] a notable acceleration of malfunctions in the second and third years. While fewer than 5% of laptops failed from malfunctions in the first year, an additional 8% fail in each subsequent year”
    • That’s still weird. You’d expect the overall failure rate to start high, then dip, then increase over time. They just show it starting low (well, not sure 5% is ‘low’) and then increasing after the end of Y1.
• “We find 5.8% of netbooks to have a malfunction in the first 12 months, over 20% more than entry-level laptops and nearly 40% more than premium laptops. Note that this data excludes failures from accidents.”
  • Not entirely surprising; again, ‘netbooks’ are just cheap laptops. When you’re building a machine to a low pricepoint (they define ‘netbooks’ in the paper as anything selling for under $400 MSRP), you’re probably going to cut corners.
  • I appreciate their attempt to exclude accidents, but c’mon, we know people are lying through their teeth to a warranty company. Accidents are almost certainly a big part of ‘defect’ claims, making the actual difference in reliability between categories of machine smaller than it appears. (I think we can safely assume most people are more careful with a $1500 machine than a $350 one.)
• ”[After 3 years,] we expect netbooks to have a 25.1% malfunction rate, entry-level laptops to have a rate of 20.6%, and premium laptops to have the lowest rate at 18.1%.”
  • While the numbers are probably exaggerated vs. non-warranty-purchasers, the overall conclusion probably holds water. If you want a reliable machine, you’re going to pay for it.
  • It would be interesting to know what the 3-year reliability numbers are like, if you only buy pre-owned/refurbished units that are 12 months old or something. That would mean you’d avoid the early failures, but be closer to the end-of-life wearout failures. Unfortunately Squaretrade’s data doesn’t help us with this question.
• Failure rates by manufacturer
  • Here it is, the money shot: ![[laptop-malf-rates-2009.png]]
  • No error bars, so your guess is as good as mine whether the differences between, say, Asus and Toshiba are significant.
  • Doesn’t seem to be any consistency by country of headquarters: Apple, Dell, Gateway, and HP are notionally “American” brands, while Toshiba and Sony are Japanese, Asus and Acer are Taiwanese, and Lenovo is Chinese. I guess maybe you could say that Japenese brands as a rule are more reliable.
  • 2022 Update: Well, not everyone made it through the last decade or so. Of the 9 brands noted in the 2009 report:
    • Asus: still around, and still making notebook computers, although the “EeeBook” low-end netbooks are gone. It’s perhaps notable that Google selected Asus to manufacture its idiot-proof Chromebooks. Fell off the top-5 brands list in 2018, pushed by Acer.
    • Toshiba: still around, but some unfortunate accounting problems resulted in the near-bankruptcy of the company and forced them to sell a number of otherwise-profitable divisions to keep Toshiba itself afloat. One of these was the notebook division, which was sold to Sharp, and now sells computers under the ‘Dynabook’ brand name.
    • Sony: still around, but the PC business and “VAIO” brand name was sold off in 2014 to Japan Industrial Partners (really inspiring name). Interesting that the two Japanese brands, both well-reviewed in 2009, ended up getting sold by their corporate parents. My guess is that they just couldn’t keep the margins they were used to getting, as computers have gotten more and more commodified.
    • Apple: still here, still attracting the occasional class action lawsuit for selling products with known defects. Still, nobody seems to care. Technically the fifth-largest PC vendor by market share, although I suspect it’s probably much higher in certain socioeconomic tranches (and coffee shops) in the US.
    • Dell: still here, still boring. Third largest in the world by market share in 2020.
    • Lenovo: not just still here, but now, somehow, the world’s largest computer manufacturer. Definitely wouldn’t have guessed that when IBM spun them off; I figured the new owners would just ride the old IBM branding into the ground, throwing it on cheap parts-bin machines and calling it a day. But they didn’t, and they’ve done pretty well. The machine I’m writing this on is a circa-2012 Lenovo, and still has some of that old IBM solidity to it. (I haven’t used any of their more recent machines though. They could still fuck it up.)
    • Acer: still around, although they went through a major reorg in 2013. I don’t know anyone who owns an Acer, and come to think of it, I’m not sure I’ve ever seen one in the wild? Wikipedia suggests maybe they’re more popular in Europe than the US. Sadly, they don’t seem to be using the Gateway name or branding anymore, despite paying fairly good money for it back in 2007. Now the 4th-largest manufacturer by market share, just edging out Apple.
    • Gateway: RIP. Sold to Acer in 2007, who basically stripped it for parts. Maybe they’ll bring it back eventually for Millennial nostalgia value.
    • HP: In 2015 the company formerly known as Hewlett-Packard split up, with the consumer division getting only the initials “HP”, and the actual names being retained by the enterprise-hardware company. (I’m not sure how they avoid getting sued in Britain for violating the well known sauce’s trademark.) Now the second-largest computer hardware manufacturer in the world, after Lenovo (and only 2 points behind them, too).