If the each independent product value being manufactured is relatively low, and/or downtime results in lost production, but with little additional cost or damage to the process, the PLC is the likely choice. If the value of a batch is high, either in market value or raw material cost, and downtime not only results in lost production but potentially dangerous and damaging conditions, the selection should be DCS. A plant that has a $10 million batch of a cancer drug in production that relies on strict and continuous temperature control, for instance, has a lot at stake in the event of a glitch. In some chemical manufacturing, maintaining the process at steady state is critical, because if the system goes down, the product could solidify in the pipes. If the cat cracker in a refinery lower down it could be days before it can be brought back on line so that production can resume. This means lots revenue lost for the refinery. The DCS system is probably worth the additional upfront investment for these applications.
In contrast, the brewery bottling operation that only needs to run 10 hours a day to meet production schedules, and which can be shutdown for system maintenance, troubleshooting, or upgrades with very little impact on the bottom line, it is a classic PLC application.
Downtime is one of the gremlins you try to avoid at all cost in process applications running 24/7/365. And money is not only the deciding factor. Dangerous downtime is clearly another deciding factor in the process of system selection. For instance, a refinery has flares that are continuously burning off gas. The system controlling those flares simply can not fail, because if the gas is not burning, it's collecting and pooling, causing an extremely dangerous situation. The more volatile the application, the more it may need a solution with lots of redundancy to ensure that the system is available when needed.