How do I interpret Outlier Data or Missing Information?

Make the most of your data and quicky review unexpected readings faster by understanding outlier readings.

• Sensors scan the inside of the bin every 4 hours and send that information back 
• If there is consumption at the site, the reading you're seeing could be a snapshot of a past level. 
• Points on the Level History graph can also be viewed as a raw readings which can be turned on or off of the graph. You may see outlier when reviewing raw data.
• The graph is color-coded to display volume: 
  • Green (100-50%)
  • Orange (50-30%)
  • Red (30-0%)
  • Grey (outlier data point)
• Example 1 - Slightly less feed that what was delivered
  The current reading on the below graph is 3.38MT. This reading was taken at 10:56am. If you're looking at this bin at 12pm there may be less feed in the bin than the reading shows. 

• Example 2 - Outlier Readings / Data

  Sometimes readings aren't immediately apparent if they are outliers or not, and it can take a couple of readings before our Smoothing Algorithm will adjust the data.  Outlier data can be as simple as a scan running during a refill which can spike a level. The sensor will see that after further readings and pull the graph back to align with the rest of the readings. The outlier will still be a data point but the line graph will not reflect that reading.

   

• Example 3 - Sensor Covered in Feed

If your sensor is buried in feed and not able to see inside the bin to report back a level confidently it will report an outlier level. In the example below, the first 6 readings were outliers until enough consumption happened at the bin until the sensor could recover. Without the raw readings turned on, the sensor may appear like it is not reporting after the delivery was initially received. 

• Example 4 - Obstructions in bin 

This obstruction was blocking the sensors view inside the bin. The sensor was not confident in the level when the bin was full and reported outlier readings. Without the raw readings selected the graph would appear to have a gap at that time. 

We also see these type of outlier levels being reported if the sensor is knocked off it's aim. This is rare but can happen during a refill and results in the sensor camera aimed at the side of the bin wall.

• Example 5 - Overfilled Bin
  
  If a Bin has been overfilled it may appear that its not reporting as we're not confident in the level.  In the example below, the fill was registered as an outlier reading as the sensor was buried in feed. Once the sensor could view the feed level again the level came back to the graph. 
  
  
  
  

• Example 6 - Ratholing
  As feed is consumed, typically the feed begins to "rathole" or "cone-down", which means the bin level in the center of the bin is lower than the sides. In some cases these ratholes can be very large and when they collapse, can cause small shifts in the level. In the example below we can see an example of a rathole collapsing on the October 2nd outside of the graph.

• Example 7 - Missed Readings

The graph trend line in the example below supports a reliable feed level. However, the sensor has missed some readings so our algorithm is still showing you the correct trend line. Missed readings can also be caught by our service department and you may see a banner on your bin page if we need to take a look at the sensor proactively. 

• Example 8 - Not all refills are registered 

  On rare occasions, small additions of feed do not always register as a refill if they do not meet the requirement threshold. The graph below shows raw readings below the smoothing graph with small increases, you may need to use the raw readings to find your small refill. 

  

• Example 9 - Refill has taken place but is not reported on the graph

The bin received a refill but the sensor did not recognize or report the feed level change. If the bin was filled passed the sensor, the camera may not be able to clearly see the full feed pile and report this type of outlier reading as it cannot be confident on the feed level. If there hasn't been a reading since the refill, you may need to wait for the next reading to see the delivery. 

• Example 10 - Incorrect projections or sensor showing no projections

The blue dotted line on the consumption graph is showing an empty date sooner than you would expect, or not showing one at all. The Predicted feed level is reactive to reported levels in your bin, if there is a large drop in a reading the empty date of the predicted level could update to reflect a sooner than expected empty prediction. Similarly, a large increase in feed level (refill), will impact the empty date, wait a reading or two and if consumption is present the predicted date should align with your expected empty date. 

• Example 11 - Extreme cold weather events 

During the winter months in the northern hemisphere, we can experience extremely low temperatures for several days at a time. Our sensors houses a lithium-ion (Li-ion) battery which is charged by a solar panel. In extremely cold environments the battery may not have enough voltage and will prevent the sensor from operating reliably. As weather warms the sensor will recover and begin reporting again. We have several checks in place to preserve battery life in these situations.