So did I miss something in that process or do we copy and paste the info I’ve already entered into the newly discovered device?
Our helping with entering the devices specifics doesn’t apply to just our installations.
I believe the entries we provide helps them on the backend with mainly future detections across the entire community.
There is no way for us to force Sense to detect a device. I think the purpose of this new feature probably gives them an idea of “this is what I have, see if you can find them”. Sense has never had the ability for us to direct detections.
I know it seems like this would be easy, apparently it’s much more difficult than it appears.
There are numerous discussions here with the reasons.
Think of your entries as helping build data for future users. Where Sense will see a device that it recognizes because we’ve provided specifics.
Feel free to copy and paste ! What you see is product development priorities and integration in a scrappy startup in action. The device naming box has been around since the product first popped out of the oven - it was needed to get minimum viable product to the marketplace.
The new inventory is the result of longer term product vision, plus customer requests, though I think many of the customer requests came with the simplistic expectation that by adding the inventory, Sense would be able to check off found devices, making it easier for the software to immediately find the remaining devices (hint: machine learning won’t really work that way). The inventory function is more helpful for what @samwooly1 describes: Sense can get a better idea statistically of how they are doing WRT to the full customer base, not just the most vocal ones, plus understand what devices/device types are still the most unknown. That probably helps most in steering the data science team, and pointing them to datasets that need investigation for improvements.
Given that they were added at greatly different times, Sense is sensibly letting the inventory feature stabilize and get populated before integrating. I’m guessing they will discover a few adjustments they need to make in the inventory so that it aligns with the naming feature, prior to integration. There are lots of subtle issues they need to deal with given there are different naming variants and behaviors for different types of device naming boxes, plus the whole question of what happens when you make a change - does it change it in the other view and what are the side effects. Examples of different variants of naming boxes:
- standard detected device - can be deleted for redetection, tied to specific real device
- combined devices
- smartplug devices - can inherit names from smartplug app, can’t be deleted, can be associated with detected device or not.
- other integration devices - Hue, NDI TVs.
It would seem logical when going through the predetermined inventory list of already named devices (e.g., Dryer) and I am entering my dryer’s DNA information, that when I confirm with the Sense app that, yes, you did discover my dryer, it should allow the already populated device info to be assigned to this now confirmed device.
I understand me telling the software/service/engineers what ”stuff” I have won’t speed things up in the detection arena; I can appreciate this will happen organically and take time.
What seems easy is having a database filled in with a home’s make/model of devices, that one took the time to do, and confirm what Sense sensed then match up the newly found validated machine with the information in the inventory that had already been entered.
I know now there will be an eventual confluence of these two features although at present they are yet to be integrated into one.
I agree with you, that would be logical. But I think Sense needs to run the scenarios before building the automated matchup and sort out some of implementation details, especially if they have to consider things like shared names between the inventory and the device name box.
You would probably recognize a few of the challenges if you had been through the Hue and smartplug betas where both the smart device apps (Kasa and Hue) and Sense have control of the naming and other device data. Questions come up about which is the authoritative source (the Hue propagates bulb model numbers back to Sense), and should things always be named the same in both places or does Sense need to maintain an alias. Another example from the Hue side is that room names control how bulbs can be grouped causing the name values Hue propagates back to Sense to potentially not match up with the real room names. Or what happens when the TV model number coming back from the NDI detection doesn’t match the model number in your inventory ? Or how to handle the addition of a new device - do you have to do it in the inventory first, or in the device box, or is either OK ?
Bottom line is that there are probably dozens of these kinds of user model questions that Sense has to sort through before they do an automated linkup. but I’m betting they are already doing the the comparison you suggest behind the scenes. That’s a good reason to push out the inventory function ahead of the automation - to collect data on the kinds of issues that might be encountered.
I won’t rehash all the points given about because they’re correct: This is a feature we plan to implement but it’s not as simple as it sounds (I feel like I say that a lot…). One additional complicating factor that I don’t think was mentioned above is mistaken duplicate devices or actual duplicate devices. There isn’t a simple way to map your inventory device with the “correct” fridge or map X and Y fridge with the correct devices. In time, we’re hoping to solve these problems. In the meantime, this is giving us really great data.
I’m not saying it has to be automated, just give the option to access the prefilled inventory data so I can manually select the device from said inventory.
In the future:
”We’re 58% certain we’ve detected a dryer, we see you entered a GE Dryer model XYZ in your inventory, is this correct?”
There are design choices that have to be made before even that simple addtion:
- Do you allow yes, no and maybe ? (that’s an option in the current device name box)
- What happens differently if you select maybe vs. yes ? I can see a definitive yes being treated differently than a maybe by machine learning.
- Is that choice of linkage undoable ? How do you fix if you make a mistake ?
- What happens when you delete/rename/remodel/re-room the device from the inventory ? What happens when you do the same from the device list ?
- Bigger picture - do you eventually offer all the device naming and management features in the inventory as well ? Does the inventory eventually become the main control panel.
I’m not saying that all these questions are all hard to answer, but elaborating and answering all of them plus making sure there is consistency with other naming and model sources takes time and bandwidth.
This is not meant to be negative, but anybody who suggests that “It should be easy to change the software so that …”, has never been a product manager or support manager for a piece of sophisticated software, especially one that interfaces with other software products. One running joke is:
Normal guy/gal: How was God able to create the universe, earth, life and humans in only 6 days ?
Product manager: No installed base (or integration) to deal with …