An Overview Of Ookla Intelligence Back Ground Data
All over the world Ookla is an established way to measure and compare internet performances between network operators. Different operators world wide get data from ookla intelligence. Operators then apply different data analysis techniques to generate insights from the data. In this article I will try to cover some aspects of this data and its benefits.
1- Data Types
There are two types of data shared by ookla intelligence.
1.1 Foreground Data
Foreground data is data from the users who performed speed test. This data is available for both Android and Apple users. However both contain different details. Processing and information of this data will be discussed soon in another article.
1.2 Back Ground Data
Back ground data is shared by devices which are not performing speed test however they are sharing different details in the background. This data is only available for android phones. In this article we will discuss this data in detail.
2- Reading Back Ground Data
One day file for background data is around 4GB . So this data cannot be run by conventional use of excel/power bi. Even with pandas we have to read in small chunks. However there is another library dask[1]. Dask loads the data and run computations only when needed and handles multitasking. This is very helpful.
3- Filter data of Operator /Competitors Per Techs
Whatever data is needed for analysis can be filtered using dask. Dasks only loads data once we ‘compute()’ the data.
Here for elaboration purpose data of operator and 2 competitors is loaded and also dds are separated per technology.
4- Analysis
As of today background data has 171 columns. There is detailed information about network , user location, cellid, signal strength and quality per technology. We will proceed with the discussion of some columns which give basic information.
Following is a utility method that can be used to analyze any column in a dask dataframe.
4.1 Basic Information
We can analyze basic information available using above function e.g trigger, devices, manufactures,roamers share.
Trigger if properly recorded over time can give a trend of mobility in different areas in summer vs winter. Also MMIMO beam profiles changing hours can be calculated.
Based on Manufacturer data marketing teams can launch different campaigns and based on device data network band allocation strategy. Roaming information is also available.
4.2 Technology share.
Get Muhammad Irfan zafar’s stories in your inbox
Join Medium for free to get updates from this writer.
we can get technology share.
4.3 Filter Clusters with undesired Technology
We can filter areas where users are stuck to legacy technologies , Operators may want to look into these areas
4.4 Coverage
Lets plot coverage as rsrp is directly available as a column.
A single point on one cell is dragged over time and there is no change in signal this makes rsrp unreliable
Well situation is not that bad we can process it and get first entry of every device id per cell. We will loose a lot of information but remaining info will be reliable.
These plots are shown for one day data , based on weekly data these can bypass requirements of any drive test.
4.4 5G Supported Devices on 4G
We can filter device id’s which camped on 5G and on same day they were on 4G. This can help operators to filter areas where 5G services can be improved.
4.5 Inter Operator Benchmark
Since this data is available for all competitors we can generate inter operator benchmark for coverage for tech share coverage etc.
5-Conclusion
Ookla background data is very useful and if properly analyzed can reduce drive test load and cost. In this article I have tried to cover basic available information. If you come across some other useful information please share . Code is available at github[2]
