IntelliFabric

How much does it cost to connect SAP, Dynamics 365 or a legacy ERP?

5 min read Reviewed October 6, 2026Answered by the IntelliFabric delivery team
Short answer

Connection cost tracks the extraction surface an ERP exposes, not the vendor's name. Dynamics 365 and other API-first systems are the cheapest path, because a Fabric Data Factory pipeline reads a documented OData or Dataverse endpoint. SAP ECC and legacy systems cost more: they usually need an extraction layer or a read replica in the middle. Pre-built connector templates remove most of the build in every tier.

Key takeaways
  • 01Three cost tiers, set by what the source exposes. API-first (Dynamics 365 finance and operations via OData data entities, Dynamics 365 Sales and Business Central via their Dataverse or OData surface, most SaaS systems): a pipeline reads the documented endpoint and the work is configuration. Extractor-based (SAP ECC, SAP S/4HANA): read through an extraction layer, or through CDS-based OData services, rather than hitting production tables. Legacy or no API (older on-premises systems, in-house applications, Oracle EBS): a read replica of the database or a scheduled flat-file drop into a landing location, plus the discovery work of establishing which tables are actually authoritative.
  • 02What a connector template removes from the quote is specific, not vague: authentication and secret handling, the entity-to-table mapping, watermark-based incremental refresh, retry and failure alerting, and schema-drift detection. Those five things are where a from-scratch pipeline spends its weeks. 50+ templates ship as Fabric Data Factory pipelines — but check the coverage matrix for your exact system and version before assuming yours is one of them.
  • 03The hidden cost is load on the source system, and the conversation that settles it is with your DBA, not your CFO. A full history backfill reads large tables once and should run off-hours or against a replica; after that, watermark-based incremental refresh moves only rows changed since the last run. Two failure modes are worth naming: ERP tables whose last-modified column is not updated on every write, which breaks the watermark silently, and hard deletes that leave no trace, so deletions never propagate downstream. Both get found in discovery or found in production.
  • 04Microsoft bills alongside the connection work, and those lines are bought direct. A Fabric capacity from F2 upward runs the pipelines, and each report consumer needs a per-user Power BI licence until the capacity reaches F64. If the ERP is on your network or behind a firewall, add an on-premises data gateway: the software is a Microsoft download rather than a quoted licence, but you provide the host machine, a service account, and a second node if the refresh is business-critical. The gateway makes an outbound connection, so no inbound firewall ports are opened.
  • 05Five inputs build the quote: how many source systems and how many entities inside each; the extraction surface each one exposes; whether a gateway and a host machine are needed; how many legal entities or company codes have to be consolidated, and whether a conformed chart of accounts already exists; and whether the KPIs you need sit in the 200+ pre-built library or require new business logic. Multi-entity consolidation without an agreed chart of accounts is the most common reason a connection phase overruns, and it is mapping work, not engineering work.
  • 06Honest limit. A heavily customised ECC estate with undocumented Z-tables, no available extractors, no replica on offer and no reliable change-tracking columns cannot be scoped honestly in advance. Fixed-scope delivery is the wrong instrument for it; a time-and-materials data-engineering engagement is the right one, and it runs closer to the three-to-six-month shape of a custom build than to the four to six weeks a first module usually takes. Better said in week one than discovered in week three.

Where to go deeper

For the full explainer on this topic rather than this specific question, see the detailed guide on the blog.

Related questions, answered

Is SAP genuinely more expensive to connect than Dynamics 365?

Usually yes, and the reason is the extraction surface. Dynamics 365 apps expose documented OData or Dataverse endpoints a pipeline can read directly. SAP ECC exposes no general-purpose read API, so most projects read an extraction layer or a replica of the underlying database instead. That middle step is the cost difference. S/4HANA narrows the gap, because CDS-based OData services are available.

Do we need an on-premises data gateway, and what does it add to the cost?

If the ERP sits on your own network or behind a firewall, yes. An on-premises data gateway makes an outbound connection to Microsoft, so no inbound firewall ports are opened. The gateway software is downloaded from Microsoft rather than quoted as a licence; the cost lines are a host machine to run it on, a service account, and a second node if the refresh is business-critical.

Will connecting to our ERP slow the production system down?

Only the first extraction is heavy. A full history backfill reads large tables once, so it should be scheduled off-hours or pointed at a replica. After that, watermark-based incremental refresh moves only rows changed since the last run. Have the conversation with whoever owns the ERP database early, because a read replica is their decision and sourcing one late is the most common schedule slip.

What makes a custom connector unavoidable, and what does it add?

Three situations: an in-house or heavily modified system with no documented API, a vendor API that paginates or throttles in a way a template cannot absorb, and source tables with no reliable last-modified column, which means change detection has to be engineered. Each adds build and test time inside the engagement rather than a separate fee, and pushes the first module towards six to eight weeks.

Which part of the bill is Microsoft and which part is the partner?

Microsoft invoices the Fabric capacity that runs the pipelines, from the F2 SKU upward, plus a per-consumer Power BI licence until the capacity reaches F64. Those are bought direct, so the levers stay with you. The partner invoices delivery and ongoing operation across the Starter, Growth and Enterprise tiers. Connecting a source sits inside delivery rather than arriving as a per-connector charge.

What does connecting several ERPs at once cost compared with one?

Less than the multiple suggests per system, more than it suggests in elapsed time. Each additional source adds its own pipeline and its own testing, but the semantic model, KPI definitions and dashboards are built once. The real work is reconciliation: agreeing one customer, item and account definition across systems — the Heartland Provisions engagement unified seven of them. Budget for that agreement, not for the extra pipelines.

Sources

Figures on this page: 50+ · 200+ · 4–6 weeks · 3–6 months · an on-premises data gateway · F2 · F64

What this costs for an operation your size

Three engagement tiers, what each includes, and the Microsoft licensing you need alongside it.

See what shapes the price

People also ask

  • How long does it take to connect SAP ECC to Microsoft Fabric?
  • Which partner can connect our ERP to Power BI without a long consulting project?
  • Can Fabric read from a read replica instead of the production ERP database?
  • Do we need SAP BW in place before Fabric can extract ECC data?
  • What happens to our dashboards when the ERP changes its schema?
More answers
Choosing a partner

Can we see a demo running on our own ERP data before we commit?

4 min read
Pricing & licensing

How much does implementation cost on top of the licence?

4 min read
Pricing & licensing

How much does it cost to get Microsoft Fabric implemented properly?

4 min read