When I had a 486 laptop at home, OS/2 felt like it was worth the money because it allowed me to access the power of the computer I already owned. Disk caching, lots of windows and lots of apps, multitasking, all stuff DOS didn't do on its own back then.
So I didn't resend spending it.
But Microsoft provided beta for free and it's long been quite easy to pirate MS OSes.
IBM did no blocking of pirates, but it expected journalists and evaluators to buy it, which is madness. It also released a time-bombed demo version of the OS, which is insane: it 100% will expire and nuke your PC, even if it works perfectly, which mostly it didn't.
Compare when I tried to review the new MS Exchange Server against Lotus Notes.
IBM sent me a copy of Notes and that was it.
MS sent me a copy of Exchange and Outlook, it sent me client licenses and OS media, and then it put me on a short off-site training course to make sure I knew what I was doing and was competent to deploy it, attach clients, and get it working.
(It wasn't amazing: for instance it couldn't collect mail over POP3 or IMAP. MS expected you to have a fixed IP, an MX record, and be receiving mail over SMTP like an ISP.)
But still, MS went the extra mile. IBM didn't. It spent fortunes on marketing, TV ads, everything, but it never occurred to IBM to get permission from Paramount Pictures to use the "warp speed" line from Star Trek until after the product had been named.
IBM spent billions but missed obvious stuff.
MS spent its money more wisely and worked to get the press, the developers, the vendors and so on on-side.
By the time IBM went solo, its vendors had been burned by missing the Windows boat, and were shy from pouring more good money after bad.
MS clearly knew what they were doing and they executed well. It's made fun of, but "developers, developers, developers" is right from Ballmer's time in the trenches, and he saw and knew that it was developers that got Windows over the hump.
IBM marketed it like an enterprise product, because they felt they only needed to sell it to "a few suits" like things they'd done before. Microsoft knew they needed to get everyone on it and developing for it, even if they had to eat piracy for decades to do so.
It's hard to see how else this story could have played out. OS/2 (1.x and 2.x) were whole new OSes, with the device-driver and compatibility issues that unavoidably entailed.
Secondly, in the era of ISA and EISA slots, PCs weren't really ready for this. MS had the clout to push initiatives like PnP, and PCI really helped too.
(As it also helped Linux. I remember the pain of trying to work out the right incantation of
linux aha152x=io:230,irq=11
... or some such nonsense to try to get a Slackware boot floppy to see my SCSI port, on which was the CD-ROM holding Slackware itself.)
Win9x was based on DOS, and so bypassed all this. But that couldn't help OS/2.
The one trick IBM could have executed that MS did, which would have helped, I tried to explain at the time to OS/2 communities online, to general bafflement.
NT could be installed from DOS. Right up to XP, and maybe later.
You copy the files from CD to a FAT16 partition, then run WINNT.EXE... just like with Win9x where you run SETUP.EXE.
This works very well and you don't need bootable media of the new OS that can access optical media or the network.
I deployed a whole network of NT 4 workstations this way. No optical drives, no USB.
1. Boot from DOS. Partition disk. Make a bootable DOS partition small enough to be efficient with FAT16 - say, 511MB. Reboot.
2. Install Netware client. Reboot.
3. Connect to server. Copy NT install files from server to C: drive.
4. CD to C:\WINNT, run WINNT.EXE.
WINNT.EXE runs in DOS, builds a new minimal NT system on the C drive. Then it reboots into that, and that formats the NTFS D: drive, and installs the OS into it.
It was quick (especially if you ran the SMARTDRV disk cache first) and it saved several thousand pounds in the cost of fitting optical drives across the fleet. That also stops users listening to music or installing their own software, and reduces potential failures, power use, maintenance costs, etc.
It means you didn't need to boot into NT to install NT. No compatibility issues, no SCSI or EIDE drivers, no network client, no faffing around modifying boot disks to see the source files to install from.
This was 100% perfectly possible for OS/2 but like the short filenames issue, I think it never occurred to them.
Agreed - I don't really think they cared much about "installing on existing systems" because IBM assumed everything would be OEM'd.
Everything about OS/2 feels like "the only existing market is DOS programs, and we can run those great, let's roll" - and they might not even have been wrong! I personally went straight from DOS 5.0 to Windows 98 SE (though I was using other people's Win95 machines during that time).
But installing on existing hardware is exactly what developers and "user fans" will do - it's almost the most important thing, until you get traction.
But I ran OS/2 2.0 for years and it was a thing of beauty for 1992. It made Windows 3 look very sad indeed.