One of the face palm moments I had while researching PM's role in SaaS was the timeline for platform and app development. The traditional path for on-premise products was platform first, then application. The cloud products flip this sequence, putting the applications first, then the platform. We're so used to seeing this pattern that it's easy to take it for granted, or mistake the reasons why it exists (hence the face palm).
There were two reasons for this evolution in the natural history of a tech vendor's portfolio. The first is distance from the customer. In an on-premise world, where there's a lot of geographic and organizational distance between the producers and consumers of technology, the collection mechanisms about customer adoption (infrequent "How ya doin'?" conference calls, cryptic bug reports, etc.) are labor-intensive. For however much effort you put into this research, the information returned is often incomplete, distorted, or just plain wrong. For example, the game of Telephone played among product teams, customers, and the salespeople in between often leaves PMs with more questions than answers about what a customer really wants.
Therefore, the product strategy for on-premise products have to accommodate a lot of uncertainty about how customers use the technology. While it's not the only reason for customization and custom application development, it's certainly an important one. Rather than guess wrong about customer adoption, on-premise vendors tend to say, "Here's a toolkit. Go build what you want."