Those versions of IE might represent 25% of web users, but any given website will probably see them as much less than 25%.
I run multiple sites, and most of them have IE (any version) as under 10% of users. One of them doesn't; IE represents ~13.5% of users on that one.
My untested theory about this is that IE users are generally less adventurous web users, sticking to sites like Gmail, YouTube, local news sites, etc.
TLDR: unless you're working on Gmail or a site where you expect to soon see the world turn up at the doorstep, it might be safe for you to mostly ignore IE.
This is true, but when these particular browser account for less than 1% of your traffic then you can make a call on whether it's worth supporting those users. Perhaps that loss is acceptable.
This, of course, is going to be different for every project and website. Each team would need to look at _their_ browser usage stats and make a judgement on what resources you want to spend on supporting the stats that you're going to see.
I agree; that's where it gets a bit tricky. You're not going to have many repeat customers who can't use your product.
You have to either be enhancing an existing product where you know that the segment in question isn't worthwhile supporting, or be prepared to estimate whether that segment is going to be meaningful for you.
Another thing to keep in mind is that sometimes the result isn't a deal-breaker. Suppose I try to get vertical centering working on one of my sites using flexbox. Perhaps the 5% IE users simply see the content of each div at the top rather than being vertically centered. Not ideal, but not the worst possible outcome, either.
Out of interest, does anyone know how flexboxes break on IE if you try to use them for vertical centering? Does the content actually just sit at the top?
The problem is that 25% of your customers could easily represent 100% of your profits. So by the time you're done with all your fixed costs and still need to turn a profit having a few % more or less users may make the difference between a company that will succeed or one that will tank.
That's why any commercial entity will be very wary of closing the door on paying users no matter how old their setup is.
Yes, if you center vertically with a flex-box, and you aren't doing anything weird with margins or overflow, and that's the only thing you do with it, you'll get non-centered results on IE8-9.
Not a problem.
But if you're only looking for that, then it seems very churlish not to use table layout, since that is the same amount of CSS and works in IE8+. There's no excuse for breaking layout in a browser if the alternative is just as simple and has no downsides.
Once you need to use flexboxes for more than vertical centering, it can cause more problems. Using it in place of inline-blocks, for example (one of its useful applications), can leave your content very weirdly laid out on older IEs.
The market share varies dramatically by niche. If you're running sites aimed at developers, you'll get very different results to a site aimed at teens on mobile, or one aimed at large blue-chip corporates.
Supporting IE < 11 shouldn't be a technical decision, I'd say. It is simply not technically difficult to do (outside of a couple of very niche web technologies). It is a business decision. So, you need to ask the CXOs if they're happy to leave those customers to your competitors to save you however many hours of dev and testing. There's a point where the answer will be 'yes'. Mostly we've passed that for IE6 now, but not always. I'm aware of one company targeting government clients, particularly in the developing world, for whom IE6 support is essential.
I run multiple sites, and most of them have IE (any version) as under 10% of users. One of them doesn't; IE represents ~13.5% of users on that one.
My untested theory about this is that IE users are generally less adventurous web users, sticking to sites like Gmail, YouTube, local news sites, etc.
TLDR: unless you're working on Gmail or a site where you expect to soon see the world turn up at the doorstep, it might be safe for you to mostly ignore IE.