SAP Analytics Migration to Azure¶

Migrating SAP BW, BW/4HANA, SAP Analytics Cloud, and SAP reporting to Microsoft Fabric, Power BI, Databricks, and Azure AI.

Overview¶

SAP analytics is typically the most complex migration workstream because it touches every business domain. SAP BW systems accumulate decades of InfoProviders, process chains, queries, and reports that represent institutional knowledge. This guide provides a structured approach to migrating SAP analytics to the CSA-in-a-Box platform --- Microsoft Fabric, Power BI, Databricks, and Azure AI --- while preserving business continuity.

1. SAP BW / BW/4HANA migration to Fabric¶

1.1 SAP BW architecture mapped to Fabric¶

SAP BW concept	Fabric / CSA-in-a-Box equivalent	Notes
InfoCube	Fabric Lakehouse Delta table (star schema)	Fact table + dimension tables in Delta format
DSO / ADSO	Fabric Lakehouse Delta table (staging)	dbt incremental models replace DSO activation
InfoObject (master data)	Fabric Lakehouse Delta table (dimension)	Master data with text, hierarchy, attributes
CompositeProvider	Fabric SQL endpoint view / Databricks view	Virtual layer for cross-source queries
Open Hub Destination	ADF SAP BW connector	Data distribution / extraction
InfoSource	dbt source definition	Source system mapping
Transformation	dbt model (SQL)	Transformation logic in dbt
DTP (Data Transfer Process)	ADF pipeline activity	Data loading orchestration
Process Chain	ADF pipeline	End-to-end orchestration
BEx Query	Power BI DAX measure + report	Business queries
BW Workbook	Power BI report	Excel-based analysis → Power BI
BW/4HANA HANA views	Fabric SQL endpoint views	SQL-based analytical views

1.2 Migration strategy: phase by InfoArea¶

BW Migration Phases (phased by InfoArea / domain)
├── Phase 1: Finance (FI/CO)
│   ├── Extract: ADF SAP BW connector → ADLS Gen2
│   ├── Transform: dbt models (bronze/silver/gold)
│   ├── Serve: Power BI financial reports
│   └── Decommission: Finance InfoProviders in BW
├── Phase 2: Supply Chain (MM/PP/WM)
│   ├── Extract: ADF SAP BW connector → ADLS Gen2
│   ├── Transform: dbt models
│   ├── Serve: Power BI supply chain dashboards
│   └── Decommission: SC InfoProviders
├── Phase 3: Sales (SD)
│   ├── Extract: ADF SAP BW connector → ADLS Gen2
│   ├── Transform: dbt models
│   ├── Serve: Power BI sales reports
│   └── Decommission: SD InfoProviders
├── Phase 4: HR (HCM/SuccessFactors)
│   └── [similar pattern]
└── Phase 5: Custom/industry-specific
    └── [similar pattern]

1.3 Data extraction from SAP BW¶

ADF SAP BW via Open Hub connector¶

{
    "name": "ExtractBWFinanceData",
    "type": "Copy",
    "inputs": [
        {
            "referenceName": "SapBwOpenHub_FinanceCube",
            "type": "DatasetReference"
        }
    ],
    "outputs": [
        {
            "referenceName": "ADLS_Bronze_Finance",
            "type": "DatasetReference"
        }
    ],
    "typeProperties": {
        "source": {
            "type": "SapOpenHubSource",
            "excludeLastRequest": true,
            "baseRequestId": 123456
        },
        "sink": {
            "type": "ParquetSink",
            "storeSettings": {
                "type": "AzureBlobFSWriteSettings"
            }
        }
    }
}

dbt model for SAP finance data¶

-- models/gold/finance/gl_journal_entries.sql
-- Replaces SAP BW InfoCube 0FI_GL_14 (General Ledger Line Items)

{{ config(
    materialized='incremental',
    unique_key='journal_entry_key',
    partition_by='posting_date',
    tags=['finance', 'sap-bw-migration']
) }}

WITH source AS (
    SELECT * FROM {{ ref('silver_sap_acdoca') }}
    {% if is_incremental() %}
    WHERE posting_date > (SELECT MAX(posting_date) FROM {{ this }})
    {% endif %}
),

enriched AS (
    SELECT
        s.rclnt AS client,
        s.rldnr AS ledger,
        s.rbukrs AS company_code,
        s.gjahr AS fiscal_year,
        s.belnr AS document_number,
        s.docln AS line_item,
        s.racct AS gl_account,
        cc.gl_account_name,
        s.rcntr AS cost_center,
        s.rprctr AS profit_center,
        s.rhcur AS local_currency,
        s.hsl AS amount_local_currency,
        s.rwcur AS transaction_currency,
        s.wsl AS amount_transaction_currency,
        s.budat AS posting_date,
        s.bldat AS document_date,
        s.usnam AS user_name,
        CONCAT(s.rclnt, s.rldnr, s.rbukrs, s.gjahr, s.belnr, s.docln)
            AS journal_entry_key
    FROM source s
    LEFT JOIN {{ ref('dim_chart_of_accounts') }} cc
        ON s.racct = cc.gl_account_number
)

SELECT * FROM enriched

2. Fabric Mirroring for SAP (near-real-time)¶

Fabric Mirroring eliminates the need for traditional ETL from SAP. It provides near-real-time replication of SAP HANA tables to OneLake as Delta tables.

2.1 How Fabric Mirroring for SAP works¶

SAP HANA (Azure VM)                    Microsoft Fabric
┌───────────────────┐                  ┌────────────────────────┐
│  SAP Tables       │                  │  Mirrored Database     │
│  ├── VBAK (SO hdr)│──CDC Stream──►  │  ├── vbak (Delta)      │
│  ├── VBAP (SO itm)│──CDC Stream──►  │  ├── vbap (Delta)      │
│  ├── EKKO (PO hdr)│──CDC Stream──►  │  ├── ekko (Delta)      │
│  ├── EKPO (PO itm)│──CDC Stream──►  │  ├── ekpo (Delta)      │
│  ├── ACDOCA (FI)  │──CDC Stream──►  │  ├── acdoca (Delta)    │
│  └── MARA (matl)  │──CDC Stream──►  │  └── mara (Delta)      │
└───────────────────┘                  └────────────────────────┘
                                              │
                                              ▼
                                       ┌─────────────┐
                                       │  Power BI    │
                                       │  (Direct Lake│
                                       │   mode)      │
                                       └─────────────┘

2.2 Supported SAP tables for mirroring¶

SAP module	Key tables	Typical row count	Notes
Finance (FI)	ACDOCA, BKPF, FAGLFLEXA	10M--500M	Universal Journal is the primary finance table
Sales (SD)	VBAK, VBAP, VBRK, VBRP	5M--100M	Sales orders, billing documents
Purchasing (MM)	EKKO, EKPO, EBAN, EKBE	5M--50M	Purchase orders, requisitions
Materials (MM)	MARA, MARC, MARD, MAKT	500K--5M	Material master data
Inventory (MM)	MATDOC (S/4), MSEG (ECC)	10M--200M	Material movements
Production (PP)	AFKO, AFPO, AFRU	1M--20M	Production orders
Plant Maintenance	AUFK, AFIH, EQUI	500K--10M	Maintenance orders, equipment
Master Data	KNA1/BUT000, LFA1/BUT000	100K--2M	Customer/vendor (Business Partner in S/4)
HR (HCM)	PA0001, PA0002, PA0008	50K--500K	Personnel master (if not on SuccessFactors)

2.3 Configuration steps¶

See Tutorial: SAP Data to Fabric for step-by-step configuration.

3. SAP Analytics Cloud to Power BI¶

3.1 Feature mapping¶

SAP Analytics Cloud feature	Power BI equivalent	Parity	Notes
Stories (dashboards)	Power BI reports	Full	Richer visualization library in Power BI
Models (imported data)	Power BI semantic models	Full	Direct Lake eliminates import latency
Planning (BPC integration)	Power BI + Azure AI (limited)	Medium	SAC planning has no direct PBI equivalent; evaluate Power BI writeback or retain BPC
Smart Predict	Azure ML + Azure AI	High	ML capabilities exceed SAC Smart Predict
Smart Discovery	Power BI Quick Insights + Copilot	High	Copilot provides natural-language insights
Smart Assist	Copilot for Power BI	High	AI-assisted report building
Data connectivity (live)	DirectQuery / Direct Lake	Full	Direct Lake provides near-import performance with live data
Data connectivity (import)	Import mode / Power Query	Full	Scheduled refresh
Collaboration	Microsoft Teams integration	Full	Power BI tabs in Teams channels
Mobile	Power BI Mobile app	Full	iOS, Android, Windows
Embedded analytics	Power BI Embedded	Full	Embed in custom applications
Multi-tenancy	Power BI workspaces + RLS	Full	Row-level security for multi-tenant

3.2 SAC story to Power BI report migration¶

Migration step	Tool / approach	Notes
Export SAC model metadata	SAC Content Network or manual export	Document dimensions, measures, hierarchies
Recreate data model	Power BI Desktop (import from Fabric Lakehouse)	Use Direct Lake mode for SAP data in OneLake
Recreate visualizations	Power BI Desktop	Manual recreation; no automated converter
Recreate calculated measures	DAX measures in Power BI	SAC formulas → DAX conversion
Recreate filters/prompts	Power BI slicers and filters	Interactive filtering
Migrate users/permissions	Entra ID groups → Power BI workspace roles	RLS for row-level security
Validate numbers	Side-by-side comparison	Compare SAC and Power BI output for key KPIs

4. SAP BusinessObjects to Power BI¶

SAP BO component	Power BI equivalent	Migration approach
Web Intelligence reports	Power BI reports	Recreate using Power BI Desktop
Crystal Reports	Power BI paginated reports	Use Power BI Report Builder (SSRS-based)
BO Universes	Power BI semantic models	Recreate semantic layer in Power BI
BO Explorer	Power BI Q&A + Copilot	Natural-language query
Analysis for Office (AO)	Analyze in Excel (Power BI)	Similar Excel-based analysis experience
BO Server (CMS)	Power BI Service (cloud)	SaaS; no server management
BO scheduling	Power BI scheduled refresh	Automated report refresh

5. SAP data to Azure AI (process intelligence)¶

CSA-in-a-Box enables AI-driven insights on SAP data that were never possible within the SAP ecosystem.

5.1 AI use cases on SAP data¶

Use case	SAP data source	Azure AI service	Business value
Invoice anomaly detection	BSEG/ACDOCA, BKPF	Azure OpenAI + Anomaly Detector	Detect duplicate payments, unusual amounts, timing patterns
Demand forecasting	VBAK (sales orders), MATDOC (consumption)	Azure ML + Time Series Forecasting	Predict demand for inventory optimization
Supplier risk scoring	EKKO (POs), LFA1 (vendors), EKBE (GR/IR)	Azure OpenAI + custom ML	Score suppliers on delivery, quality, financial risk
Contract analysis	SAP document management (attachments)	Azure AI Document Intelligence + OpenAI	Extract terms, obligations, renewal dates from contracts
Predictive maintenance	EQUI (equipment), AFIH (notifications)	Azure ML + IoT integration	Predict equipment failures from maintenance history
Cash flow forecasting	BSID/BSIK (open items), BKPF (postings)	Azure ML + Time Series	Predict cash position for treasury management
Customer churn prediction	KNA1 (customers), VBAK (orders)	Azure ML	Identify at-risk customers from order patterns

5.2 Architecture for SAP AI¶

SAP HANA (Azure VM)
    │
    ▼ (Fabric Mirroring)
OneLake (Delta Tables)
    │
    ├──► Databricks (ML model training)
    │       │
    │       └──► Azure ML (model registry, deployment)
    │               │
    │               └──► Real-time inference endpoint
    │
    ├──► Azure OpenAI (prompt-based analytics)
    │       │
    │       └──► "Summarize AP aging for company code 1000"
    │            "Which suppliers have the highest late delivery rate?"
    │            "Detect anomalies in GL postings for Q4 2025"
    │
    └──► Power BI (AI visuals)
            │
            └──► Key Influencers, Decomposition Tree, Smart Narratives

5.3 Example: Invoice anomaly detection with Azure OpenAI¶

# Databricks notebook: Invoice anomaly detection on SAP data
from pyspark.sql import SparkSession
from openai import AzureOpenAI

# Read SAP finance data from OneLake (Fabric Mirroring)
df_invoices = spark.read.format("delta").load(
    "abfss://sap-finance@onelake.dfs.fabric.microsoft.com/acdoca"
)

# Aggregate invoice patterns by vendor
vendor_stats = df_invoices.groupBy("lifnr").agg(
    F.count("*").alias("invoice_count"),
    F.avg("wsl").alias("avg_amount"),
    F.stddev("wsl").alias("stddev_amount"),
    F.max("wsl").alias("max_amount"),
    F.min("budat").alias("first_posting"),
    F.max("budat").alias("last_posting")
)

# Flag statistical anomalies
anomalies = vendor_stats.filter(
    (F.col("max_amount") > F.col("avg_amount") + 3 * F.col("stddev_amount"))
)

# Use Azure OpenAI to summarize findings
client = AzureOpenAI(
    azure_endpoint="https://aoai-sap-analytics.openai.azure.com/",
    api_version="2024-02-15-preview"
)

anomaly_summary = client.chat.completions.create(
    model="gpt-4o",
    messages=[{
        "role": "system",
        "content": "You are a financial analyst reviewing SAP invoice data."
    }, {
        "role": "user",
        "content": f"Analyze these vendor invoice anomalies and provide recommendations:\n{anomalies.toPandas().to_string()}"
    }]
)

6. Migration timeline for SAP analytics¶

Phase	Duration	Activities
Discovery	4--6 weeks	Inventory BW InfoProviders, SAC stories, BO reports; categorize by domain and complexity
Pilot domain	6--8 weeks	Migrate one InfoArea (e.g., Finance) end-to-end; validate with business users
Parallel run	4--8 weeks	Run old and new analytics in parallel; compare outputs
Phased rollout	3--6 months per domain	Migrate remaining domains (Supply Chain, Sales, HR, etc.)
Decommission	4--8 weeks	Shut down BW/SAC after all domains migrated

Last updated: 2026-04-30 Maintainers: CSA-in-a-Box core team Related: Feature Mapping | Tutorial: SAP Data to Fabric | Integration Migration