The things you can do with statistical methods nowadays might blow your mind if you think there is not much information in oven usage.
You could infer likely occupancy periods of the home, since obviously if the house is empty no one is pushing buttons on the oven, and use some basic priors (9-5 jobs, x number of kids) to develop what is likely a pretty accurate model that would be profitable to sell to advertisers. Perhaps there are sensitive voltmeters on the power supply that can detect usage of other devices on the same circuit. Not to mention that it is yet one more channel of information alongside the flood of data being generated about you and your home by the rest of the devices you have incorporated into your life, and these things tend to have superlinear benefits when combined a la "multi-task" ML contexts. This is not an exhaustive list.
You could infer likely occupancy periods of the home, since obviously if the house is empty no one is pushing buttons on the oven, and use some basic priors (9-5 jobs, x number of kids) to develop what is likely a pretty accurate model that would be profitable to sell to advertisers. Perhaps there are sensitive voltmeters on the power supply that can detect usage of other devices on the same circuit. Not to mention that it is yet one more channel of information alongside the flood of data being generated about you and your home by the rest of the devices you have incorporated into your life, and these things tend to have superlinear benefits when combined a la "multi-task" ML contexts. This is not an exhaustive list.