Power BI and Tableau Reporting on Reconciled Data

A dashboard earns its cost when somebody moves budget because of it. Every page we ship carries one decision, one owner and one action, and no page goes live until its totals reconcile to the platform interface within 2 percent across the trailing 30 days, with whatever remains explained in a note printed on the page itself rather than in a document nobody opens. How the build runs: The semantic model comes first and the visuals come last. In Power BI we work in the .pbip project format, so the model lives in git and a change arrives as a diff a colleague can review instead of a binary file emailed around. Measures are written once in DAX under a naming rule that puts the variation last, so a reader can guess a measure name before opening the field list. Tabular Editor handles model edits. DAX Studio's VertiPaq analyzer tells us which column is eating the memory, and it is nearly always a raw creative identifier that should have been hashed upstream. Incremental refresh on facts, full refresh on dimensions, row level security bound to Entra ID groups instead of a text column an operator can retype. When Tableau is the right answer: We use Tableau where the client already runs it and the analysts are fluent, and then we publish the data source separately from the workbooks so one metric definition cannot fork into three versions across three teams. Extracts over live connections to BigQuery in almost every case. If you are choosing between the two tools with no incumbent, we will say which one fits your refresh pattern and licence position, and the answer is not always Power BI. Currency, because we work from Istanbul: Clients reporting in both lira and euro get one written FX policy applied inside the model: the TCMB buying rate on the date of spend, never the rate at refresh time. Without that rule a Turkish account's cost per acquisition appears to improve month over month when the only thing that changed was the currency, and someone reallocates budget on the strength of it. What we will not build: The forty tile screen a leadership team asks for after seeing one at a conference. Real time refresh for a decision made on Mondays. Any page sitting on data that has never been reconciled against an invoice, because a fast wrong number travels further than a slow right one. For health and wellness clients we will not model user level rows into a reporting layer at all; the grain stops at cohort and campaign, and we explain that limit to your growth team rather than quietly enforcing it. Who this is not for: Teams who want the dashboard before the warehouse. Teams who want figures to agree with a slide already written. Anyone whose real requirement is a monthly deck, in which case buy the deck. We also decline work where nobody will be named as the owner of a page, because an unowned page is dead within two quarters. What handover looks like: A measure dictionary defining each metric in a sentence a finance director would accept, the reconciliation log showing every gap we found and its cause, the repository holding the model, and two recorded working sessions with your analysts. After that your team edits the model. We are available for a follow up quarter at day rate, and most clients do not take it.