Analytics Platform Cost & Architecture Model

Original work

The question

A BI license line item is easy to put in a spreadsheet. It is also incomplete. Storage, pipelines, warehouse or lakehouse compute, capacity, creators, self-service analysts, viewers, training, and governance all move the total. Comparing tools only on seat price answers the wrong question.

Licenses are one cost and architecture component. They are not the platform. The original evaluation asked how Power BI and Tableau compare on cost, ecosystem fit, adoption, visualization, modeling, governance, scale, AI roadmap, and implementation approach for an enterprise healthcare analytics audience.

Modernized

What changed

License-only models understate platform cost and hide architecture. Fabric attribution matters: one capacity can cover DE and BI, so allocation labels must not double-count. F64 changes viewer economics. Non-cost criteria from the original comparison still decide close races once hard constraints pass.

1. Power BI Premium capacity is no longer the correct forward-looking capacity assumption; Fabric F SKUs are the capacity path.
2. Fabric is not just a BI feature. It adds shared compute, OneLake storage, pipelines, warehouse/lakehouse, and other DE workloads.
3. F64 is a licensing breakpoint because Free viewers can consume Power BI content at F64+.
4. DE compute/storage/network and BI seat licensing are modeled separately, then recombined into total TCO without double-counting shared capacity.
5. CU, Snowflake credit, and Databricks DBU stay distinct; workload calibration is required.

Modernized / New extension

Architecture model

Data engineering (ingest, transform, storage, warehouse or lakehouse) plus business intelligence (semantic models, authors, self-service analysts, viewers, capacity). Total analytics platform cost is both layers together, for a normalized workload, not a one-to-one conversion of CU to credit to DBU.

• DE options: Microsoft Fabric, Snowflake, Databricks.
• BI options: Power BI, Tableau.
• Combinations: six default pairings including Fabric + Tableau. Affinity is a transparent criterion, not a hidden preference.

New extension

Recommendation

Transparent multi-criteria analysis: hard constraints first, then weighted scores, then soft penalties. If nothing is feasible, the page says so. This panel is a teaching and decision artifact, not a republished private client recommendation.

What this demonstrates

• Financial modeling — public list rates, native billing units, and inspectable cost lines instead of a black-box total.
• Systems thinking — data engineering and BI priced as one platform, with shared Fabric capacity attributed without inventing two capacities.
• Decision design — priorities (what matters most) separated from constraints (what cannot be compromised), with scenario-specific overrides when local evidence differs.
• Communication — a recommendation you can audit, including what would change it.

Financial modeling that keeps units honest, systems thinking across DE and BI, decision design that separates priorities from hard constraints, and communication that leaves the math inspectable.

The artifact is the model. No client savings, adoption metrics, or negotiated discount is claimed here.

How the model works

Reference material. The interactive model above is the primary artifact.

Sources and methodology

Centralized config distinguishes list price, estimate, and negotiated (negotiated is never defaulted here).

1. Microsoft: Power BI pricing
2. Microsoft: Microsoft Fabric Azure pricing
3. Microsoft: Fabric licenses and capacity
4. Microsoft: OneLake storage tiers
5. Tableau: Tableau Cloud pricing
6. Snowflake: Snowflake service consumption table
7. Databricks: Databricks product pricing