It is almost always a definition problem.
Before blaming an integration or a report builder, take one number that disagrees and ask each system what it thinks it is counting. "Revenue this month" might mean invoiced, collected, or work performed. "Jobs completed" might include or exclude cancellations, warranty callbacks and jobs reopened after a complaint.
Both systems are usually right by their own definition. The disagreement is between the definitions, and it has been there since whoever configured each system made a reasonable choice without knowing the other one existed.
Write the definitions down before building anything.
For each measure that leadership actually uses: what it counts, what it excludes, which system is authoritative, at what moment it is counted, and who owns it. That document is boring and it is the entire project. Most reporting work that fails, fails because it skipped this and went straight to connecting data sources.
A useful test: give the definition to two people who have not discussed it and ask them to produce the number. If they get different answers, the definition is not finished.
Fewer measures, produced reliably.
A leadership team needs a small number of measures it trusts, not a dashboard with thirty charts. The second one produces a meeting where everybody finds the chart that supports what they already thought.
Choose the handful that answer whether the month went well. Fix where each one comes from. Then automate the production of them, so the number is produced rather than compiled by someone on the last Friday of the month.