Preinstalls discourage other operating systems for two reasons:
1. they set a functionality mark that a on-preinstall OS needs to surpass in order to interest someone in switching (especially if the new OS costs money),
2. they enable inaction. What I mean by this is that you may find the features of a new OS interesting enough to want to try it, but doing so actually will take some effort. If you procrastinate, you've still got the old OS.
Before Win95, the preinstalls were generally either DOS, or DOS + Win3.
When Win95 came out, it was an immediate hit via retail sales. On launch day, many stores were open at midnight to sell it as soon as they could, and people lined up to buy it. I don't have the numbers handy, but I'd be surprised if Win95 on the first day of availability did not surpass OS/2.
This shows that there was a significant number of people in August 1995 for whom an OS of Win95's level could entice them to want to try it sufficiently to get them to actually go out and buy it and install it.
If IBM had properly marketed OS/2, those people could have been their customers. IBM could have established themselves as the preferred 32-bit PC OS before Win95 got out of the gate, and vendors would have been rushing to sell preinstalled OS/2 systems.
One can make a good case that after Win95 came out and was getting preinstalls, no amount of effort by IBM could have overcome that--Win95 was good enough that OS/2's advantages probably were not enough to overcome procrastination. But by the time Win95 was getting preinstalls, Microsoft had already won from IBM's indifference pre-Win95.