Windows 7, ran solid, fast, no ads or weird shit, clean UI
Earlier pre-Clippy versions of Word before they added 8 billion useless features when it was still fast and did its job well
Wordpad, great slimmed down and fee version of Word. I wrote many college papers on it
Earlier Excel for single user smaller use cases (but was always inappropriate for large or shared file use for which it was never intended but used a lot)
Minesweeper, who hasn't played and then tried again and again?
The pattern here is that like all software companies, they create something good, then they destroy it by continuing to work on it. There's no reason to buy software again if it does everything you need, and Microsoft is too good at compatibility, so you aren't forced to buy it every three years because it broke in some Windows upgrade.
The only alternative for milking existing software is renting software, which sucks from a consumer point of view, but provides continuous revenue.
The right thing to do, of course, is to declare the software feature-complete and move on, dedicating only minimum employee time to maintenance, so that you get way less profit from that team every year, but have basically no costs, and shift the headcount to work on some new piece of software that will produce your next source of new profits.
But tech companies are lazy. They'd rather create something once and milk it forever and ever. Why is this a problem? Simple. 95-year copyright durations. If we rolled back to the original 1970 duration, when copyright was 14 years and renewable for 14, this wouldn't be an issue. Companies would have a limited amount of time to make money off of the software, because in 28 years, the software would be free for anyone to use, so with reasonable virtualization, you'd have folks using 1996-era versions of Word for free right now.
Add to that a second change in the law — making the cost of renewal be 5% of the total gross revenue for the product to discourage renewal — and you'd have a copyright policy that actually makes sense. If a company is making enough money for it to be worth that cost, then they should be able to extend it for another 14 years, but it should be expensive enough that they don't just automatically do so out of spite. That would mean that most software from 2010 would no longer be under copyright, and companies would not be able to milk software for every last drop of potential profit like they do now, and instead would be forced to actually innovate regularly with new products, thus promoting the progress of science and the useful arts.