Since you mentioned customer perspective, here's one:
For some of your customers (or potential customers) there might even be regulatory benefits if you, if you publish the driver as open source and take the effort of having it accepted into the mainline kernel and then further down the line.
Namely, in some fields (say medical), there are regulatory standards for code quality. If I'm using off-the-shelf code, like a driver for a device, I might need to "qualify" it as medical code.
For certain levels of criticality, in some regions, some of the arguments which might be used to qualify code are if it's in wide-spread use, and if there's a clearly documented development process using which the code was made, if there's a clear bug reporting and solving process and so on.
If the code is a part of Linux, then I can qualify it basically "for free", as a part of the whole kernel package. If it's not, then I'll probably have to do reviews on it, and might have to do an audit of your development processes, write many pages of reports on it and so on and so on.
In practice, this means, that if there's a competitor which has his drivers in kernel, I'm going to pick his product, even if it's more expensive or of lower quality, instead of spending a boat-load of money doing paperwork for your, and then repeating the procedure of every device I want to use your product in.