Power BI report ADX -> Regions show incorrect position on map

[fajterini]

Hi, take this as question first, I created internal PoC for Finops hub v14 and was showing Power BI reports.
And received question why it shows that we have cost spending in China, when we cannot have anything there.
Then I noticed all of the location are in incorrect position.

🐛 Problem

PowerBi CostSummary.Report have tab, where users can see cost spending on map view. Problem is the point on map correctly don`t represent data.

👣 Repro steps

In PowerBi report go to regions tab

🤔 Expected

Regions (datacenter locations) should be correctly show with mirror error+/- 100-200 km in reality error is more than 1000 km in many cases.

📷 Screenshots

West Europe -> Netherlands -> But shown in eastern Europe -> Slovakia

Southeast Asia -> Singapore but shown in Thailand

East Asia -> Hong kong -> but shown in middle of China

🔧 Environment

• FinOps hub version: 14
• Billing account type: MCA
• Power BI report type: ADX (Azure Data Explorer)
• Power Bi report: Cost summary
• Cost Management export: FOCUS 1.2-preview

Thanks

[microsoft-github-policy-service]

Thank you for submitting this issue! We try to respond within 3 business days. In the meantime:

• 📖 Check the troubleshooting guide for common issues
• 💬 Ask questions in Discussions
• 📅 Join our biweekly office hours for live help from the team

[RolandKrummenacher]

Hi @fajterini, thanks for the detailed write-up and the clear screenshots — they made this easy to pin down.

What you're seeing is expected with how the map is currently built, though I agree it's confusing. The Regions page map plots each point using the friendly region name (for example "West Europe", "East Asia", "Southeast Asia") and lets the mapping service geocode that text. Those names resolve to the geographic center of the broad region they describe, not to the actual Azure datacenter city:

• "West Europe" → centroid of western/central Europe (lands around Slovakia/Hungary) instead of the Netherlands
• "East Asia" → centroid of the East Asia region (central China) instead of Hong Kong
• "Southeast Asia" → centroid of Southeast Asia (Thailand/Cambodia area) instead of Singapore

So the bubbles are off by hundreds to 1,000+ km, exactly as you measured. Importantly, only the dot position is wrong — the region labels and the cost values attached to each bubble (and everything in the table/totals) are correct. There is no spending actually being attributed to China; that bubble is "East Asia" cost drawn in the wrong spot.

This is a known limitation today. The documentation for the Regions page already calls it out: "The map shows approximate locations and isn't exact." That said, your examples are a good illustration that "approximate" is currently quite far off, so this is useful feedback.

In the meantime, if you need precise placement you can rely on the cost-by-region table/values rather than the bubble positions, which are accurate.

Nothing more is needed from you — your repro and screenshots are sufficient. Thanks again for flagging it!

[RolandKrummenacher]

🔍 Root-cause analysis (engineering)

Root cause: The CostSummary Regions map is an Azure Maps visual bound only by Category = Costs.RegionName, with no Latitude/Longitude projections. ADX-based report: src/power-bi/kql/CostSummary.Report/report.json:1212 ("visualType":"azureMap", "projections":{"Category":[{"queryRef":"Costs.RegionName"}],"Size":[Sum(Costs.EffectiveCost)]}). Azure Maps geocodes the free-text friendly name ("West Europe", "East Asia", "Southeast Asia") to the centroid of that named geographic region rather than to the Azure datacenter city, producing the 1000+ km offsets shown. No coordinate data exists in the model: the Regions open-data file has only OriginalValue, RegionId, RegionName (src/open-data/Regions.csv:1), the Regions table partition imports just those 3 columns (src/power-bi/kql/Shared.Dataset/definition/tables/Regions.tmdl:4-36), and the RegionName column carries no dataCategory (e.g. Place) annotation to improve geocoding (src/power-bi/kql/Shared.Dataset/definition/tables/Costs.tmdl:461-468). The "latitude/longitude" hits in en-US.tmdl are Q&A linguistic synonyms, not real columns (confirmed none are projected by the visual).

Scope / impact: Affects both report variants, not just ADX: the storage-based report uses the identical pattern (src/power-bi/storage/CostSummary.Report/report.json:2273 azureMap, Category=Costs.RegionName). Independent of billing-account type / export type (FOCUS 1.2-preview here, but cause is purely visual binding). Cosmetic only — labels, bubble sizing, table values and totals are all correct; no cost is misattributed numerically. Present in v14 and prior. Documented as a known limitation: "The map shows approximate locations and isn't exact." (docs-mslearn/toolkit/power-bi/cost-summary.md:180).

Fix options:

1. Add an authoritative region→lat/long mapping (extend Regions.csv with Latitude/Longitude columns, or ship a companion open-data file) and rebind the map to Latitude/Longitude projections — most accurate; requires maintaining coordinates for ~101 distinct RegionName values and bumping the published open-data asset both reports download.
2. Set the RegionName column dataCategory to Place/StateOrProvince to give the geocoder a stronger hint — low effort, but still relies on Azure Maps text geocoding and will not reliably land on datacenter cities; likely marginal improvement only.
3. Do nothing / document only — it is already documented as approximate; zero effort but the discrepancy (cost "in China") keeps surprising users and erodes trust.

Recommendation: Treat as a low-priority quality fix; prefer option 1 (coordinate lookup + Lat/Long binding) since it is the only option that actually places bubbles on the datacenters and would fix both report variants. Acceptable to keep as a documented known limitation short-term.

[fajterini]

Hi @RolandKrummenacher

Thanks for reply and good RCA.
I had the same idea, to extend regiosn.csv with RegionLatitude, RegionLongitude, but it's not so simple as I was thinking ...
For someone new to Datafactory an ADX, but i like the idea to continue with same naming's standard, is it shoes data source, in this case ResourceLocation, RegionName, Region ID, Region Latitude, ...

Thanks