Divya Shroff

It began quietly late last year when officers from the Income Tax Department walked into a few popular biryani restaurants in Hyderabad for routine checks. These were not dramatic raids. Kitchens were busy, customers were eating, and the billing counters were working as usual. But one thing did not add up. The number of diners inside the restaurant did not match the numbers blinking on the billing system. Officers noticed that some cash bills briefly appeared in the system and then vanished. The printed summaries looked clean, but the software logs told a different story. It was not a clumsy accounting mistake. It felt scripted. At first, the officers thought it was a local trick. Then they discovered that all these restaurants were using the same billing software. That was the moment the investigation stopped being routine. When officials traced the deletion pattern back to the software provider's backend in Ahmedabad, they unlocked nearly sixty terabytes of stored billing information from more than one lakh restaurants across India. A small mismatch at a biryani joint had suddenly turned into a nationwide data trail. Inside the tax department's digital lab in Hyderabad, specialists began reconstructing the deleted bills. Every transaction clicked into the system had left tiny digital marks, even if it had been "deleted" on the surface. HOW AI HELPED Artificial Intelligence tools helped read these marks and rebuild the missing pieces. What officers found was striking. Restaurants across the country had generated around Rs 2.43 lakh crore worth of bills during the six years examined. More than Rs 13,000 crore worth of those bills had been quietly erased after being recorded. This was not accidental. Some restaurants removed a few cash bills every day. Others erased entire blocks of dates, sometimes up to thirty days at a stretch. A few businesses kept all their records intact inside the software but still declared a much smaller turnover in their tax filings. The intention was the same in each case. Lower reported income meant lower tax. The scale grew clearer as officers checked individual states. Karnataka showed the highest value of deleted transactions at about Rs 2,000 crore. Telangana followed with about Rs 1,500 crore. Tamil Nadu, Maharashtra and Gujarat were not far behind. A closer look at 3,734 PANs in Andhra Pradesh and Telangana revealed suppressed sales worth over Rs 5,000 crore When investigators studied a sample of forty restaurants in the two states, they found nearly Rs 400 crore in unreported turnover. Some outlets seemed to hide almost a quarter of their sales WIPING DIGITAL RECORDS DOESN'T WORK Well-known investment banker and financial advisor Sarthak Ahuja explained what the probe revealed about modern tax detection. According to him, the tax department can now see exactly who is deleting invoices and manipulating digital accounts He noted that officers had analysed the billing data of 1.77 lakh restaurants since 2019 and found that many of them were suppressing sales by about 27% on average In several cases, transactions had been recorded and then wiped from the system either individually or for entire date ranges. He pointed out that the under-reporting added up to about Rs 70,000 crore in concealed turnover, which could mean a tax loss of over two billion dollars Ahuja also warned that wiping digital records is no longer a safe tactic because even cloud-based systems leave visible trails for investigators to examine As the puzzle pieces fit together, the probe spread from Hyderabad to Visakhapatnam and other towns in Andhra Pradesh and Telangana The Central Board of Direct Taxes (CBDT) expanded the investigation to other states, noting that the findings so far relate to one billing platform and that other software may come under scrutiny as well The department is now matching reconstructed bills with tax returns and bank records, preparing to issue notices and calculate penalties

I googled why one hour is 60 minutes and one minute is 60 seconds and the answer wasn’t even that exciting