[Open Data] Fix ISF ratio scale, region coverage and culture handling, and keep retired-SKU reservations visible

.github/workflows/opendata-instance-size-flexibility.yml

@@ -94,9 +94,13 @@ jobs:
             echo "replacing the deprecated \`AutofitComboMeterData.csv\` and \`isfratioblob.csv\` files"
             echo "that were hosted on \`ccmstorageprod.blob.core.windows.net\` (retired 30 Aug 2026)."
             echo ""
+            echo "The dataset is additive: the Catalogs API only returns purchasable SKUs, so the"
+            echo "result is merged over the published file rather than replacing it. SKUs the API"
+            echo "no longer returns are retained."
+            echo ""
             echo "## Test plan"
             echo ""
-            echo "- [ ] Review the diff for unexpected changes (large drops in group/SKU count, etc.)"
+            echo "- [ ] Review the diff for unexpected changes (the dataset is additive, so expect additions and ratio updates only)"
             echo "- [ ] Spot-check a few SKUs against the Azure Pricing Calculator"
             echo ""
             echo "🤖 Generated by the **Update Instance Size Flexibility** workflow"

docs-mslearn/toolkit/changelog.md

@@ -78,6 +78,11 @@ The following section lists features and enhancements that are currently in deve
 
 - **Added**
   - Added a new [Instance size flexibility](open-data.md#instance-size-flexibility) dataset that maps each ARM SKU to its instance size flexibility group and ratio, sourced from the Azure Reservations Catalogs API. It replaces the deprecated ISF ratio files hosted on `ccmstorageprod.blob.core.windows.net` ([#2090](https://github.com/microsoft/finops-toolkit/issues/2090)).
+- **Fixed**
+  - Fixed the instance size flexibility dataset omitting SKUs that are no longer available to purchase. The Catalogs API only returns purchasable SKUs, so retired series that reservations still cover (Av2, D, DS, Dv2, Dv3, Ev3, F, G, H, LS, NC, NV, and others) were missing. The dataset is now additive and backfilled from the retired ratio file, growing from 211 to 318 flexibility groups and from 1,353 to 2,574 SKUs ([#2300](https://github.com/microsoft/finops-toolkit/issues/2300)).
+  - Fixed the generator sweeping a hardcoded list of 26 regions, which silently omitted SKUs that launch only in regions outside the list. It now enumerates every physical region from the ARM locations API, 63 at the time of the change ([#2300](https://github.com/microsoft/finops-toolkit/issues/2300)).
+  - Fixed the dataset publishing a SKU whose ratio is `0`, which the Optimization Engine's benefits simulation divides by. Non-positive ratios are now dropped ([#2300](https://github.com/microsoft/finops-toolkit/issues/2300)).
+  - Fixed ratios from the retired files sitting on a different scale than the Catalogs API. The retired files normalized each group so its smallest SKU was `1`; the API reports absolute vCPU counts. Because Azure retires individual sizes rather than whole families, the API covers only part of many groups, so combining the two left single groups holding both units. The backfilled rows were converted onto the API scale, and the generator now fails rather than publishing a group in two units ([#2300](https://github.com/microsoft/finops-toolkit/issues/2300)).
 
 **[Commitment discount eligibility](open-data.md#commitment-discount-eligibility)**
 

docs-mslearn/toolkit/open-data.md

@@ -4,7 +4,7 @@ description: Use open data to normalize and enhance your FinOps reporting, ensur
 ms.topic: concept-article
 author: flanakin
 ms.author: micflan
-ms.date: 09/01/2026
+ms.date: 09/08/2026
 ms.service: finops
 ms.subservice: finops-toolkit
 ms.reviewer: micflan
@@ -193,10 +193,11 @@ Sample data:
 
 A few important notes about the data:
 
-- It covers Virtual Machines, Redis Cache, and Dedicated Host.
+- It covers Virtual Machines, Redis Cache, and Dedicated Host across every physical Azure region.
 - `ArmSkuName` is unique across the file and can be used as a join key on its own.
-- Ratios are the raw Microsoft values within each group (the smallest SKU isn't always `1`).
-- Data is updated weekly via a GitHub Actions workflow and always reflects the current API catalog, so retired SKUs are removed automatically.
+- Ratios are the Microsoft values within each group (the smallest SKU isn't always `1`). Compare ratios only within a flexibility group.
+- The retired ratio files normalized each group so its smallest SKU was `1`, while the Catalogs API reports absolute vCPU counts. Both express the same proportions. Rows carried over from the retired files were converted onto the API scale, except for the fully retired groups that the API doesn't cover, which keep the normalized scale.
+- Data is updated weekly via a GitHub Actions workflow. The dataset is additive: the Catalogs API only returns SKUs you can still purchase, so SKUs it no longer returns are kept because reservations might still cover them.
 
 <!-- prettier-ignore-start -->
 > [!div class="nextstepaction"]