How can transactional analysis bring value to banking business?
A quick insightful analysis on bank transactions.
There’s a sea of opportunity regarding data these days and banking is one of the industries that generate loads and loads of new data every minute. Being able to analyze it in order to get insights is, or should be the goal of every big bank or small fintech.
For this analysis we’re looking at a transactions dataset over the period of the entire year of 2021.
Here are a few questions that can quickly be answered with transactional data.
Q1. What kind of transactions the users mostly do?
Understanding the type of transactions can seem a pretty basic analysis but that alone can be a relevant hint to tell if you’re attracting the right kind of users for your business. You want to make sure the business model and the customers are all on the same page.
Clearly the TYPE_1 transaction is the most used type of transaction by the users, reaching about 65%. This makes a lot of sense for the bank since for this type of transaction is charged a fee.
The remaining 35% of transactions are represented by TYPE_2, TYPE_3 and TYPE_4 together. All of these are free of charge transactions so it's good for the bank that they are the less used ones.
We could go a little deeper here and look for the amount of fees this bank is charging per TYPE_1 transactions so that we have an idea of the amount of profit the bank had over this type of transaction.
The overall mean amount of fees is $9.41. The mean amount of TYPE_1 fees is $14.31 while the mean of all other transactions fee is only $0.20. So it’s pretty clear that TYPE_1 transactions are more profitable.
The mean amount of fees for TYPE_1 is 52% higher than the mean amount of fees for all transactions together. Now that’s impressive!
Indeed the bank’s profit over fees on TYPE_1 transactions is the most significant. It’s a great sign that the users are actually using the product we want them to.
Q2. Is the number of transactions significantly raising over time?
Taking a look at the transactions volume over the months is a great start to undersanding how the company is growing.
It seems the users significatly increased the usage of transactions services during the second and third quarter of last year but had a sudden drop at the last quarter.
As the goal is to increase the number of transactions over time, or at least maintain it, this plot isn’t exactly as expected. This could be a source for a metric that will monitor how strong the users membership is.
We could go a little deeper here and investigate if the number of transactions is being affected by new users joining in.
It looks like the rise in the number of transactions can be strongly related to a big raise in the number of new users at the begining of the second quarter of last year. It seems like that’s when the company started to get their first significant customers.
The recent users could start getting comfortable with the bank’s services and that made the number of transactions grow beautifully until August.
The sudden drop on the number of transactions can be partially explained by the number of new users not keeping up with the raise after June but it could also be linked to something else like customer satisfaction.
Q3. Let’s say the company is facing some database maintanence challenges that causes a need to take down the server for a couple hours. When would be the least caotic day of the week and time to do that?
Let’s take a look at how the transactions are spread over the week days.
Monday and Friday are the busiest days of the week for transactions. Those would be bad days to have the server taken down for maintenance. The number of transactions made during the weekends (Sat and Sun) are much lower than during the week, so that could be a good strategy.
Now let’s take a look at the time of the day.
Overall the busiest times for transactions are from 11am til 8pm and the less busy times are from 10pm til 9am.
Now that we have the less busy times regarding days of the week and hours of the day, let’s take a look at the less busy day’s (Sunday) times and determine a time window for the maintanence service.
The less busy time window on a Sunday is usually from 2am til 5am or from 9am til 12pm.
Those would be the chosen time windows to perform database maintainence and have as little customer friction as possible.
Wrapping up
Throught this analysis we could explore some situations where transaction data comes in to help us with decision making for a company.
- We understand that the company’s goals regarding the most profitable type of transactions is aligned with what the customer are actually using. The mean profit with fees for the transaction type most used is 52% higher than the mean profit over all transactions together.
- Regarding the number of transactions over the months, although we can’t afirm that is what caused it, there seem to be a strong correlation between it and the number of new users. To further understand that and the big drop on number of transactio after a few months, we would need some other data on customer satisfaction or so.
- We simulated an issue that would require taking down the server for a couple hours. We found that the best window time to do that would be on a Sunday either from 2 til 5am or from 9am til 12pm. This finding can be used to let our users know about it and aims for the smallest chance of customer friction.
I leave you with new questions…
- What other kinds of business questions can we answer with transactional data from banks?
- What other kind of data would we need for that?
- How can we improve customer experience with these findings?
Thank you for your time!
Let me know if you have any questions or feedbacks, I’d love to hear from you!
You can find the code for this analysis here and you can get in touch with me here.
