### Two exact answers with different meanings
An executive can approve a figure after checking every source value and still approve the wrong account. I did so when our fictional research publisher described an equal-contract mean as the average across a combined purchased volume. The arithmetic looked ordinary, which made the interpretation easier to overlook. Two rates appeared in the source record and the report needed one summary. We added the rates and divided by two. That sequence answered a question, but the published label promised an answer to another one. The mistake was mine to explain.
The example involved one MWh at ₹4,000 per MWh and nine MWh at ₹6,000 per MWh. Both were invented energy-only contracts for the same study period and on the same stated basis. An equal-contract mean gave ₹5,000 per MWh. It gave each contract the same influence, regardless of its volume. That can be a legitimate description if the intended question is the average of those two contract rates under equal treatment. Calling it simple does not make it false. The difficulty arose because our report described the experience of the combined volume instead.
For that pooled-volume question, the first contract contributed ₹4,000 of expenditure and the second ₹54,000. Together they accounted for ₹58,000 across ten MWh, or ₹5,800 per MWh. Nitin's challenge was not that we had copied a rate badly. He asked me to follow what each rate represented in the summary. The resulting difference exposed a decision about influence that our label had hidden. A reader seeing only the word average had no way to recover the intended unit or recognise that we had substituted one question for another.
It would be convenient to turn the correction into a general instruction to weight every price by volume. That would repeat the original mistake at a larger scale. The correct method depends on the research purpose, the population being described and the comparability of the observations. A study of contract offers might deliberately give contracts equal influence. A study of pooled expenditure per unit asks something else. Neither purpose becomes legitimate solely because its calculation is more complicated. The method needs to follow a stated question and appropriate specialist assessment, rather than acquire authority from its appearance.
### The label is part of the calculation
A short label often travels farther than the analysis supporting it. It appears in a heading, a board paper or a subscriber's presentation, while the analyst's qualifications remain elsewhere. That makes the language attached to a result part of the research work. A number whose meaning survives only when its author is present is not ready for every use its format may invite. We should ask what a reader can reasonably infer from the published account, including the population, period and basis that the summary actually covers.
In our correction, the common period and energy-only basis mattered. Without them, the numerical illustration could imply a broader comparison than the records supported. Neither result described an all-in bill, a real current tariff or a recommendation to purchase electricity. The ₹800 difference between the two means was not a loss suffered by a customer. It was the difference between answers to two differently weighted questions in an invented study. Converting that difference into a commercial harm figure would have added a second unsupported interpretation while claiming to repair the first.
An editor reasonably worried that the correction would make accurate source prices look unreliable. We therefore retained the rates and quantities and explained where the description had gone wrong. The original equal-contract result remained intelligible under its own definition. The revised publication identified the pooled-volume purpose and the calculation serving it. Review and correction expenditure was ₹43 lakh, a cost of our response. We did not present that expenditure as a return achieved or as evidence that a customer's actual decision had improved. The record was narrower and more useful for being so.
This experience changed how I listen when someone says a figure is merely descriptive. Description still selects a unit, a population and a relationship among observations. A summary may give equal influence to firms, contracts, units or some other relevant element. Each choice has consequences for interpretation even before anybody makes a prediction. I want those choices discussed by the people responsible for the research. The data function can help preserve them and expose inconsistencies, but it should not quietly assume methodological authority because it operates the calculation or manages the source files.
### What an approval should preserve
Before a concise result is approved, I ask for an account of what one observation means and why it receives the influence assigned to it. I also want the claim written in ordinary language. The point is not to create a lengthy ritual around every spreadsheet. It is to make an important interpretive decision visible while it can still be challenged. If the sentence and the calculation describe different units, an accurate output does not resolve the disagreement. The approval needs to establish which question the publication intends to answer.
The supporting account should retain enough detail for another qualified person to reproduce the result and examine its meaning. That includes the relevant quantities, definitions and study scope, together with any exclusions that affect the interpretation. Reproducibility alone is insufficient: several people can reproduce the same mislabelled mean. But without recoverable support, an apparent correction may become another figure whose authority depends on trust in its author. The useful combination is a calculation that can be followed and a purpose that can be debated without reconstructing the original meeting from memory.
For a leadership reader, the strongest question may be surprisingly plain: what does this average describe? That question should not be treated as evidence that the reader lacks technical sophistication. It can reveal a hidden shift between an individual unit and a pooled total, or between the respondents observed and a larger population implied. A specialist can then explain whether the method serves the intended purpose. Directors and executives need not perform every calculation themselves to require an account in which the claim, method and supporting population remain aligned.
I keep my error in the profile because it prevents the later discipline from sounding like an abstract preference I always possessed. I approved the description that needed correction. The remedy was not to celebrate a more impressive formula, nor to insist that the larger result was more truthful simply because it was larger. We restored the relationship between the research question and the method. That is the contribution I now expect from data leadership: helping an organisation preserve what its figures mean when they leave the people who produced them and enter decisions those people do not control.