Learning what a complete table can conceal

I began work at Vesquli Information Services in Indore in June 2000 after completing a statistics degree. The company prepared business-information studies for regional clients. My first tasks were to check records, reconcile source references and assemble tables under a research manager's supervision. I enjoyed the apparent precision of the work, but I soon saw that a neat table could answer a different question from the one its reader thought it answered. A missing company was sometimes treated as if it had never belonged to the study. A category could mean one thing to the person collecting data and something else to the person writing the summary.

In April 2006 I joined Dalqeri Credit Research as a data-methods analyst. The business supplied research information to corporate credit teams. My role was to examine how source material became a usable dataset, including the definitions applied by researchers. I did not assess a client's lending decision. I worked with the research team to document categories and retain source references so that a conclusion could be checked. The appointment taught me that a method must be understandable to the colleagues applying it every day, not only to the analyst who designed the first version.

I became dataset manager in April 2012. Eleven colleagues worked with me on company-information collections and release checks. The role added responsibility for delivery dates and for the effect that a methodological change could have on existing users. I could not simply improve a definition and assume that every earlier comparison remained valid. We began documenting changes alongside the release and explaining where a revised category should not be compared directly with its predecessor. I also spent more time with product colleagues, learning why a coverage figure mattered commercially and why that pressure needed an evidence-based response rather than a general objection from the data team.

In April 2018 I became head of data stewardship. The company had more sources, more products and several different understandings of what a source agreement permitted. I worked with legal and product colleagues to connect the approved use of material with the outputs in which it appeared. Legal specialists interpreted the terms; my team made the resulting conditions visible in the data process. We also reviewed the samples used to evaluate research classifications. Disagreement between experienced reviewers could reveal an unclear question rather than poor performance by one of them. That distinction changed how we trained colleagues and how we reported quality.

I took the CDO appointment in April 2023. I now report to chief executive Joseph Darquli and work closely with head of research Manisha Pelurani. Thirty-four data colleagues report through four managers, supported by a ₹7.6 crore function budget. The company has 269 employees and serves 48 client teams; those figures describe its scale rather than my personal authority over client decisions. I attend board discussions on data-dependent product decisions when invited. I have not served as a director and do not hold membership of a board committee.

The cases on this site show the responsibilities that came together in the role. A proposed external product relied on source material whose permitted use had been assumed rather than checked against the actual distribution. A single classification label hid several different judgements about what a text said. A growth comparison selected only businesses still active and therefore omitted part of the original experience. Each issue could survive a conventional check that the file was complete and the calculation ran. Resolving it required product, research and qualified legal input as well as data work. We accepted smaller coverage and additional review where the evidence required it.

My leadership has moved from trying to remove ambiguity myself to helping the organisation decide how ambiguity should be handled. A CDO cannot personally arbitrate every source or every classification. The team needs clear responsibilities, a route for unresolved questions and enough confidence to delay a release when its meaning is uncertain. I also want us to recognise when a limited result is still useful. The objective is not to produce a dataset without caveats, which is rarely a realistic ambition. It is to ensure that the caveats describe the question the client is actually asking, and that they remain available when the result leaves our own research environment.

Open to Board appointments · Executive opportunities

Languages / English · Hindi · Urdu