I don't even want to put IO into the device at all. Not only because it increases cost and size, but because I don't what the user having to interact. We have to find better ways to fit the device in your life, so you don't even think about it.
For example, I helped someone transfer their stuff from their old iPhone to their new one a few years ago. The way you're supposed to do it is touch your old iPhone to the new one and it'll just work. Needless to say, it didn't. I think it was about an hour of rebooting the old and new ones before it finally caught. Since there weren't any logs or settings to change or any way at all to influence the process it was more frustrating than magic.
Now, it's possible your product really is as simple as turning it on and it'll just work, in the same way a lamp is "turn on and it works", but if there's any configuration at all that the device does, please expose it to the users. Human brains are incredible at finding patterns, generally better than computers, and if there's a mismatch between the human's model of how something works and the device's model, it's best to allow the human to change the device's model
An app to my brain screams "depends on your phone and will be outdated at some point; requires picking and unlocking your phone to use it; will have updates that change/ruin it at some point".
I just want to feel a button and press it, especially for things supposed to be used in the dark while sleeping.
People's subjective experience may vary, but relying on an app objectively entails more complexity and risk exposure for the user than exposing on-device configuration.
It's quite possible that many people who say they prefer using apps do indeed experience a higher level of frustration over the full span of their usage, and are only expressing their immediate-term evaluation at the outset of usage.