Exploring the Impact of Lockdown on Mobile Data Speeds
Data visualisation of mobile internet speeds of different service providers during the COVID-19 pandemic lockdown in India.
Preface
A few days into the fifth (or sixth? I’ve lost track) COVID-19 lockdown in my city of Mumbai, there was a sudden power loss and my Wi-fi got disconnected. I was in the middle of watching an episode of Kingdom on Netflix and I was really desperate to complete the last few minutes.
So I switched on my mobile data and hoped for the same seamless HD quality that I was watching a minute ago. What I got instead was grainy, highly pixelated crap that took 15 minutes to buffer.
The. episode. had. only. 5. minutes. left.
So that got me thinking about mobile data speed in general and my download speed in particular. I really couldn’t remember the last time I had such a bad case of low speed data.
Motivation
I was curious to know if the lockdown and sudden surge in mobile data usage across the country had negatively affected average data speeds. My assumption was that my service provider, Airtel, was struggling with high speed data, much more than the others (call it user bias?!) and to test this, I carried out this small project.
Data Source
A quick search through Google led me to the Telecom Regulatory Authority (TRAI) website and their MySpeed Portal. MySpeed is an app developed by TRAI to allow users to monitor the download and upload speeds of their service providers (a.k.a Telecom Service Provider or TSP).
This website had state-wise and technology-wise (3G/4G) historic data from March 2018 and covered all the major TSPs in the country. They also have a well-written white paper on measurement of wireless data speeds, which makes for some interesting reading (as long as you don’t read it after lunch, like I did).
I downloaded three years of (2018 to 2020) data for March, April, May and June, because these were the crucial months of the nation-wide lockdown this year (so far!).
The entire project can be found on my GitHub repo.
Data Preprocessing
The data was distributed in different .csv files and I aggregated them into a single data-frame. The features of interest were —
month_year— The month and year of the datatsp— Name of the service providertech— 3G or 4Gtest_type— Either a download or upload speed testdata_speed— Data speed in Mbps (Mega bits per second)lsa— Licensed service area, basically the Region or State or City.
I then proceeded to clean the data of spurious values in each feature and reformatted the data types in the numerical features. As you can see the dataset is huge and takes up nearly 550 MB of memory!
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 8985844 entries, 0 to 8985843
Data columns (total 8 columns):
# Column Dtype
--- ------ -----
0 tsp object
1 tech object
2 test_type object
3 data_speed float64
4 signal_strength float64
5 lsa object
6 year int64
7 month object
dtypes: float64(2), int64(1), object(5)
memory usage: 548.5+ MBA few things to remember…
- Each data in the dataset is essentially a record of a speed test by a user through the MySpeed app and willingly sent to TRAI for quality control and analysis.
- The speed test represents the short duration upload and download speed.
- This data is crowdsourced and reflects the number of times the data speed was tested using the MySpeed app in a given month.
Exploratory Data Analysis
Number of speed test
Q1: How many speed tests were conducted over time?
The number of speed tests were the highest in 2018 than in 2020. The month-on-month comparison showed that there was no significant trend apart from a decline in speed tests from April to June across all three years.
Q2: Where were the most speed tests done?
Looking at the states, it was clear that across all three years, Uttar Pradesh led the country in the number of speed tests, followed by Maharashtra. This contrast was especially prominent during the pandemic, although the numbers were less than the previous years.
Q3: Who did the most speed tests?
During the Lockdown (March-June 2020), more than 50% of the speed tests were done by users of Jio (Reliance Jio Infocomm), with Airtel (Bharati Airtel Limited) coming a distant second. This vast disparity is due to more Jio users in the country. (Jio has a 33.47% wireless subscriber market share, followed by Airtel at 28.31%)
Q4: What kind of speed tests were done?
During the Lockdown months, across all the TSPs, the speed tests were nearly equally divided between the upload and download speed tests, but this is expected, since the app automatically carries out a dual speed test.
Variation of Data Speeds
Q5: How did the average speed vary across competitors?
Looking at the average data speed (in Mbps) across the three years, one can observe that Jio users have consistently enjoyed very high data speeds that were nearly twice that of other TSPs.
The absolute values of the data speed indicate that Jio and Airtel users faced a significant drop in speed during the initial Lockdown months in 2020 as compared to their competitors. This could be attributed to their large subscriber base and hence higher load.
Q6: Was there really any drop in the data speed during Lockdown?
The year-on-year variation (comparing values in a given year to its values in the year before) in average monthly data speeds was plotted as bar charts. It is immediately obvious that all major TSPs suffered a drop in average speed in the first four months of Lockdown. Cellone, which had shown the largest increase in data speed in April 2019, showed that largest drop in speed in April 2020.
The lowest data speeds were observed in April 2020 for the majority of the TSPs, after which there seems to be some minor respite. Surprisingly, Idea and Vodafone are the only companies which have shown an increase in data speeds in the latter two months of the Lockdown.
Q7: How did the data speed vary across each service provider?
A look at the distribution in data speed showed that Jio, Airtel and Vodafone offered speeds exceeding 10 Gbps during the Lockdown, albeit inconsistently. The most often observed data speed was ~ 2 Gbps for most of these TSPs with Jio users consistently enjoying even higher speeds (From the bimodal nature of its distribution).
Q8: How did the average data speed vary across the country?
To answer this, I looked only at the top two TSPs — Jio and Airtel. Jio users in Madhya Pradesh enjoyed highest average data speed during the Lockdown, followed by Maharashtra. In Maharashtra (Mumbai is its capital) the average download speed was between 6–7 Gbps.
Airtel users in Tamil Nadu and Andhra Pradesh enjoyed the highest download speeds, followed by Madhya Pradesh and Haryana. Maharashtra users experienced speeds between 7 and 8 Gbps during the lockdown.
Conclusion
By the end of this project, I understood that —
- All service providers saw a drastic drop in download and upload data speeds during the first few months of the Lockdown in India.
- Jio users are far more than Airtel users in the country.
- Despite this disparity, Jio and Airtel offered comparable average download speeds during the Lockdown.
- People in Madhya Pradesh enjoyed high average download speeds across both the major service providers during Lockdown months
I guess my initial assumption stands corrected!
Some Interesting Links
Hope you liked this post.
Ciao!
