ADDED: Sort of. Let me explain.
Apple does slow down the clock speed on the main processors of your phone as the battery wears down. I assume there is a good technical reason to do this, and it kind of makes sense. So, yes, they slow down your phone but not to sell you a new one, but rather, to help your phone be a better phone.
But, the slowdown can be reversed by replacing the battery. And, Apple has never made even the slightest move to inform people that this is a thing. So, it is like the time Homer Simpson was told by Marge to not eat a pie she had just made. Homer found himself walking across the kitchen with his mouth making an up and down scarfing motion in an arbitrary direction that happened to lead directly to the pie. “If that pie doesn’t get out of the way, I’m going to accidentally eat it” he proclaimed. Sure enough, the pie remained still and Homer ate it.
Similarly, people will buy a new phone because performance is way down, when all they had to do was to replace the battery. Apple is Homer pretending to innocently happen to eat a pie. The phone is homer walking along. You are the pie. Not a pretty picture.
So, really, the slowdown is a) engineered into the phone, b) causes people to buy a new phone, not a new battery, and c) the fix that would be so much cheaper is kept out of the available information from Apple.
So, yes, Apple intentionally slows down your phone to make you want to buy a new one, it turns out. Effectively.
How do I know this? From this excellent and well documented source.
And now, back to my original post in which I argue that something suspicious is going on but I don’t quite know what it is:
It is a widespread belief that Apple, as well as other computer manufacturers, do things that make your device, be it a desktop computer, a notebook, a smart phone, or anything, slow down as they lead up to and release, and begin to sell, a new version of their product.
I want to point to a study done that concludes that they don’t do this. The study is by FutureMark which is basically a benchmarking software producer. They to not explain in their methodology where they get their data from, but I will guess that it is from the phones of people who install their benchmarking app.
If so, then right there we have a problem with the study. Without describing the sampling design, the study is useless right out of the gate. But if it includes the sorts of users that will install a benchmarking app on their phone, the that’s a bias (and uncontrolled mysterious one at that).
The study has other problems. It does not specify how it would detect an intentional slowdown. It focuses on graphics production and display, not on the things that we know actually slow down, like how long it takes for your phone to open up the sharing icon thingie when you want to share something, or how long it takes for any kind of internet or bluetooth connection to form or work, etc. Also, old beat up phones that drive you crazy and make you want a new one are not simply slow. They are quirky. They stop doing things right, and they start doing new evil wrong things. Those things, not a slowdown in graphics rendering, are what drives us to the cell phone store.
So the study is flawed because there is no methodology or a suspicious one, it measures possibly irrelevant variables, and it may ignore important variables. Therefore the concususion they reach, that there is no shenanigans going on with the manufacturer, is not valid.
But wait, there’s more. Look at these two graphs and tell me what is wrong with the graph and associated conclusion.
You can’t draw that conclusion from the chart for several reasons. First, the Y-axis is four or five times taller than the highest bar, so variation in the graphed data is not visible. Second, there is no attempt to show variation. If Apple is being both nefarious and smart, they would slow down selected phones, the ones that are nearing the point where they are paid for, for instance. We would not be able to detect a slowdown in month-long data averages. But we would see an increase in variability by month.
There is the same problem with this graphic as with the last one, but worse. Notice that there IS a slowdown over time in performance. But, since the Y-axis is so tall you can’t see it. Here, let me fix that for you.
This fix, by scaling the original diagram, is somewhat bogus, but do notice that over time, each of the first two OS implementations shown slow down in the measure used (which as I argue is not the best measure). Also, I used the zero mark on the Y-axis. You may believe as many do that a Y axis must always have the zero at the bottom. This is simply not true. You can prove this to yourself with two projects. First, draw a graphic of the average daily temperature per month in Minneapolis. Use zero F as the bottom of the graph. Good luck with that. You have to put zero up on the Y axis to make it work. Now, take the same data and convert all the numbers to Kelvin. Now draw the graph with Zero Kelvin at the bottom. The variation will disappear.
For all I know, 450 units on this graph is an Atari 2600 made in 1977. Indeed, when the same article shows the iPhone 6s and 7 performances, which might be used to compare just to see how bogus or useful this scale is, there is no scale given! For all I know, an iPhone X gets a 540 while the iPhone 5 got an average of 538 on this scale.
Having said all that, I doubt Apple slows down your phone on purpose. But I can tell you one thing that I’m pretty sure most manufacturers of most computers, across companies and across operating systems, do in fact do. They stop caring if your machine is slow or works funny. The degree to which companies will support and keep optimized older hardware varies across companies and across time and even across hardware types. Your computer, a desktop or smart phone or whatever, will stop working properly when new operating systems come out, and new software is running on it, and where applicable, where new hardware is integrated with it, because design will tend to focus on the here and now, and short term future, rather than the then and done.
Historically this applied least of all to Linux, which famously could be run on any computer and ran well. However, even major distributions of Linux are starting to drop old hardware. If you have a 32 bit desktop or laptop you will soon have to start maintaining specialized versions of the OS, or an old version, because everything will be 64 bit soon (don’t worry, there will always be a 32 bit version of some kind. Probably.)
Here’s the thing. Manufacturers don’t have to plan obsolescence. It happens by itself. They could do a better job of battling obsolescence, and probably, of all the manufacturers out there, Apple is the most famous, I think justifiably, for making their fans buy new hardware.
So, maybe the conclusion of the Futuremark study is valid, that “[Our] benchmark results provide a unique insight into the everyday performance of each iPhone model over time. And, as you’ll see, there are no signs of a conspiracy,” but not for good reasons, in my opinion. This study will be cited (already has been cited) as proof that there is no shenanigans. But at the study stands, it is not without better documentation and better analysis, and even then, I doubt a study of this measure is very helpful.