As digitisation is inducing transparency in systems across the government and the private sector, more and more people are responsibly filing their tax returns. Though it is pleasing to the ears, the huge data that this generates is hard to handle manually. Hence, tools like data analytics with machine learning play a crucial role in optimising revenue generation through taxes for the governments. Deliberating on similar lines, Showkat Ahmad Parry, Additional Commissioner, Excise and Taxation, Government of Punjab, addressed the Revenue Management and Intelligence Summit 2021.
He started on the note that “data is the most important aspect of human life.” Citing examples like Aadhaar, GSTN, income tax, etc. he said that “data is running the country today.”
Speaking on the significance of data analytics, he said, “I am currently, working in an organisation that deals with taxpayers in the State of Punjab. Only in Punjab, we have 3.6 lakh registered taxpayers. Each one of them, excluding the composition taxpayers, have to file the returns twice a month. This generates nearly 7.2 lakh returns a month and each of the returns has data spread across some 30 to 40 rows and columns. Through this, we can imagine the humongous amount of data available with the officer who is working with around 1000 dealers. How do we expect an officer, dealing with such data, to scrutinise it?” Therefore, it becomes extremely essential that we have enough tools and appropriate technology to empower the staff to deliver its duties efficiently and effectively. These technology tools are important so that the officers can focus on high-risk cases where a significant amount of revenue gets stuck. Data analytics have come in handy to scrutinise such data and help manage revenue, he added.
Apart from the data, the officers also need to predict the behaviour of the taxpayers. In almost all the departments, the government officers have to keep an eye on how consumer behaviour is changing. In this regard, there is a need to have interdepartment collaborations and strong data sharing policies. This is important as the number of units of electricity a firm is using is dependent on its productions and that decides the amount of tax it has to be paid. Therefore, all the departments are linked to each other. It is important that data from different departments have to be available. There are issues in the government departments that the interaction between departments is low due to which the required data is not readily available. If interdepartmental data sharing is established it will improve fraud analytics and other control measures. Therefore, there should be proper data sharing policies to ensure the flow of the data and transparency in processes.
To Watch Complete Video Click Below
Deliberating on technology tools, he highlighted, “Once the action is taken through a system then the system should be able to capture the data and learn from that to enable itself to detect and generate alerts automatically. This will improve our efficiency.” Along these lines, the department is currently working to develop a tool using machine learning technology in partnership with a private firm. So, the department is working with them to develop a machine learning tool to identify bogus firms. “Similarly, we have partnered with SAS. We have purchased their software and we’re working with it and we have done some analysis on that,” he added.
Drawing out a conclusion, he said, “Data is the new lifeline. All the departments need to synergise with each other, coordinate with each other and share. Also, centralising the data collection can be a great way to establish stringer communication between different sets of data and realise the potential of the data to optimise the revenue generation for the government.”
Supporting Siddarth Jain’s words, he said, “I agree that in today’s time, a lot of efforts are not needed. At the click of a button, we have so much data available that it is possible to get cases worth over four or five crores with limited efforts in a short period of time.” Adding on to the availability of data, he mentioned that in the initial days of GST, the government was lenient and wanted to make things easy for the business community. However, many took advantage of the situation and committed fraud. So, there are cases that the departments are fetching out almost every day. “We have our hands full of the available information and actionable data. In Punjab, we have cancelled suo moto for more than 30,000 firms which were proven to be bogus. And, the bogus billings from these companies were nearly Rs 35,000 crore involving a tax of Rs 3.500 crore,” he added.
He raised concern that mostly the actions are taken when the fraud has already happened. The machine learning-driven tools should be placed at the prevention level. “We are developing a machine learning tool with a private firm to predict bogus firms. We have been working for the past six months and recently the company shared with me a list of 223 firms, as per the machine learning tool developed. I verified all the 223 firms and of those over 190 firms were found to be bogus,” he pointed out. The tool analyses a firm on as many as 300 parameters like the type of addresses used, the pattern of filing of returns, the kind of PAN numbers, and more which is nearly impossible for a human to do. So the tool is definitely effective and has been of great help in recovering revenue.