Sense Blog

Sense Gets Smarter About Devices

Learn about the latest improvements we’ve made to our device detection methodology.

As you know, Sense is a learning device and we’re constantly improving how it learns. Hopefully, you’ve seen the results of our development team’s efforts this year in faster, more plentiful, and more accurate device identification. In addition, we’ve recently announced integrations with smart plugs from Wemo and Kasa. These don’t just close the loop from insight to control within the Sense app, but they also improve our device detection algorithms by giving us valuable “ground truth” data from specific devices in your home. These integrations are just the beginning and we can’t wait to release more in 2019 and beyond.

As Sense continues to learn, so do we. In the past two years that Sense has been in users’ homes, we’ve been cautious about how frequently and to what extent Sense will make changes to existing devices that Sense has already recognized. But we’ve compiled data showing that some devices are conflated, duplicated, or their identification is incorrect, so we believe that making these kinds of updates will be a good thing for most users. Recently, we analyzed heaps of data from the past year and made refinements to our existing algorithms to improve the accuracy and speed of our updates to existing devices, and you should start to see more of these kinds of updates going forward.

The first category of changes you will see impact device model grouping and name/type accuracy. Updates will be communicated to you on your timeline, and, in some cases, via in-app notifications when you open the Sense app. These new types of device updates include:

  • Device Renaming and Re-typing – If your Sense monitor has detected a device but is uncertain about its specifics, it has likely named it something general like “Motor 1” or “Unknown Heat.” Going forward, if it is confident that one of these devices is a specific device in your home, it will automatically rename and re-type it for you. This can help with unknown devices  as well as mis-typed devices, like a coffee maker that’s really a kettle or a “Heat 3” that’s really a coffee maker.
  • Device Auto-Combining – Sense may realize that certain devices (such as a Heat and a Motor) are actually part of a single device (such as a Dishwasher), and automatically combine and rename them for you.
  • Device Removal – At times, Sense may have thought it found a unique device, but after further analysis decided it was not a unique device after all. Keeping this device in your Devices list is not helpful for two reasons: (1) You might be wasting your time trying to figure out what it is, and (2) this model could actually prevent updated and more accurate models from running. Now Sense will start removing these unreliably detected devices from your Devices list more frequently, after notifying you.

In some cases, multiple updates might appear at once, such as a removal and a rename, in the case of conflated devices.

Device detection is not yet a perfect science, but your energy data helps Sense keep improving. Our Data Science team is excited to make these changes, both to improve your overall Sense device detection experience and to work towards the app helping you understand changes that are happening in the background.