I had made the register simpler by asking it to assume too much

The customer register looked simpler after the proposed merge. That was what persuaded me in 2019. Records sharing a telephone number could be gathered into one party account, removing apparent duplication from a management problem that had been discussed for months. Rina Zeqvamni asked us to inspect the relationships hidden by that shortcut. Among 920 candidate pairs, 610 had support for a same-person link, 230 concerned different people and 80 remained unresolved. The contact was useful evidence of a connection. I had allowed it to become a claim of identity.

I grew up in Shillong, where a household or workplace contact was often a practical way of reaching someone rather than a private identifier. That ordinary experience should have made me more cautious. I completed a statistics degree at the fictional Qezvami Statistical College in May 1995 and joined Nezqavri Information Services the following month. The firm had 72 employees and ₹9.4 crore revenue for FY ending March 1996. I was one of seven junior reconciliation analysts, with no direct reports or spending delegation. Our work concerned the relationship between records supplied by different operational teams.

At first I thought of reconciliation as finding the correct common value. Supervisors taught me that two records could disagree because they described different objects or different responsibilities. A reference used by a service desk was not necessarily the identifier used by the contracting party. I learned to preserve the source and ask what a proposed connection meant before treating an apparently neat combined account as an improvement. That lesson was more demanding than finding a spelling variation or checking a total, because it required listening to the people who created the record.

In September 2002 I became reconciliation lead. Nezqavri employed 188 people and recorded ₹42 crore revenue for FY ending March 2003. Thirteen colleagues worked within my ₹1.2 crore annual function expense. I reviewed client matching instructions and explained the cases our team could not settle. The position made me responsible for the consequences of a recommendation, although clients retained control over their own records and business decisions. I wanted to understand those consequences beyond the end of a service assignment, which led me into insurance in June 2009.

My first role at Vezqerthi was head of customer-reference stewardship. The insurer then had 410 employees and ₹540 crore gross written premium for FY ending March 2010. Sixteen data colleagues worked within ₹2.4 crore annual function expense. We dealt with policyholders, contacts, financing interests and people acting on behalf of others. A convenient common register could help colleagues find a record, but it could also conceal a distinction they needed to make a sound decision. Underwriting and claims teams did not acquire their authority from our reference model, and our model could not replace their interpretation.

I became director of information integration in January 2016. This was a functional title, with no statutory directorship. The employer had 780 employees and ₹1,460 crore gross written premium for FY ending March 2017. My team grew to 31 within ₹5.7 crore annual operating expense. I was under pressure to make information easier to combine across business functions. The shared-contact proposal came from that ambition. When its weakness became clear, I authorised ₹62 lakh of correction and accepted a six-week delay. We withheld the automatic merge, kept the 230 different-party pairs separate and left 80 unresolved. The 610 supported matches entered the controlled stewardship route; the review did not certify every other customer record.

The CDO appointment began in November 2021. I now lead 64 people through five heads and report to Anil Qezvathri. My ₹14.2 crore annual expense supports the data function; it is not authority over premiums, payments or the employer's capital. The role includes executive presentations on data matters, while I remain outside the statutory board. Rina leads the stewardship work that makes a proposed link explainable. I want disagreement from that function to reach me before a delivery timetable turns an uncertain relationship into a settled assumption.

Subsequent cases widened the question. A financing-party label had been translated into an ownership category. Repeated partner submissions needed to be distinguished from separate notices about similar circumstances. A matching evaluation improved one error measure by making another worse. A proposed access connection needed evidence of the person's actual permitted role, even when the underlying identity link was sound. Each situation required a decision about the meaning attached to data, rather than another general declaration that the database should be clean.

I hold 0.15% of Vezqerthi through its employee vehicle. My fictional databank record shows registration in January 2026, with proficiency assessment still outstanding and no exemption asserted. Twenty-three days is a provisional annual allowance for a possible outside contribution, including preparation and exceptional demands. I work in English, Hindi and Khasi. A first board role would require learning to examine these relationships through collective oversight, rather than directing another company's data team. The useful part of my experience is not a claim that I always identify an unsafe assumption. It includes the occasion when I approved one, and the responsibility to change the recommendation when colleagues showed what I had missed.