2 Week Feedback

GeoDrops Lounge
1

Hi folks -

I’ve had my sensors for a bit over 2 weeks now and wanted to provide feedback based on my experience so far.

Overall, the hardware build quality is impressive and for a first release, the software intuitive and I’m looking forward to seeing how it can integrate with/augment my Rachio irrigation system. Now that I’ve got 3 sensors fully installed (my 4th is still training), I can appreciate how important sensor placement is, so focusing on simplifying this process (consolidating some of the posts here into the app, for example) is time well spent.

I’m one of those folks dealing with shallow, gritty, heavily compacted soil and had significant challenges in getting my sensors properly installed. Specific questions that came up for me during that process were

  1. How wide is this thing vs. my hand trowel? Turns out my trowel was about half the width o the battery compartment and installation was much easier once I figured that out.
  2. Do I dig down deep enough to create a cavity for the entirety of battery compartment or is it expected to create space for itself as we push it down?
  3. Do I backfill space around the battery compartment and sensor itself with the soil that I removed to get it into place?
  4. If I can’t seem to get the soil sensor to confirm a sensor lock, should I try to improve my install spot (filling in with soil, using more water to soften, etc.) or try another spot? I’ve got a bunch of divots in my lawn now from defaulting to the the latter =)
  5. Install language in the app specifies a specify a solid blue LED light, followed by green blinks. In practice, this is a solid green led followed by green blinks. Just a documentation bug.

Specific pieces of feedback (many you’re likely already aware of, I’m sure):

  1. Keep Insight Zoom level (1/2/4 week) setting across sessions, instead of resetting. Right now, I’m spending a lot of time at the 1 week granularity because I’m trying to give feedback on the watering events and the need to zoom down each time is annoying. Definitely still present the slider as a graphical hint to the user that the scale can be changed though. Perhaps also change the default zoom level setting dependent on the amount of time the device has been in service? (staying zoomed in for new devices, then shifting out by default as more data is available).

  2. Fix odd “stretching” graphical behavior for water (possibly during training periods before max water calibration?) in insight view

  3. Data for previous day “disappears” regularly (at least the date is gone, although the graph is still present). It does appear later - perhaps an indication of some type of batch processing?

  4. Is it possible to indicate to the user when they’ve already submitted feedback on specific segments via Insights? I’m submitting feedback on watering, but find myself not sure which events I’ve already submitted feedback for and don’t want to confuse the model. Maybe highlight the submitted segment in green/red depending on whether it’s valid or invalid feedback?

  5. Add drop-down fields for timeline feedback - finger-based time highlighting is intuitive, but also imprecise as it usually shifts by a couple of hours when you take your finger off the screen. Makes it challenging to submit precise feedback for model training.

  6. Provide method to get back into soil sensing mode from the app? Several times during installation, I’d see the sensor pop out of soil sensing mode indicating that it had found a soil lock, but I knew from just touching the sensor that it wouldn’t stay where it was. In these situations, removing the battery cover to get back into soil sensing mode is annoying and risks debris incursion (since the sensor had just been in the ground). It seems like you’ve got hooks for something similar in the app (Move Device/Relearn Soil Profile), but the instructions (as I recall them) indicate that we should pop the battery cover.

Questions/Future Ideas :

  • Should we be providing feedback on rain as well as watering to help train the model?

  • Support for calculating AWC to improve performance in 3rd party apps (like Rachio). You reference that you’re not actually calculating this here though: Grass Moisture Experences - #4 by homedigy-lawrence and I recognize this may be considered a competitive disadvantage to facilitate.

  • Support for syringing (is this another name for pulse watering?) along with canopy temperature.

  • I, like many others, am hopeful for Home Assistant support. One way to potentially facilitate this at a relatively low power cost and (hopefully) lower development time would be to use BLE broadcasts that can be picked up and decoded by the Home Assistant Bluetooth proxy integration.

    I leverage this to monitor the temperature/humidity of 9 (small) greenhouses in my house via Govee hygrometers and use that data along with PlantMonitor/OpenPlantbook to check for plants that may be outside of acceptable humidity/temperature ranges. Home Assistant automatically detects the Govee BT broadcast after it powers on and prompts me to add the device to my setup. More details here: GitHub - custom-components/ble_monitor: BLE monitor for passive BLE sensors. Just an idea.

My devices in case you’re curious:

334JTQ
334MHW
337EAT
337RPF

Hope the feedback is helpful! Happy to answer any questions you might have.

2 Likes
2

Hi @Scythe ,

Thanks for the detailed feedback. Engineering is aware of this, but we will need more time to get back to you in detail.

Stanley
Marketing PM
Team GeoDrops

1 Like
3

Hi @Scythe ,

I want to let you know that we’re still looking into this. @homedigy-lawrence or someone else on the team will be responding soon.

Stanley
Marketing PM
Team GeoDrops

1 Like
4

Hi @Scythe ,

Hope you had a good Labor Day weekend. Just wanted to let you know that we’re still looking into this.

Stanley
Marketing PM
Team GeoDrops

1 Like
5

Hi @Scythe

Thanks for your very detailed feedback, it’s amazing to see have all the amazing users like you contributing to the GeoDrops community!
I wanted to make sure all of your points are addressed before my reply, so sorry it took me a while.

See responses below!

Cheers,
Lawrence

Re: Hardware Quality

I’m glad you liked the hardware quality.
We worked so hard on both the engineering design and manufacturing QC & Test for it. :slight_smile:

Re: Soil Installation

+1 to your point that it makes a HUGE difference how carefully the sensor probes are installed into the soil.

  • Good quality install = Stable soil moisture readings in ~1 week
  • Poor quality install = Can take 1+ months before getting stable moisture readings

Do I dig down deep enough to create a cavity for the entirety of battery compartment or is it expected to create space for itself as we push it down?

[A]: Yes, the idea is actually to dig about as deep as the sensor probe, in order to have enough room for the battery compartment. For compacted (less aerated soil), the battery compartment’s “shape” won’t be able to push the soil away, and you may find yourself having difficulty pushing the device flush against the ground.

Do I backfill space around the battery compartment and sensor itself with the soil that I removed to get it into place?

[A]: Yes please, and try to make the soil as tight as possible. Once the device is flush against the ground, you can step on it with your entire body weight.

If I can’t seem to get the soil sensor to confirm a sensor lock, should I try to improve my install spot (filling in with soil, using more water to soften, etc.) or try another spot? I’ve got a bunch of divots in my lawn now from defaulting to the the latter =)

[A]: Next, yes it’s actually better to pick a new location.

I’ve been thinking. Do you mind watching this 1min GeoDrops installation video, and and let us know if you find it more useful than our App instructions:

We actually sent this video to our Beta users back then, and got mixed feedback. Some preferred reading quick bullet point text, and found this 1min video too long. Would love to get your take on this!

ps - The “small divots” created by the hand trowels (shovels) actually aren’t that much wider than core aeration holes, so they actually help loosen up and aerate your soil, and should naturally go away in about a month. :wink:

pps - In parallel, @homedigy-leon is revisiting the whole installation instructions again and seeing how we can make the install instructions better.

Install language in the app specifies a specify a solid blue LED light, followed by green blinks. In practice, this is a solid green led followed by green blinks. Just a documentation bug.

[A]: Thanks for catching this! @homedigy-leon owns the entire App and will update in the next release.

Re: Interacting with Soil Moisture in GeoDrops App

Keep Insight Zoom level (1/2/4 week) setting across sessions, instead of resetting. Right now, I’m spending a lot of time at the 1 week granularity because I’m trying to give feedback on the watering events and the need to zoom down each time is annoying. Definitely still present the slider as a graphical hint to the user that the scale can be changed though. Perhaps also change the default zoom level setting dependent on the amount of time the device has been in service? (staying zoomed in for new devices, then shifting out by default as more data is available).

Fix odd “stretching” graphical behavior for water (possibly during training periods before max water calibration?) in insight view

[A]: Thanks for these feedback! We’ve seen people who preferred this other way around too, we’ll keep your feedback in mind!

Data for previous day “disappears” regularly (at least the date is gone, although the graph is still present). It does appear later - perhaps an indication of some type of batch processing?

[A]: Yes, there’re multiple tiers of batched processing in order to lower the cost of running Cloud AI Learning – it’s just so expensive to run!
As part of the large irrigation control Cloud software release, we’re actually in the middle of moving to a new data pipeline architecture where the pipelined data updates are done in smaller increments so this will look more seamless for everyone.

The only thing we unfortunately cannot work around are historic weather data. Weather forecast data are always there, but historic “observed” weather data (not forecasted data) can often take up to 12 hours or more to show up and get consolidated into our BigData database.

Is it possible to indicate to the user when they’ve already submitted feedback on specific segments via Insights? I’m submitting feedback on watering, but find myself not sure which events I’ve already submitted feedback for and don’t want to confuse the model. Maybe highlight the submitted segment in green/red depending on whether it’s valid or invalid feedback?

Add drop-down fields for timeline feedback - finger-based time highlighting is intuitive, but also imprecise as it usually shifts by a couple of hours when you take your finger off the screen. Makes it challenging to submit precise feedback for model training.

[A]: These are great suggestions! We may include this in our next updated “user feedback UI update”. @homedigy-leon

Provide method to get back into soil sensing mode from the app? Several times during installation, I’d see the sensor pop out of soil sensing mode indicating that it had found a soil lock, but I knew from just touching the sensor that it wouldn’t stay where it was. In these situations, removing the battery cover to get back into soil sensing mode is annoying and risks debris incursion (since the sensor had just been in the ground). It seems like you’ve got hooks for something similar in the app (Move Device/Relearn Soil Profile), but the instructions (as I recall them) indicate that we should pop the battery cover.

[A]: Another great suggestion, this one is actually a bit hard to do, but we will keep this in mind!

Should we be providing feedback on rain as well as watering to help train the model?

[A]: Great question. Feedback on inaccurate rain data unfortunately are NOT as useful, since we get these weather data directly from Official NWS Database – Our government provides really precise weather forecast down to Zip code level, and we currently don’t have plans to do our own AI Training to detect and improve weather forecast data directly from NWS

Re: Soil Science and Data API

Support for syringing (is this another name for pulse watering?) along with canopy temperature.

I, like many others, am hopeful for Home Assistant support. One way to potentially facilitate this at a relatively low power cost and (hopefully) lower development time would be to use BLE broadcasts that can be picked up and decoded by the Home Assistant Bluetooth proxy integration.

I leverage this to monitor the temperature/humidity of 9 (small) greenhouses in my house via Govee hygrometers and use that data along with PlantMonitor/OpenPlantbook to check for plants that may be outside of acceptable humidity/temperature ranges. Home Assistant automatically detects the Govee BT broadcast after it powers on and prompts me to add the device to my setup. More details here: GitHub - custom-components/ble_monitor: BLE monitor for passive BLE sensors. Just an idea.

[A]: Thanks for sharing this! We’ll actually actively working on a Google Cloud BigData access API, hopefully we can make it live within a month or so. I’m hoping people in our community can help write a Home Assistant plugin for it. :slight_smile:

There;s also a public dashboard coming, so people can directly view their histric soil moisture data on a large browser on their computer, instead of the tiny phone screen.

Also, instead of BLE, we’re actually working on a proprietary BTLR-based protocol in the background for the upcoming GeoDrops Station. You can find a bit more info about the GeoDrops Station on our website’s specs page here (We finally updating and porting all the Kickstarter and older contents to our new website 2 days ago):

Support for calculating AWC to improve performance in 3rd party apps (like Rachio). You reference that you’re not actually calculating this here though: Grass Moisture Experences - #4 by @homedigy-lawrence and I recognize this may be considered a competitive disadvantage to facilitate.

[A]: Actually it’s the opposite, we have spent so much time and effort to create something we think is better. I ended up wirting a seperate post for it just now. Can you take a look and let me know what you think?

See link below:

2 Likes
Weather Information Source
6

Appreciate the detailed response - I know you guys are trying to push out the IFTTT release, so I had expected a lag in response, so no worries on that front - just wanted to make sure I was providing detailed feedback that may be helpful in a (hopefully?) coherent format.

Currently living this exact experience =) 3 of my sensors locked within 10 days, the 4th is on its second 2 week learning period (seemed to reset itself at Day ~16). I’m about to core aerate, topdress and overseed, so I’ll get another chance to reseat all 4 sensors (yay?) =)

Yep, that’s what I eventually settled on, but it took likely a dozen attempted installs across 3 devices to stumble my way through, so hopefully we can decrease that learning curve for later users. I’ve seen the video that you’ve linked before - I’m a visual learner, so I was referencing it during my installs. The “shearing” graphic for the battery compartment at :24 seconds in the video was the reason for my question about the cavity creation.

Heard about the preference for a new location, although I had more success when I improved the install spot (went to this approach later in my install attempts). Obviously this runs the risk of creating an “artificial” environment that’s not actually reflective of real soil conditions. Caveat emptor applies!

Sorry, I phrased that badly. I should have asked more directly - do you consider both irrigation and rainfall “watering events” that users should provide feedback on? I wasn’t sure if you had only wanted feedback on the irrigation side.

Yep, you and I talked previously about the possibility for Thread support as well =) Completely get that you’re going the BTLR route with the Station and that BLE broadcasts would benefit only a very small portion of your install base, but figured I’d mention it as another possibility since you’re already turning on your antenna at least every 2 hours (I think that’s the polling interval for device status?) and a broadcast doesn’t require any type of handshake.

I like the article. It’s a bit hard to judge the value of it right now given that end users can’t control irrigation via GeoDrops (so we don’t know what targeting a moisture level would work, especially with regards to troublesome installs), and also assumes that a user will utilize GeoDrops to control irrigation versus providing ground truth to augment the predictive nature of most controllers.

I can only speak for myself (as a hardcore geek who likely doesn’t represent the vast majority of your targeted customer base), but I haven’t decided what will actually drive my irrigation system - will it be GeoDrops, Rachio, or some amalgamation thereof (e.g. I could use the data from GeoDrops to update Rachio with ground truth moisture (since it’s already pulling in forecast information from my Tempest).

So, when I’m asking about AWC, I’d be looking for input from GeoDrops to help adjust the advanced settings of my Rachio zones. That’s an ask that i think many customers may ask for, assuming that they’re not driving irrigation via GeoDrops yet.

There’s also a cost component that most will consider - in my case, I lean towards a hybrid model where GeoDrops feeds data to Rachio because I have 8 Rachio zones, but only 4 Droplets, so GeoDrops can only accurately control half my zones, but better calibration data (provided by moving Droplets from zone to zone over time) can benefit my entire system. Hopefully I’m explaining that in a clear manner.

Ok, overly long-winded response over. Very much appreciate the hard work you guys are doing on this stuff and I’m looking forward to consuming/acting on the data that GeoDrops provides!

2 Likes