Analytics Migration: Looker to Power BI¶

A hands-on guide for BI developers and analytics engineers migrating from Looker (LookML, Explores, dashboards, embedding) to Power BI, DAX, semantic models, and Power BI Embedded.

Scope¶

This guide covers:

• LookML model to Power BI semantic model
• Explores to Power BI report pages
• Looks and dashboards to Power BI visuals and dashboards
• LookML dimensions and measures to DAX
• Looker embedding to Power BI Embedded
• Looker Studio to Power BI Desktop/Service
• Data modeling philosophy comparison

For BigQuery compute migration, see Compute Migration. For the worked example of a Looker explore to Power BI semantic model, see the Migration Playbook Section 4.7.

Architecture overview¶

flowchart LR
    subgraph Looker["Looker Stack"]
        LM[LookML Model]
        EX[Explores]
        LK[Looks + Dashboards]
        LE[Looker Embed]
        LS[Looker Studio]
        LA[Looker Actions]
    end

    subgraph PowerBI["Power BI Stack"]
        SM[Semantic Model - Direct Lake]
        RP[Report Pages]
        PD[Power BI Dashboards]
        PE[Power BI Embedded]
        PBD[Power BI Desktop/Service]
        DA[Data Activator]
    end

    LM --> SM
    EX --> RP
    LK --> PD
    LE --> PE
    LS --> PBD
    LA --> DA

Data modeling philosophy comparison¶

Understanding the philosophical differences between Looker and Power BI is critical for a successful migration.

Looker (LookML)¶

• Model-first: LookML defines the semantic layer in code (views, explores, measures, dimensions). All downstream analysis flows through LookML.
• Symmetric aggregation: LookML's symmetric_aggregates feature prevents double-counting across joins automatically.
• Git-native: LookML lives in Git repositories. Version control is first-class.
• Centralized: One LookML model governs all analysis. Analysts explore within the model's boundaries.
• SQL-generated: LookML generates SQL at query time. The semantic layer is an abstraction over SQL.

Power BI (Tabular Model)¶

• Model-first (converging): Power BI semantic models define relationships, measures, and calculation groups. TMDL (Tabular Model Definition Language) enables Git-based workflows.
• DAX for measures: DAX (Data Analysis Expressions) is a formula language for defining measures, calculated columns, and tables.
• Star schema: Power BI works best with star schemas (fact + dimension tables). This aligns naturally with the medallion architecture (gold layer = star schema).
• Direct Lake: In Fabric, semantic models read Delta Lake files directly without import -- equivalent to Looker querying BigQuery at analysis time.
• Copilot: Natural language queries, measure suggestions, and report generation.

Key differences¶

LookML to Power BI semantic model¶

Views to tables¶

A LookML view maps to a table in the Power BI semantic model.

LookML view: sales.order_lines  -->  Power BI table: order_lines
LookML view: sales.dim_product  -->  Power BI table: dim_product
LookML view: sales.dim_region   -->  Power BI table: dim_region

In a Direct Lake semantic model, each table points to a Delta table in OneLake. The LookML SQL definition (sql_table_name or derived table SQL) is replaced by a Delta table reference.

Explores to relationships¶

A LookML explore defines a base table and its joins. In Power BI, this becomes the relationship model (star schema).

LookML explore:

explore: order_lines {
  join: dim_product {
    type: left_outer
    sql_on: ${order_lines.product_id} = ${dim_product.product_id} ;;
    relationship: many_to_one
  }
  join: dim_region {
    type: left_outer
    sql_on: ${order_lines.region_id} = ${dim_region.region_id} ;;
    relationship: many_to_one
  }
  join: dim_date {
    type: left_outer
    sql_on: ${order_lines.order_date} = ${dim_date.date_key} ;;
    relationship: many_to_one
  }
}

Power BI model relationships:

• order_lines[product_id] → dim_product[product_id] (many-to-one)
• order_lines[region_id] → dim_region[region_id] (many-to-one)
• order_lines[order_date] → dim_date[date_key] (many-to-one)

These relationships are defined in the Power BI model view (UI) or in TMDL (code).

Dimensions to columns¶

LookML dimensions map to columns in Power BI tables. Most require no translation because they are simply column references.

Measures to DAX¶

This is the most labor-intensive part of the migration. Every LookML measure becomes a DAX measure.

Filtered measures¶

LookML measures with filters become DAX measures with CALCULATE and FILTER:

LookML:

measure: us_revenue {
  type: sum
  sql: ${revenue} ;;
  filters: [dim_region.country: "US"]
}

DAX:

US Revenue = CALCULATE(
    SUM(order_lines[revenue]),
    dim_region[country] = "US"
)

LookML persistent derived tables (PDTs)¶

PDTs are precomputed tables in Looker. They map directly to dbt incremental models in the CSA-in-a-Box pattern.

Looker dashboards to Power BI¶

Visual mapping¶

Dashboard layout¶

Looker dashboards use a grid layout. Power BI dashboards use a canvas layout with pinned tiles. For the closest match, create Power BI reports (not dashboards) -- reports provide the full canvas layout with interactions, filters, and drill-through that Looker dashboards offer.

Power BI dashboards (pinned tiles from multiple reports) are a lighter concept, more similar to Looker's dashboard-level tiles.

Filters¶

Looker embedding to Power BI Embedded¶

License model difference¶

• Looker: Embedded analytics requires Looker embed licensing (negotiated)
• Power BI: Embedded analytics uses A-SKU (Azure) or F-SKU (Fabric) capacity -- no per-user licensing for embedded consumers

This is often a significant cost advantage for applications with many external users.

Looker Studio to Power BI Desktop/Service¶

Looker Studio (formerly Data Studio) is Google's free self-service BI tool. Migration to Power BI Desktop is straightforward because both are visual report-building tools.

Looker scheduled deliveries to Power BI subscriptions¶

Looker Action Hub to Data Activator¶

Looker Action Hub triggers external actions from BI events. The Azure equivalent is Data Activator (for alert-based triggers) plus Power Automate (for workflow automation) plus Event Grid (for event routing).

Migration sequence¶

1. Inventory all LookML projects, explores, looks, dashboards, schedules, and embeds
2. Prioritize by business criticality and user count
3. Build the semantic model -- translate LookML views to Delta tables (gold layer), define relationships, convert measures to DAX
4. Rebuild reports -- recreate dashboards as Power BI reports, visual by visual
5. Configure subscriptions -- migrate scheduled deliveries to Power BI subscriptions
6. Migrate embedding -- replace Looker embed SDK with Power BI embed SDK
7. Parallel run -- run Looker and Power BI side-by-side for 2-4 weeks, validate measure parity
8. Decommission -- retire Looker instances after parallel validation

Validation checklist¶

After migrating analytics:

• All LookML measures produce matching values in DAX (< 0.5% variance for aggregations)
• Star schema relationships match LookML explore joins
• Row-level security matches Looker user attribute filters
• Dashboard visuals match original Looker dashboards
• Scheduled subscriptions deliver on the same cadence
• Embedded analytics render correctly in target application
• Copilot / Q&A produces reasonable answers for common questions
• Direct Lake semantic model refreshes on Delta table updates

Last updated: 2026-04-30 Maintainers: CSA-in-a-Box core team Related: Compute Migration | ETL Migration | Complete Feature Mapping | Migration Playbook