Forum Discussion
Impactful Individualtotals dont match
i removed the one FYTD table from the model. I converted floats to decimal (18,6). I moved some calc responsibility away from DAX into the sql views. I left some applied sort , rename and maybe dupe steps deleted. I removed one critical flag that was probably working but now cant even get to pbi thanks to a sql predicate blocking it.
All numbers match now not only to our warehouse $'s, but between the desktop and service as well. Hard to explain (rounding errors) to the user when the main focus was on making the report smarter on the 1st of the month. The report always tried to report thru "Yesterday". With previous authors so fixed on the Today() function, that created some challenges.
Impactful Individualhi all , i reset filters in the service for prod and today's FYTD yyy,yyy,700 didnt change.
i also dropped down bookmaks in the service and see neither personal ones nor "more" bookmarks.
i downloaded the prod semantic model and FYTD shows as yyy,yyy,700.
i refreshed the downloaded prod pbix and FYTD shows as yyy,yyy,753.
Todays prod Semantic model in the service refreshed at 7:36 am. Its 10:13 am right now.
i refreshed the semantic model in the service, then jumped right to the report and still see yyy,yyy,700.
i hit the refresh button in the service for prod left of comment and FYTD shows yyy,yyy,699 which to me is essentially the same as yyy,yyy,700.
i hit the refresh on the browser in the service for prod and FYTD remains yyy,yyy,699.
In sql, i see the right value should be yyy,yyy,683.
In the clone in the service i see yyy,yyy,698.
In the clone on the desktop after hitting refresh, I see yyy,yyy,753.
in an old ssrs report i see yyy,yyy,683.
While both prod and the clone seem to have an issue, possibly the same one, i'll focus on getting the clone straightened out. It does worry me that neither's values remain the same aftre refresh on the desktop.
i saw an insights button in the sevice and as i recall there is probably one on the desktop too. hopefully it will give me a count of records even though the # will be north of 1.6 million records.
Community SupportHi db042190
To address the inconsistent FYTD totals observed between Power BI Service, Desktop, and SQL, it is necessary to review the use of the float data type for currency values. Since floats are approximate, they can introduce rounding errors that accumulate during aggregation, especially in large datasets. This can cause totals to vary slightly with each refresh, even if the underlying data remains unchanged. Converting the float column to a fixed-precision type like decimal(18,6) at the SQL source or within Power BI Power Query will help ensure accurate and consistent calculations across all platforms.
Additionally, please review any Power Query steps that remove duplicates. If these are not based on stable business keys, they may result in inconsistent row counts depending on refresh timing or data order. Refining this logic will help prevent unintended data loss or duplication. Comparing record counts and analyzing totals by facility or customer can further pinpoint where discrepancies arise. By ensuring proper data type usage and transformation steps, FYTD totals will remain consistent and reliable across Power BI and SQL.
Regards,
Karpurapu D,
Microsoft Fabric Community Support Team.