A total can be calculated correctly from the wrong set of rows. A historical field can be accurate and still arrive too late for the prediction we are testing. I want the business meaning to survive the technical transformation, so a director can understand what the number counts before deciding what to do with it.
Selected works
Decisions in context.
I
Correcting the order amount repeated by a shipment join
In 2018 I approved a management extract showing ₹18 crore against a defined group of 240 orders.
The record must keep its meaning on the way to the decision.
A management extract I approved reported ₹18 crore against a defined group of orders whose actual header values totalled ₹12 crore. The query had attached shipment records to order headers and then summed the repeated header amounts. Eighty orders had two shipment rows; their values appeared twice. The source orders were intact, but our analysis had changed the unit of observation without changing the calculation.
I am CDO of Velqenri Industrial Spools, a fictional Hubballi manufacturer of reusable polymer spools for non-hazardous industrial cordage. The company employs 217 people and recorded ₹102 crore revenue and ₹8.2 crore operating profit in FY 2025–26. I lead fourteen data colleagues through two leads, with ₹3.1 crore annual expense and ₹1.2 crore separately approved capital. CEO Geeta Qernvashi retains enterprise authority, operations head Mohan Pelqeshi owns operating decisions and finance remains responsible for statutory reporting.
My work also examines time and identity. A delivery-risk prototype used thirteen fields that would not have been available when the proposed decision had to be made. A regional customer report added memberships as though each region contained different customers. In each case, the useful question came before the technical fix: what does this row represent, what could we know at that moment and which entity are we claiming to count?
Shipment detail repeated an order amount
A 240-order group became 320 joined rows. Eighty orders worth ₹6 crore appeared twice, and 160 worth ₹6 crore appeared once, producing ₹18 crore instead of ₹12 crore in the management extract.
A later fact was not an earlier predictor
Thirteen of 120 candidate fields were unavailable at the intended prediction point. We held the prototype and funded ₹54 lakh over ten weeks for point-in-time reconstruction and evaluation.
Regional memberships were not distinct customers
Two regions reported 120 and 90 active legal-customer identifiers with 30 in both. The combined population was 180, not 210; the overlap remained useful for regional work.
Current focus
I am making the unit, time and identity behind important reports explicit before they reach a management paper. The data team must show how a result changes when an order acquires more detail, when information arrives later or when the same customer appears in another region. These are business questions expressed through data work, and their answers need accountable owners as well as functioning queries.
Conversations I welcome
I would consider an industrial CDO mandate or a first compatible outside board contribution on the evidence behind management information. I have no shares or statutory seat. Twenty-four proposed annual days require employer agreement, further preparation and assessment of the actual appointment.
The leadership record
Trusted with consequential decisions.
Formation
Bachelor of Science in StatisticsZerqelni Statistical College · 1996
Parvez explained both the inflated extract and why its correction did not mean that the business had lost six crore in cash.
Geeta QernvashiChief Executive Officer, Velqenri Industrial SpoolsThe full record ↗
Contact
I would consider an industrial CDO mandate or a first compatible outside board contribution on the evidence behind management information. I have no shares or statutory seat. Twenty-four proposed annual days require employer agreement, further preparation and assessment of the actual appointment.
