Compare two versions of the same year
A revision story should compare the same crop year across releases. Comparing a new forecast with last year’s outcome answers a different question. Here the August 20 and September 25 outlooks both describe 2026–27, so their difference isolates a change in the published scenario.
The interactive revision table keeps that comparison explicit. We subtract the August value from the September value, and divide by the August value for the percentage change. The calculation is ours; the inputs remain attributed to their government releases.
Reconcile the revision before interpreting it
Forecast supply rises by 625 thousand tonnes between the two releases. The disposition side also reconciles: domestic use increases by 450 thousand tonnes, exports decrease by 300 thousand, and closing stocks increase by 475 thousand. Those three changes sum to the same 625 thousand tonnes.
This check matters because a larger crop does not mechanically become a larger export number. The destination of the available material is part of the forecast. The bridge helps a reader distinguish an increase in availability from a change in how that availability is expected to be used.
The price row has its own boundary
The simple crop-year average price forecast moves from CAD720 to CAD775 per tonne. That is a change in a dated forecast specification, not a current bid, a futures settlement or a price available at a particular elevator. We keep that price out of the supply/use chart because dollars per tonne cannot share a tonnes axis.
For an actual marketing decision, a grower still needs the relevant delivery location, contract terms, quality settlement and current buyer information. Our comparison establishes what the published outlook changed. It does not establish what a grower could receive today.
A revision is a question generator
The useful follow-up is to watch whether later observed movement, processing and crop estimates support the scenario. Those series arrive on different calendars. A weekly movement observation cannot simply be substituted for a full-year forecast, and an expected release cannot be described as completed.
We preserve both vintages in the downloadable dataset. When the next verified outlook is available, the same method can extend the history. A sequence of revisions can become useful original research without inventing a proprietary measurement or obscuring where the numbers began.