Inventories: the number that resolves the argument

Supply estimates argue. Demand estimates argue. Inventories are where the argument ends, because stocks are what physically remains when both sides have had their say.

The verdict property

Every balance-sheet claim implies a stocks path: if supply is really this and demand really that, storage must fill or drain at a computable rate. When the weekly count arrives and stocks did something else, one of the estimates was wrong - and the market reprices toward whichever side the physical evidence indicts. This is what module 2 meant by calling inventory data direct evidence: it is the only line of the table that gets audited against reality, every week, in public.

Level, trend and the seasonal normal

• The level says little alone: module 2's lesson. Gas storage is always high in November - that is the calendar, not information.
• The deviation from seasonal-normal is the signal: stocks two standard deviations below their usual November is a market genuinely tighter than its history, whatever the raw number reads.
• The trend of the deviation is the sharpest read of all: a gap versus normal that widens for six consecutive weeks is a balance sheet being proven wrong in the same direction, repeatedly - the closest thing commodity analysis has to a trustworthy trend.

Using the verdict prospectively

The analytical move is to run the identity forward: take the consensus balance, compute the stocks path it implies, and watch the weekly counts against that path. Each release becomes a hypothesis test you set up in advance - agreement is the consensus surviving, divergence is your early notice that the table is wrong, often weeks before the agency revises it. That converts lesson 1's demand for testable views into a routine: the view names its stocks path, and the calendar audits it.