Twenty-five minutes with a dataset you have never seen is enough to know what it can support and what it cannot, provided you run the checks in a fixed order rather than browsing.
What are the six checks?
Shape comes first: rows, columns, units, period, currency, and what one row means. The Indian brief supplies 24 rows, six cities by two segments by two years, with five columns each. The American brief supplies eleven rows, one per terminal, with nine columns. Say out loud what a single row is: "one city's pharma capacity and occupancy in one year." A dataset whose row you cannot describe in a sentence will produce a chart nobody can read.
Ties asks whether the parts sum to the stated whole. Revenue built up city by city for Tarangini gives ₹35.2356 crore, and revenue built up segment by segment gives the same ₹35.2356 crore, so the two views are consistent and either can be used without contradiction appearing on a later slide. Shares sum to 100.00%. Occupied slots never exceed capacity in any of the twelve city-segment cells, which sounds trivial until a dataset fails it and you discover the capacity column is a year-end figure and the occupancy column is an average.
Advertisement
Coverage asks what is missing. The Indian dataset has no monthly detail, so any seasonality claim is unsupported. It has no waiting-list or turned-away-volume field, so the statement "pharma demand exists above 92% occupancy" cannot be made from this data. It has no customer-level detail, so concentration is invisible. Write those three absences down, because two of them will become risk lines on the recommendation page and one of them will become an information request.
Definitions asks what each column actually measures, and what its denominator is. The Tarangini rate column is revenue per occupied pallet slot per month, and the cost column has two parts with different denominators: a variable handling cost per occupied slot and a cost of capacity per capacity slot, occupied or not. Mixing those two denominators is the single error that produces the wrong recommendation, and it is easy to make, because the pack presents both as "₹ per slot per month."
Impossible and implausible values comes fifth, and nothing here is impossible, which is itself worth noting. In the American dataset the operating ratios run from 78.30 to 96.86 and the daily shipment counts from 380 to 1,450, both plausible for a regional carrier. What is implausible is the simple average of the eleven operating ratios, 90.0436, which is 2.25 points worse than the revenue-weighted figure of 87.7909 and describes a network that does not exist. An average of ratios across units of wildly different size is a number waiting to be quoted in a title.
The question this data cannot answer is the sixth check. State it explicitly, in a single sentence, and put it in the rationale. For Tarangini: the data cannot say whether pharma demand exists above 92% occupancy, which is the assumption the headline return rests on. For Harborline: the data can identify which terminals underperform their own density benchmark, and it cannot say why, which is the difference between a finding and a diagnosis.
The ties check pays for itself once more here, in a way nobody plans for. The Tarangini cover note asserts that food "is now a quarter of revenue". Recompute it: ₹6.1776 crore of ₹35.2356 crore is 17.5323%. The cover note is 7.47 percentage points high, which is 42.6% high in relative terms, and it is the sentence that has been shaping how the board talks about the business for two quarters. Finding it took ninety seconds. Reporting it, gently, in one line of the appendix, is worth more to that board than any exhibit in the deck.
What should the written page contain?
Twenty-five minutes ends with a written page: what one row is, what ties, what is missing, what each column means, what is implausible, what cannot be answered, and one correction. That page is the difference between analysis and browsing.