Why is the output from the multicell location REST API (Ground Fix) changing day to day despite sending the same cell information? Why is the ADV parameter not used when doing Cell Locationing (Ground Fix)?

[leo_nam]

The output from the REST API when doing multicell locationing changes from day to day even when the inputs are the same. Why is this the case and is it possible to keep the location being returned the same? I am using postman to send and receive the cellular location info.

On October 17, my cell information and output for the Ground Fix API was

Input:
{
  "lte": [
    {
        "mcc": 302,
        "mnc": 610,
        "eci": 197711621,
        "tac": 55005
    }
    ,
    {
        "mcc": 302,
        "mnc": 610,
        "eci": 141566989,
        "tac": 55005
    }
    ,
    {
      "mcc": 302,
      "mnc": 720,
      "eci": 8937225,
      "tac": 30175
    }
    ,
    {
        "mcc": 302,
        "mnc": 490,
        "eci": 102476822,
        "tac": 35004
    }
  ]
}


Output:
{
    "lat": 43.86699231,
    "lon": -79.38158322,
    "uncertainty": 433,
    "fulfilledWith": "MCELL"
}

 This was very accurate and very close to where I was.

On October 18, my cell information was the same and the output for the Ground Fix API was 

Input:
{
  "lte": [
    {
        "mcc": 302,
        "mnc": 610,
        "eci": 197711621,
        "tac": 55005
    }
    ,
    {
        "mcc": 302,
        "mnc": 610,
        "eci": 141566989,
        "tac": 55005
    }
    ,
    {
      "mcc": 302,
      "mnc": 720,
      "eci": 8937225,
      "tac": 30175
    }
    ,
    {
        "mcc": 302,
        "mnc": 490,
        "eci": 102476822,
        "tac": 35004
    }
  ]
}


Output:
{
    "lat": 43.8689033,
    "lon": -79.3758543,
    "uncertainty": 568.202,
    "fulfilledWith": "MCELL"
}

This was not at all accurate. I was not in the circle.

This difference in returned cellular location in the span of a day makes it very difficult to trust and deploy commercially.

Also, why does the adv parameter not change the returned result? The manually calculated distance to the tower has been accurate to less than 300 metres. Is it because the true tower location is not always known? 

[Gregers Nordic]

Hi, we are using multiple vendors to look up cellular locations, and worked on optimizing this system last week. The vendors have slightly different data about the cell towers, so the response for the same request will be a bit different depending on which vendor we use. For your example above the uncertainty has increased, but the median uncertainty in Canada has decreased after our adjustment last week. Indicating a general improvement for most users, but if the returned coordinate is off it is difficult to measure without being there and knowing the physical location of each device. So please let us know if you see this problem over a larger area or fleet of devices.

Cellular positioning isn't meant to get an accurate position, but it can give a rough estimate without using much power. For higher accuracy, please consider alternative solutions: docs.nordicsemi.com/.../LSOverview.html

[leo_nam]

I see. I understand prioritizing the decrease in average uncertainty across all devices vs a single random device. I was mostly thrown off by the inconsistency in returned results which, I guess, was due to work being done on the system as you mentioned.

Thank you