Chief Data Officer India

An orderly report was not enough to authorise the conclusion

Statistics and operating-data experience: 32 years; Career begins July 1994
Data colleagues: 24; Led through four leads
Annual data-function expense: ₹5.8 crore; ₹1.8 crore approved capital expenditure

Decisions & consequences

Correcting the unit before interpreting the inventory report

A review of seventy-two linen stock codes found eighteen bundle-to-piece mapping errors in the reporting path. The report presented an item count, but some source definitions referred to bundles. The discrepancy could distort a comparison without any corresponding physical movement. I asked operating and finance owners to establish what each unit meant before the team explained the difference as a stock outcome.

The reporting team proposed a quick multiplier adjustment so that the dashboard could continue on schedule. Operations wanted the catalogue definitions examined because the same apparent bundle did not necessarily represent the same reporting purpose everywhere. I supported the wider review and funded ₹62 lakh of definition and mapping work. We accepted additional work on the catalogue and affected comparisons rather than hide the problem inside a correction whose meaning few readers would understand.

The eighteen identified mappings were corrected with the relevant owners' agreement. The result concerned reporting relationships, not a finding that eighteen stock lines had suffered physical loss or that the business had created new inventory. We retained the explanation of the change so that historical comparisons could be interpreted with the appropriate caution. A more accurate current number should not make a previous definition change invisible.

The decision reinforced a distinction that matters in operational data. A unit is not merely a label added after calculation; it is part of the meaning of the observation. If the unit changes between collection and reporting, the resulting figure can be mathematically consistent and operationally misleading. I would ask a board considering a large inventory variance whether management has separated physical movement, valuation and definition changes. The answer should determine the response, rather than allow the most alarming or convenient interpretation to become the story before the underlying record has been understood.

Holding twenty-seven allocations outside the billing decision

Among 320 processing bags reviewed, twenty-seven did not have dependable customer identifiers. Their processing events were recorded, but the evidence needed to allocate them to a customer was incomplete. A billing report expected a populated field, and the temptation was to use the most likely customer based on nearby activity. I did not consider that an adequate basis for a financial record.

The commercial team wanted to close the reporting period using provisional allocations and adjust later if a customer queried them. Finance argued that the evidence should be resolved before the allocation entered billing. I supported finance and held those twenty-seven allocations for review. A ₹41 lakh programme improved the records and handoff that supplied the customer identifiers, while the relevant teams examined the known cases.

Finance accepted the resolved allocations after review. That outcome applied to the twenty-seven identified bags within the 320-bag population. It did not prove that every allocation elsewhere was correct or that no later question could arise. We also kept the processing event separate from its financial use: the service record existed, but it was not yet sufficient for the proposed customer charge. The decision accepted a delay in completing those billing allocations instead of treating a full field as the primary objective.

I learned to make the unknown state useful. A blank or disputed field can be uncomfortable in a report, but it may be the most accurate description of what the organisation currently knows. The data process should help an accountable owner resolve it, rather than reward the quickest plausible entry. Directors can ask how management handles that uncertainty when it affects money. The quality of the answer depends on the evidence and authority behind the allocation, not on whether the final table contains a value in every cell.

Keeping the raw times while the sequence was reviewed

We reviewed ninety processing batches and found fourteen timestamp discrepancies associated with different time settings. The displayed sequence could appear inconsistent even though the underlying records needed a more careful interpretation. My team could identify the discrepancy, but it could not decide the technical significance for processing or release by changing the display alone.

The systems team proposed normalising the stored times in place to simplify future reporting. Quality head Nadia Qezhari wanted the original entries preserved so that the interpretation could be examined. I supported retaining the raw records and recording the corrected representation separately. We funded ₹87 lakh of configuration and record-review work, accepting the additional complexity of an explicit history rather than overwriting the evidence that had prompted the question.

The fourteen identified discrepancies were reviewed with their source settings and the interpretation recorded. Quality retained assessment and release authority throughout. The result did not establish sterility, infection prevention or a clinical outcome; it clarified a data representation in a defined review. A tidy sequence could not be used as a substitute for the professional examination of what the processing record meant.

The case made me more cautious about the word 'clean' when it is applied to data. A record can look cleaner after its awkward history has been removed, while becoming less useful to somebody investigating the event. The appropriate response should preserve the facts needed to understand the correction and its limits. I would want directors to know whether management can explain what changed, who accepted the interpretation and which technical questions remain outside the data team's authority. That account gives a correction meaning without allowing a revised timestamp to become an unsupported assurance about the service itself.

Postponing a ranking until eleven sites counted the same activity

A proposed linen-consumption comparison covered eleven contract sites and used occupied bed days as its denominator. Four sites had applied definitions that were inconsistent with the others. The resulting ranking looked precise, but it could not support the efficiency conclusion being proposed without further work. I asked the team to examine the basis of the denominator before placing the sites in an order.

The contract team wanted to publish the ranking with a short note about variation so that the performance discussion could proceed. I argued that the definition difference affected the central comparison, not a minor qualification. We postponed the ranking and funded ₹24 lakh of definition work with contract and operating owners. The decision delayed a visible management output and required agreement across parties who collected activity for different purposes.

All eleven sites subsequently used an agreed basis for that comparison. This made the denominator more consistent. It did not reduce linen consumption by itself or establish that every remaining difference was caused by efficiency. Service mix and other operating conditions could still require examination before action. The revised measure was a better starting point for a decision, not the decision's completed outcome.

My lesson was that a ranking can outrun the evidence because it supplies an immediately understandable order. A director can see who appears first and last without seeing the assumptions that created those positions. The data team's responsibility is to make the basis clear enough that the order is not mistaken for an explanation. I would ask management which differences have been controlled, which remain and what conclusion the comparison can actually support. Sometimes the responsible next step is to improve a definition before asking a manager to explain a performance gap that may partly exist only in the way the activity was counted.

Discuss a leadership challenge