## The arithmetic survived the wrong interpretation
The eighty bookings did not appear as late journeys because no journey had been completed. That was precisely why they mattered. In 2018 I approved a report showing approximately 96.1% on-time performance among 920 completed trips. It counted 884 on time and 36 late. The eligible booking population was one thousand, including eighty that were unserved. The arithmetic answered a question about completed trips. The presentation allowed that answer to describe the whole service commitment. I had accepted the narrower population without giving its boundary enough prominence in the conclusion.
The distinction can sound technical until someone asks what happened to the person whose booking never became a trip. That experience did not become less relevant because it lacked a completion record. Nor was it the same as arriving late. A report about the whole eligible cohort needed to retain three states: served on time, served late and unserved. Across that cohort, 884 on-time journeys represented 88.4%. The smaller percentage was not a rival calculation invented to make the organisation look worse. It answered the broader question that our headline had already encouraged the customer to ask.
I do not regard every completed-trip measure as defective. Operations may need to examine the journeys that actually occurred, and that subset can support useful questions about their timing. The error begins when the subset loses its name while its result keeps the authority of a whole-service statement. A measure can be accurate within its boundary and misleading outside it. I needed to make the boundary part of the communication, not leave it for a reader to discover in supporting material after accepting the headline.
Some account colleagues wanted to retain the familiar reporting basis. Their concern was understandable: customers had seen earlier reports in the same form, and a changed headline would require explanation. But consistency of presentation could not resolve the mismatch between the population counted and the population implied. Continuing the old account would make the comparison easier while preserving its weakness. I chose to reopen it, explain my own approval and show the full cohort. The responsibility belonged in the discussion alongside the revised arithmetic.
We spent ₹34 lakh rebuilding the reporting process and corrected the customer account. That expenditure is part of the case because clearer reporting required practical work, not simply a new sentence in a presentation. It is not a measured financial benefit or evidence of punctuality gained. The eighty bookings remained unserved in the historical cohort. The 36 late journeys remained late. No passenger's experience changed retrospectively when we chose a more complete denominator. I wanted the correction to be judged for what it did: make the service account more faithful to the commitment being described.
## Unserved and late are different experiences
The distinction also affected how the operating problem could be discussed. If unserved bookings were blended into lateness, colleagues might lose sight of the different questions each state required. If they were removed altogether, the report would conceal part of the demand the service had accepted. Keeping separate states did not prescribe an operational response. It gave the responsible team a clearer starting point for examining why the outcomes occurred and which actions were appropriate. Data leadership should support that examination without pretending that a classification has already solved the underlying problem.
There is a temptation to respond to a reporting failure by creating one supposedly perfect measure. I resist that conclusion. Different questions can require different populations, provided their boundaries are explicit and the measures are not exchanged casually. A completed-trip account and a full-booking account can coexist. Their relationship should be intelligible enough that a director, customer or operating colleague understands why both are present. The task is to prevent one from silently borrowing the meaning of the other, especially when the more flattering figure is easier to remember.
I also ask how a report changes the attention of the people reading it. A visible unserved category brings a neglected experience into the decision. That does not establish the reason for every unserved booking or tell us whether a particular intervention would work. It does make it harder to describe the service solely through the trips that succeeded in reaching completion. The value of the correction is therefore partly in the questions it allows colleagues to ask, rather than in a numerical improvement attributed to the data team.
## Correcting the account is not improving the service
For a leadership team, the uncomfortable part is acknowledging that a persuasive measure passed through competent hands. I had statistical training and years of operational analysis. Those qualifications did not prevent me from approving an account whose scope was too narrow for its message. The experience made me more attentive to the journey from a calculation to a claim. Review needs to examine the sentence a reader will take away, not only whether the numerator divides correctly by the denominator printed beside it.
The customer conversation needed equal care. We could explain the old basis without defending the broader interpretation it had invited. We could explain the new basis without suggesting that every previous number was fabricated. We could show the two views without claiming that the customer should choose whichever looked more reassuring. The purpose was to make the difference understandable and restore a common account of the cohort. That required plain language about completed, late and unserved bookings rather than a technical argument that made responsibility harder to locate.
I would bring the same question to an outside board: which experience has fallen outside the population behind this claim? That is an oversight question, not an instruction for directors to rebuild the dashboard themselves. Management should be able to explain the scope, the excluded states and the decision each measure supports. Directors can then judge whether the account matches the commitment being discussed. In a healthcare-adjacent transport business, they should also avoid turning an operational punctuality measure into a clinical claim it has never established.
A later report should also make its comparison basis understandable. If the population changes between periods, a reader needs to know whether an apparent movement reflects service performance, a changed definition or both. This case does not supply a later result, so I cannot claim a trend. It does explain why the corrected basis needs to remain visible in future discussions. Otherwise the organisation can repair one account and recreate the same ambiguity when the next summary is shortened for a different audience.
The enduring lesson is modest but demanding. A useful report must preserve the people and events relevant to the question, including those that never reached the convenient endpoint from which the data was assembled. My original approval is still part of the story because the later correction does not erase it. The full-cohort figure made our account less flattering and more complete. What happened next in the service required its own operational evidence. Keeping those two achievements separate is part of taking both reporting and the passenger experience seriously.