Digital India Payments After Demonetisation

Ever since India demonetised high value currency notes there has been significant push for using digital payment personally from Prime Minister Naredra Modi and GOI (Government Of India) itself. GOI has launched new initiative such as BHIM and renewed thrust on UPI along with several incentives to drive down cash economy and move toward less-cash economy if not cashless. Since RBI (Reserve Bank of India) publishes payment data monthly I thought of exploring it and see if there is any interesting pattern or behaviour that Indians exhibit while making digital payment.

Load Data

We will be using R for our analysis and GitHub for code hosting. RBI provides data in excel form so we will be using following libraries to load, transform and visualise our data.

Now we will load and merge all data in memory for further processing

Loads and Clean data

In above code we wrote a generic function to read excel files properly and performed some basic cleaning operation. As USSD volume and value both were too tiny for any meaningful comparison we have removed them from data. This is how our data looks now

Untidy Data

Ask Questions

Awesome ! We got our first view of data now we need to understand it and frame questions that our data can answer us. I thought it would be great if we can get answers of followings:

  1. What is monthly and weekly distribution of payments?

2. Distribution of each form within month for volume and values ?

3. Which form of payment has highest volume and value?

4. Which form has low value and low volume transactions?

5. Which payment mode has highest variability and which is consistent across all days?

6. What is distribution of transactions for individual day and in entire month?

Now that we know what questions to ask lets transform our data so that it can answer them.

Data Transformation & Cleaning

Now we will use Tidyverse family of packages to transform and manipulate data into tidy form where each column will be a value and each row will be an observation.

Here first we have identified what unique attributed our data has i.e. column and then transform existing data according to that. In above code first we merged all column into key — value pair except day and month as we need them as column and then separated them in mode, value and volume other columns that we identified earlier. This is how our data looks now

Tidy Data

Now we have our tidy data its time to clean it. We observe that value and volume are in two different unit i.e value is in billion whereas volume is in million. Shall we convert them into singe uniform unit for fair comparison? After some pondering i thought to leave it as is as we can use different aesthetics to show them and we will never compare volume and value directly as it does not make any data sense.

Next was to order months chronologically rather then any other ordering as it makes sense and help user grasp a view of scenario over time.

Finally there is no column on which we can visualise and compare our data so we will add a date and weekday column. To complete cleaning and transformation of data.

Below is code for that.

This is our cleaned and transformed data which is ready for visualisation now.

Final Data

Visualise & Answer

  1. What is monthly and weekly distribution of payments?

As expected there were over 30 million digital transaction in December as people were having cash crunch and as money supply got better volume started decreasing steadily. March saw highest value transaction as people buy lots of insurance policy and make significant investment for tax purposes.

Here is an interesting pattern as volume are highest on Monday maybe retail customer prefers Monday of payments and business people prefer Friday as it has highest value transaction to settle all trades.

2. Distribution of each form within month for volume and values ?

Monthly distribution of all payment modes are consistent and there doesn’t seems to be any interesting pattern or outlier here.

3. Which form of payment has highest volume and value?

From monthly transactions by payment mode chart we can infer that POS(Point Of Sale) which includes both credit and debit card has highest volume understandably so, as infrastructure of POS was already in place before demonetisation. Whereas RTGS(Real Time Gross Settlement) has highest value as it caters to larger transactions of above 50000.

4. Which form has low value and low volume transactions?

Clearly without exception UPI has lowest number of transactions.

5. Which payment mode has highest variability and which is consistent across all days?

As we can observe POS(Point Of Sale) is most consistent throughout month whereas newer and smaller forms such as UPI and IMPS is more volatile then rest of others.

6. What is distribution of transactions for individual day and in entire month?

Again our earlier intuition is reaffirmed here that high volume transactions takes place early into week e.g. first Monday of April, second Monday of May, second Monday of July and third Tuesday of June are high volume transactions day.

For second part of our question:

Here an interesting pattern is visible as we reach end of the month there is spike in value in all months except April where we have dip in transaction value towards the end of the month.

Learnings

Its really interesting to have an insight on how India is going digital. Following are some of the interesting points that we established with Data:

  1. Huge volume pickup after demonetisation in December
  2. People prefer making high volume transaction early in week and month while high value transactions takes place at weekend and monthend.
  3. RTGS is most preferred for high value transactions, POS is at top for high volume transactions.
  4. UPI and PPI are fastest growing mode of payment supported by apps such as BHIM, Paytm etc.
  5. Monday is preferred for high volume transaction whereas Friday is preferred for high value transactions.

Thanks for your time and if you like this please spread the love.