The query worked and the account was wrong

I learned to distrust an elegant query only after I had approved one whose output made a commercial position look larger than it was. The source records were not fabricated. The arithmetic performed exactly as written. The error lay in what we had allowed a row to mean after attaching shipment detail to an order. That distinction became central to my work as an industrial data leader.

I began in June 1996 after statistics studies at Zerqelni Statistical College. At Terqenvi Analytical Services I prepared tabulations for business clients within a six-person team. I had no independent budget or client decision authority. Senior analysts reviewed the work, but much of my attention went to whether the calculation ran and whether its columns matched the requested format. I gradually learned that a correct total still needed a definition of the population it described.

In April 2002 I became a reporting lead, coordinating seven colleagues and ₹84 lakh annual team expense. Client requests often asked us to connect datasets prepared for different purposes. An order, an invoice and a shipment could all carry related identifiers without representing the same unit. I started documenting those differences, though I did not always insist that they remain visible when a result was summarised for management. A compact table could hide the very distinction needed to interpret it.

I joined Velqenri Industrial Spools in April 2009 as data integration manager. Eight colleagues supported reporting across commercial, finance and operating systems within ₹1.5 crore annual expense. The internal role made the consequences harder to escape:the person acting on a number was a colleague I would meet again. Operations head roles and finance owners retained their own decisions. My team had to understand the business sequence rather than treat every matching identifier as permission to combine and sum.

Promotion to head of analytics in April 2016 widened my responsibility to management interpretation. Eleven colleagues worked within ₹2.4 crore annual expense and ₹80 lakh approved programme capital. In 2018 I approved the extract that reported ₹18 crore against 240 orders worth ₹12 crore. Eighty orders with two shipment records had their header values counted twice. I explained the error to management and withdrew the affected account; the correction did not mean that the company had issued duplicate invoices or lost ₹6 crore in cash.

I became CDO in April 2021 and now lead fourteen colleagues through two leads. The annual data expense is ₹3.1 crore, with ₹1.2 crore capital approved separately. Velqenri employed 217 people in March 2026 and recorded ₹102 crore revenue and ₹8.2 crore operating profit in FY 2025–26. Founders hold 55%, an industrial investor 35% and employees 10%. I hold no equity or statutory appointment. CEO Geeta Qernvashi is my reporting manager; operations head Mohan Pelqeshi owns operating actions, and finance retains statutory reporting.

The delivery-risk prototype exposed a different change in meaning. Historical records included fields populated only after the proposed prediction point. They were real facts when the dataset was extracted, but the system would not have known them when asked to assist a new order decision. We held the prototype and funded a reconstruction of available information available at the relevant time. Removing thirteen fields did not automatically validate the other 107 or make the model ready for release.

Regional customer reporting required us to discuss identity with commercial colleagues. The same legal customer could buy through two regions. Adding their separate active-customer counts produced 210 memberships, but there were 180 distinct identifiers after accounting for 30 in both. Neither regional team had invented a customer. The combined report had asked a different question. We preserved each regional view while correcting the enterprise total and keeping legal entities separate from corporate-group analysis.

I try to make these discussions safe for the analyst who notices a problem late. A senior approval should not turn an ambiguous result into something nobody can challenge. I ask the team to explain the record population and the decision time before showing the most persuasive chart. In a prospective board role, I would use that experience to question management information without becoming the receiving company's data engineer. My governance preparation is unfinished, and the limits of both my evidence and my readiness belong in the account.

Open to Board appointments · Executive opportunities

Languages / Kannada · Hindi · English