Nest Thermostat Troubles: The Problems with IoT and Smart Devices Exemplified

The interface is confusing. Instead of showing me what the temperature actually is in my house, the main Nest interface usually displays the target temperature that it is trying to maintain. So, if I set the temperature to 68 degrees but it is currently only 65 degrees, the thermostat displays a big 68. This is kind of like if, when I logged into my bank’s app, it displayed the amount of money I wish I had, rather than the actual balance in my account. It’s much more useful to know what the temperature really is at a given moment, not what it should eventually become. I can get this information if I log into the Nest app, or click on the device itself, but then I have to perform additional, manual steps to get crucial information. I think this is just poor interface design.

Connectivity monitoring is difficult. I’d like to get an email when the Nest disconnects from my wireless network. This could be a sign that the power to my house has gone off — which, in winter, would mean I need to drain the pipes in order to prevent serious damage. Bizarrely, there appears to be no way to configure disconnect notifications from the Nest. The thing seems to be designed under the flawed assumption that it will always have power and Internet connectivity.

It tries to be smart but uses dumb data. The Nest is marketed as a “learning” thermostat because it supposedly learns your schedule and, using predictive analytics, automatically programs heating and cooling schedules to ensure that your house reaches your desired temperature when you are home, but that energy is not wasted on heating or cooling when no one is home to enjoy it. This feature is great in theory. In practice I find that it works poorly because the analytics are based on problematic data. For example, the Nest tries to determine which times of day I am home in part by monitoring how often I walk by the thermostat. The thermostat is in my dining room, and I don’t usually go into the dining room apart from when I am, you know, dining. As a result, the Nest seems to think I am home only at meal times. It should either have a better way of collecting data about my habits, or not use this information for predictive analytics because it is flawed data.

It requires manual maintenance. Occasionally, for reasons I don’t fully understand, my Nest goes offline and displays a blinking green light. This may be because of a loss of power, a failed update or something else. I’m not sure. Whatever the reason, I have to reset the device manually. That’s fine if I am home and notice the problem before the temperature in my house goes out of whack. It would not be so fine if it were the middle of winter, the furnace shut off and I arrived home to discover burst pipes and cracked radiators, all because the Nest broke and could not automatically recover.

Difficult interfaces. Most IoT devices have non-traditional interfaces. They can’t display as much information by default as you can on a traditional computer screen. Developers and designers must therefore think hard about which information to make available by default to administrators and users. This is especially important in cases where IoT networks are composed of hundreds or thousands of devices and collecting extra data manually (as I do on my Nest by logging into the app to see what the actual room temperature is) is not feasible.

Lack of constant power and connectivity. In a perfect world, IoT devices would always have power and would always be connected. But in many real-world contexts this is difficult to achieve. Devices should be designed accordingly. There are solutions like the LoRaWAN, but they are not necessarily used. My Nest would be more reliable if it connected using a non-traditional network designend specifically for the IoT, rather than relying on my local wireless network.

Poor data-based analytics. The ability to leverage data to automate the behavior of IoT devices is one of the most compelling features of the IoT. To work properly, however, analytics-driven management of IoT hardware needs to be based on useful data. Collecting data that provides only part of the picture you need to construct, and ending up with unhelpful results, is worse than not using predictive analytics at all.

Automated management. More so than traditional computers and servers, IoT devices need to be managed in a fully automated way. When you are dealing with hundreds or thousands of devices, or don’t always have humans beings in close proximity to your devices, you can’t rely on manual processes for managing them or fixing them when something breaks. This is especially true in cases where the devices manage critical infrastructure. Without full automation, the IoT won’t work well at all.