What a BI director actually decides

Most BI problems look technical. Most are decisions nobody made.

The gap is direction

Gartner surveyed 504 chief data and analytics officers¹. For 30%, the top challenge was measuring the impact of data, analytics and AI on business outcomes. Only 22% had defined, tracked and communicated business-impact metrics for most of their use cases.

That is not a tooling gap. It is a direction gap. Five decisions close it, and each one needs an owner, by name.

1. Platform: move, or stay

Microsoft and Salesforce both say Gartner named them Leaders in its 2026 Magic Quadrant for analytics and BI platforms². So "which tool is better?" is the wrong question. The right one is "which tool fits this company?"

The left column rests on Microsoft documentation and list prices. The right column rests on vendor blogs, so read it as claims³.

A move is seldom a copy. Microsoft says a one-to-one migration with no refactoring at all is rare⁴. The same series says this about little-used reports: "Sometimes, the cheapest and easiest thing to do is nothing"⁵. A director applies the same test to the platform, and keeps "stay" on the table.

2. Definitions: one KPI, one meaning, one owner

Every top KPI needs four things in writing: a definition, an owner, a source and a refresh time. Without them, a migration copies the chart and loses the meaning.

Kansas City's public 311 data shows how. Under the old software, 30.9% of PeopleSoft requests had run past their estimated timeframe⁶. The new system's public dataset, live since March 2021, has no such field. The late rate did not improve or worsen. It stopped being published.

Definitions also hide inside formulas. Whether revenue obeys the date slicer is a business rule. Tableau keeps that rule in a context filter; Power BI keeps it inside each measure⁷. Someone should decide it once, in writing.

3. Governance: who may change what

Governance is a list of owners, not a policy binder. Who certifies a data set? Who may publish? Who sees which rows?

The platform details bite. Power BI row-level security binds only the Viewer role; workspace Admins, Members and Contributors see every row⁸. A workspace role is a security decision, not an admin chore.

Both vendors publish a framework. Microsoft's Fabric adoption roadmap spans 12 areas, from data culture to change management⁹. Tableau Blueprint names three core capabilities, agility, proficiency and community, and calls governance central to self-service¹⁰. A neutral lead can run one agenda across both.

4. Adoption: measure use, retire the rest

A report nobody opens still costs money to refresh, secure and migrate. Adoption is measured, not announced.

One Microsoft customer tracked report visits, active reports and distinct users each month⁵. The target was an inverse curve: the new platform rising as the old one fell. The same company found that only about half of its thousands of reports had been opened in a year.

I worked on a Center of Excellence that trained 1,700 users¹¹. One lesson carries over. Training without a usage scorecard is hope.

5. Cost: price the real seat mix

License math is where boards get surprised. List prices, as checked on 17 October 2026:

Microsoft's prices come from its own pages. Tableau's come from the search index, because tableau.com refused direct reads¹².

The arithmetic changes decisions. Five hundred read-only viewers cost $7,000 a month on Pro, $7,500 as Tableau Cloud Standard Viewers and $17,500 on Enterprise¹³. A reserved F64 costs about as much as 357 Pro seats, and its viewers need no paid license. Every author who publishes still needs Pro.

Below F64, every viewer needs Pro or Premium Per User¹⁴. Microsoft adds that Premium Per User is the cheaper route below 250 users who need Premium features¹⁴. Pro rose from $10 to $14 on 1 April 2025, its first increase since launch¹⁵. A June 2026 comparison still quoted $10¹⁶. These are list prices, not quotes, and a parallel run pays both bills at once.

What this means for your BI team

• Name an owner for each decision. Platform, definitions, governance, adoption and cost each need one person and a review date.
• Price your real seat mix before the renewal date. After it, you negotiate from behind.
• Publish your top KPI definitions in the first month. Each gets an owner, a source and a refresh time.
• Review usage monthly. Retire what nobody opens.
• Expect an advisor to say "stay" sometimes. If the answer is always "move", it was never a question.

Three ways to work with me

Each engagement ends at a decision. The first two work in either direction, and I resell no licenses.

The Readout: two weeks, fixed scope

Should we move, what moves first, and what will it cost to run? I inventory reports, sources, calculations, users and security rules. Then I run the scans and price both platforms at your real seat mix.

You get a two-page executive summary, a technical appendix, a keep, merge or retire list, and a first wave plan. A 60-minute readout closes it. The answer can be "stay".

The Parity Pilot: four to six weeks

Can our hardest dashboards move without the numbers changing? We pick three: one high-value, one complex, one high-traffic.

Tolerance and exit criteria are signed before work starts, and the legacy reports stay live throughout. You get the migrated set in a test environment, a translation log, a parity report, rollback steps and a calibrated wave plan.

BI Direction: a fractional BI lead, monthly

Who keeps BI pointed at decisions after the move? The first 90 days follow a fixed arc, adapted from a published plan¹⁷.

• Weeks 1 to 3: trace the top KPIs to their sources.
• Weeks 4 to 6: publish definitions, each with an owner.
• Weeks 7 to 10: set governance, from ownership to security standards.
• Weeks 11 to 13: deliver a costed 12- to 24-month roadmap and a platform recommendation.

Then, each month: a one-page direction memo, an adoption scorecard, lint and drift scans, and license advice for both platforms.

Notes

1. The Register, Data is very valuable, just don't ask us to measure it, leaders say, 2025-02-21, on a Gartner survey of 504 chief data and analytics officers. Accessed 2026-10-17. ↩︎
2. Vendor statements about a paywalled report: Microsoft's Power BI page and Salesforce's analyst reports page. Both cite Gartner's Magic Quadrant for Analytics and Business Intelligence Platforms, 29 June 2026. Accessed 2026-10-17. ↩︎
3. Left: Microsoft Learn, migration overview, customer lessons and Fabric licenses; prices as in note 12. Right: Spike AI and NeenOpal. Accessed 2026-10-17; summary in creation-test/danieljin-bi/brief/research-market.md, section 8. ↩︎
4. Microsoft Learn, Gather requirements to migrate to Power BI, Stage 1, updated 2026-01-02, accessed 2026-10-17. ↩︎
5. Microsoft Learn, Learn from customer Power BI migrations, updated 2026-01-02, accessed 2026-10-17. ↩︎
6. Kansas City Open Data, audited in tableau/real-data-audit.md, claim 21 and its schema row: dataset d4px-6rwg has no equivalent field. Public data, not a client. Full story in What Kansas City 311 stopped measuring. ↩︎
7. Tableau Help, Level of Detail Expressions and Filters, accessed 2026-10-17. The mapping is in the translation table. ↩︎
8. Microsoft Learn, Row-level security with Power BI, ms.date 2026-05-13, accessed 2026-10-17. ↩︎
9. Microsoft Learn, Fabric adoption roadmap, accessed 2026-10-17. The 12 areas are listed in research-market.md, section 6. ↩︎
10. Tableau Blueprint, Core capabilities of data-driven organizations, accessed 2026-10-17. ↩︎
11. Career facts in creation-test/danieljin-bi/brief/content-inventory.md, section D. ↩︎
12. Power BI pricing; Microsoft Fabric pricing, Central US, USD; Tableau Cloud pricing via the search index. All 2026-10-17; detail in research-migration.md, section 5. ↩︎
13. Derived from the list prices above: 500 × $14, 500 × $15 and 500 × $35; $5,002.67 ÷ $14 is about 357. Arithmetic in research-migration.md, section 5d. ↩︎
14. Microsoft Learn, Microsoft Fabric licenses, updated 2026-08-05, accessed 2026-10-17. ↩︎
15. Microsoft Power BI blog, Important update to Microsoft Power BI pricing, effective 2025-04-01. Old prices from The Register, 2025-04-02. Both accessed 2026-10-17. ↩︎
16. Spike AI, Tableau vs Power BI, June 2026, accessed 2026-10-17; noted in research-market.md, section 7. ↩︎
17. The arc adapts KORE1's fractional CDO plan, accessed 2026-10-17. The areas follow the Fabric adoption roadmap and Tableau Blueprint. ↩︎