Chief Data Officer / India
The report changed when its claims were examined
- Data and research through four leads
- 29 staff
- ₹5.8 crore FY 27 expense; separate ₹2.1 crore programme
- Field studies and service evidence
- 29 years
- Career since July 1997; CDO since March 2022
- Provisional external-board capacity
- 22 days
- Preparation, consent and actual calendar remain to assess
Decisions & consequences
01Withdrawing my completion label when the trace showed only presence
In 2022 I endorsed a dashboard rule that treated a vehicle entering a defined proximity around a scheduled collection point as evidence of completed collection. The route trace was available and consistent enough to make the account look dependable. The reporting sponsor wanted to replace a slower manual process, and I supported the convenience without adequately testing the service meaning attached to the location event. Nusrat challenged that interpretation before it became the company's standard completion account.
A selected review examined 160 route events with supporting field evidence. It supported observed collection for 112, passage without collection for 28 and an unresolved account for 20. Those were classifications within the review, not an estimate of service performance across all seven towns. A vehicle might pass near a point for several reasons. The presence trace alone did not explain which reason applied, and lack of corroboration did not establish misconduct or a missed collection in every unresolved case.
We spent ₹46 lakh on the review and status redesign. I withdrew my endorsement of the automatic completion label. The revised account retained proximity as route evidence and separated it from the collection status supported by the review. Operations and field assurance owned the further investigation of service events. Technology colleagues implemented the distinction; worker-safety, contractual and disciplinary decisions remained outside the data rule. We did not replace an overconfident positive label with an equally automatic accusation.
The choice gave up a simple, rapidly populated completion view and left more work for the people responsible for resolving uncertain events. The alternative was to retain the original rule and treat exceptions as minor imperfections. I rejected that approach because the label claimed a different event from the one the system observed. The result was a more precise report, not proof that collection improved or that every location trace was accurate. I would now agree the meaning and evidence requirements of a service status before making its automatic population the main success criterion.
Keeping both improving area rates beside a falling combined rate
A 2023 service review compared two defined observation periods across core and extension areas. In the first period, the account contained 900 recorded completions from 1,000 scheduled core-area point-visits and 100 from 200 extension-area point-visits. In the second, it contained 460 from 500 and 360 from 600 respectively. Core-area completion rose from 90% to 92%, and extension-area completion from 50% to 60%. The pooled rate nevertheless fell from 1,000 of 1,200, or 83.3%, to 820 of 1,100, or 74.5%.
The reporting sponsor wanted to use the combined decline as evidence that the operating teams had deteriorated. Area managers preferred to show only the two improved rates. Neither presentation answered the whole question. The observed population had shifted toward extension-area visits, whose recorded completion remained lower. The overall rate accurately described the combined account, while the group rates described a different comparison. Changing the mix made it possible for those directions to differ without either calculation being false.
I authorised ₹34 lakh of reporting and definition work to retain all four numerators and denominators, the group descriptions and the pooled account together. We removed the unsupported team-deterioration interpretation without hiding the lower overall recorded completion. Operations investigated the practical reasons for the changing schedules and completion patterns. These were point-visits in defined records, not unique households, a town-wide coverage estimate or a controlled test of workforce performance. We did not create a preferred standardised rate and present it as an observed result.
The decision cost us the convenience of one unqualified performance arrow. It also required managers to discuss the extension area's lower level even while acknowledging its improvement. The result was a clearer basis for scrutiny, not a causal explanation of why either group changed. I would now agree the population and comparison question before selecting the headline. A total should remain visible where it describes the service received, but it cannot be used casually to assign causes that belong to a different level of analysis.
Retaining a weight increase without inventing a generation claim
In 2024 two defined four-week accounts from the same selected route set recorded 900 tonnes and 1,020 tonnes received at the designated facility. Each period covered twenty-two scheduled collection days. A management draft described the 120-tonne increase, approximately 13.3%, as evidence that households were generating more waste. The weights were recorded on the stated net-received basis, but that did not establish the proposed upstream explanation. The route set and scheduled days were only part of the conditions needed for a meaningful comparison.
Nusrat asked us to retain the collection coverage, actual movements, material mix and weather context alongside the receipts. The commercial sponsor argued that the larger total was useful for planning service capacity and should not be delayed by an academic qualification. I agreed that the received load mattered operationally. I did not agree that it proved a change in household generation, behaviour or the effect of a public programme. Moisture and collection conditions were possible influences to examine, not explanations we could simply insert because they seemed plausible.
We spent ₹41 lakh on the evidence review and revised the report to describe the two received-weight accounts with their periods and limits. Qualified operating and environmental colleagues retained physical interpretation and any further sampling method. We did not calculate an invented dry-weight correction, infer material diversion or assert that rain explained the difference. The available evidence left the cause unresolved. Capacity decisions could still consider what the facility actually received, with the separate question about generation identified as requiring different support.
The choice gave up a clear behavioural story that the initial draft could not substantiate. It preserved the increase instead of making uncertainty a reason to suppress an inconvenient number. The outcome was a better-defined account and a plan for more comparable observation, not a reduction in waste or a verified household trend. I would specify the inference sought when commissioning the next weight review. A number can be important for one operating question while remaining insufficient for the broader claim somebody wants to build around it.
Including streets that the complaint register barely described
A 2025 evidence programme began with a sixty-street frame. The recorded contact history was uneven: twenty streets formed a higher-contact group, twenty a lower-contact group and twenty had no recorded resident contacts during the defined review period. The service sponsor wanted field researchers to concentrate on the busiest complaint locations. That would have followed visible demand for attention, but it would also have left us unable to learn whether the quiet streets had satisfactory service or simply less access to the reporting route.
I proposed twelve street reviews, with four selected at random within each of the three predefined groups. The selection was an evidence programme, not a rule for responding to current complaints. Urgent operating concerns continued through their established owners. The sponsor reasonably worried that spreading research would reduce depth at the locations already asking for help. I retained that trade-off and separated research coverage from frontline response priorities, rather than suggest the data team could decide which residents deserved service.
The programme received ₹76 lakh authority. At the review cutoff, ten street reviews were completed and two remained postponed because access arrangements were unfinished. ₹61 lakh had been spent and ₹15 lakh remained unspent. We kept the two selected locations in the record instead of quietly replacing them with easier streets. The incomplete work did not support a finished comparison across the groups or a population estimate for all sixty streets. No result about the quality of service on the unvisited streets was invented.
We gave up the speed and apparent certainty of studying only the locations already visible in the complaint account. We also accepted that a broader design could remain unfinished at a management reporting date. My contribution was to make the observation question and its coverage explicit, while preserving operating responsibility for live concerns. I would now examine channel access when setting the study frame, before contact volume becomes the sole explanation of where problems exist. Silence is a reason to ask how evidence is obtained, not a satisfactory-service response supplied by default.