The people a pooled interview count could not describe

The count in the 2019 report looked reassuringly large. It combined 2,400 interview records from repeated research waves, and I approved the description of those records as different people. Zoya Darshavi’s review identified 1,650 distinct respondents in the selected pool. Some people had answered in more than one wave. The records did not have to be false for my conclusion to be wrong; I had allowed the unit of observation to change between the data file and the client-facing claim.

I grew up in Shillong and completed a Bachelor of Science in Statistics at the fictional Velrashi Statistical College in May 1996. I joined Marvenli Survey Services the following month in a nine-person support group. We prepared survey records and resolved questions for senior researchers. I did not design sampling methods or approve client conclusions. The practical lesson was that a row needed a meaning before it could become part of a useful count.

In April 2003 I moved into sample operations, leading fourteen colleagues within ₹1.8 crore annual function expense. We worked with recruitment and response records across studies whose collection arrangements differed. A completed interview, a recruited panel member and a person in the population were not interchangeable categories. Keeping those distinctions visible was part of the work I supervised. That makes my later pooled-record mistake more serious, not less: the relevant principle was already familiar when I approved the broader claim.

I joined Lersavni Audience Research in April 2011 to lead research-data management with twenty-four colleagues and ₹4.2 crore annual function expense. Moving into a company that published continuing research products changed the responsibility. A definition could persist across many client presentations and become harder to question because the chart looked familiar. I worked with methodology and commercial colleagues to keep the study’s population, collection method and response unit available when data became a report. The challenge was to make those details usable rather than leave them in a technical appendix no decision-maker consulted.

I became data director in April 2017, leading thirty-two colleagues within ₹6.1 crore annual function expense. The pooled-interview error occurred two years later. We spent ₹39 lakh over eight weeks correcting the account and the approval process. The supported statement was 2,400 interview records from 1,650 distinct respondents in that pool. I did not turn the corrected number into a claim about population reach, independent observations or the quality of every estimate. The repair narrowed the assertion to what had actually been established.

I became CDO in April 2022. Forty colleagues now work through five leads within ₹7.8 crore annual function expense, including ₹2.2 crore improvement work. Nitin retains enterprise decisions and Zoya leads methodology. In 2023 we changed a study from telephone to web collection after eight telephone waves. A 120-person paired comparison produced ninety matching and thirty different answers on a selected item. We could not attribute every difference to mode or derive a universal adjustment. We used ₹58 lakh over twelve weeks to revise the product and begin a separately labelled series while keeping the comparability question visible.

In 2025 an online panel recruited across eighteen towns was being presented as coverage of thirty-two districts. Demographic weighting could adjust the observed sample under stated assumptions; it could not create observations in places outside the recruitment frame. We spent ₹62 lakh over eleven weeks restricting the proposition to the eighteen-town panel and explaining the selection limits that remained. The narrower description did not make the panel representative of every town resident. It stopped one unsupported geographic claim while preserving the research that could still be discussed honestly.

The researcher trust carries no personal interest for me, and my executive work does not include a statutory directorship. A first outside contribution would need more preparation and a specific assessment of fit. I would bring a habit of asking what changed between the source record and the commercial conclusion, and the experience of finding that change in an account I had personally approved. I would want the chair to hear both what a study can usefully contribute and what it leaves unanswered. The point is to improve the decision, not to reward the person who supplies the longest list of qualifications or the most visually confident chart.

Open to Board appointments · Executive opportunities

Languages / English · Khasi · Hindi