When mentioning connected devices these days, we immediately think about the cutting edge. A refrigerator that can order more eggs when you’ve just run out, or a speaker that gets pizza to your door by just asking it to. The more common ones are the smartwatches, activity trackers, smart light bulbs, and so on.
The western world is literally swarmed by connected devices, but the most important one is the smartphone. Packed with state of the art sensors, each one of them benefits from the always-on connectivity of their host device. But that’s not all, your car’s Bluetooth entertainment system is a connected device, and every single Wi-Fi router you are connected to is a device that can indicate something about your daily routine.
IoT can be described as a data flow. For example, the data flow is from a controller (mostly a smartphone, but it can be a sensor like a volume sensor) to a device (light bulb) is “smart” only because it can respond to commands sent through the network instead of an analog switch. Another example is the movement from an activity tracker sensor to your phone, where the “smart” part is the way this data is displayed on the dedicated app for that tracker.
Providing these bare minimum data flows is not what should be considered “smart”, they are just new interfaces which provide just a little less friction for the underactive among us. Combining data flows and making sure they are interconnected will create a true smart environment around the center of it all: humans.
The base of this human-centered data flow is a network of sensors, but not all sensors were born the same. Let’s compare single- and multi-channel data sensors. Single-channel sensors provide one data input. For example, a thermometer provides the temperature at a point in time; steps counter – cumulative steps in a time range; Glucometer – a person’s sugar level at the time of measurement. The smartphone is a multi-channel device – it has dedicated motion sensors like a gyroscope and inferred motion sensors like the GPS receiver and the cellular tower’s reception level. It has location sensors — a dedicated one (GPS again) and an inferred one, the Wi-Fi antenna. The phone is also equipped with sensors that can detect its surroundings like light sensors, the microphone, camera, and sometimes even a barometer.
This differentiation is key when analyzing data and using it for various IoT-dependent products, because when you’re fighting for your users’ attention, context is king. Approaching them at the right time with the right messaging can change the way they relate to your product.
The smartphone, an always-on, always-connected, and almost always-carried device is the perfect choice for a daily routine descriptor. It goes wherever its owner goes during the day. Accumulating and aggregating the data from it would reveal amazing insights about the owner’s daily routine.
Looking deeper into the data and finding simple patterns can help realize more complex moments, such as when the user sleeps, when a device is idling for a long period of time, connected to a power supply, and is kept static, and when the user wakes up, the first time the device moves after being idle during the night (relative to the device’s time zone).
The next step would be to try and predict the user’s behavior. If I can identify when the person is asleep every night, I can try and see if something else is happening around that time, such as lights being turned off or an alarm system turned on. Now I know when the user is planning to go to bed. These tiny moments during the day are summed up in an actual image of someone’s daily routine. That image that influences the way we can analyze every new piece of data coming our way. All of this is directly connected to where the user is, which can be discovered by GPS and WiFi routers. An awesome example of a routine of seeing the same SSIDs all the time. For instance, my SSID at home is “It’s a trap!”
I’m always connected to it when I’m there and it can be quickly identified as my home network. This is easy to see, in part, when looking at how many times I’ve connected to that network, for what length of time, as well as how long it took for the device to connect.
While the smartphone helps us focus on the daily routine, it’s time to put single data channels to good use. For example, a connected thermometer indicates an elevated temperature of 38 degrees C/100F, which might translate to a health condition, but the daily routine of the person would show that they were outside all day, walking and running, and it was a hot day according to the weather conditions in that location. Instead of prompting the user to make a doctor’s appointment, the wellness app would suggest moving away from the sun and drinking some water. Messaging is provided with context to the real situation of the user.
And it can get more interesting. Connected devices can improve otherwise limited lifestyles and help make them as active as they can be. People with type 1 diabetes are constantly struggling with their sugar levels, from the discomfort of the hyperglycemia to the life-threatening hypoglycemia. When a person with diabetes finishes an exercise, he or she is probably in some level of hypo, and driving in that state might be dangerous. By understanding the current situation, with their fitness device connected to an Android Automobile, the diabetes app provides immediate advice to eat some fruit and make sure their glucose levels are suitable for the needed concentration levels for that activity.
Implementations are not only medical. What is a “Smart Home”? It should be a home where connected appliances are reacting to you. The smartphone identified that you arrived home? Turn the lights on, open the blinds, let the connected speaker let you know that the water heater made sure you can take a shower now. Your surroundings respond to you with minimal activity. All of this can be achieved by making sure your current “smart” devices are switched on by an event, in this case arriving home, rather than a finger tap.
In the world of content, demographics are the center of everything. Understanding the person’s location (Geography), age, and gender are considered keys to delivering the best conversion rate for ads and making sure content will get the most exposure possible. Google (based on searches) and Facebook (based on interactions and likes) are known by using these attributes for better segmentation. But at the end of March, Netflix’s VP of product said “Geography, age, and gender? We put that in the garbage heap”. Instead, they decided to concentrate on taste alone, using the thumbs up and down ratings for each show. It worked very well as a suggestion algorithm but it lacks context, which in this case is “What will I watch today?”. Some viewers will prefer some lightheaded short episodes after a busy day, while others might rather watch a food documentary if their day finished early. All based on their general preferences, but with relation to their daily routine.
Context is the invisible ingredient that turns clumsy interactions to frictionless behavior. And with even a superficial interpretation of data channels around the users, their experience will improve.
I avoided the term “Machine Learning” throughout this article to show that even a huge subject such as that can be approached with baby steps and lean thinking. By utilizing Wi-Fi, which is everywhere, GPS data, and motion sensors that are available on every smartphone out there, data can be input to a simple condition clause that can later grow into more complex decision trees and pattern matching. Eventually, to make the most out of the data provided, complex algorithms would be the next logical step. Huge leaps can be achieved by using the basic input from the data around us to actual actions right here, right now.