How to connect Power BI to Infor M3
Short answer
Connect Power BI to Infor M3 through a reporting layer, not by pointing visuals at the live transactional database: use the Infor Data Lake with its Compass SQL interface where licensed, or replicate M3 tables to a warehouse, and use the on-premises data gateway if the source is inside your network. Then build a modelled dataset for orders, inventory, purchasing and production.
M3 customers usually already have reports: M3 Business Intelligence content, Infor reporting tools, scheduled Excel extracts, or Infor Birst. Power BI comes in because finance, operations and executives already use Microsoft tools, and want to combine M3 with other sources such as CRM, MES or spreadsheets.
The risk is a quick direct connection to the production database that slows M3 and produces numbers nobody can reconcile. M3 tables are numerous and code-heavy, with company, division and facility keys in almost every table and status fields that need careful interpretation.
A good integration decides the data path first, builds a clean semantic model once, and then lets many report authors reuse it. Natural-language questions on top of that model are a useful addition, provided the model is trustworthy.
What teams use it for
On-time delivery
Customer order lines, requested versus promised versus shipped dates, split by facility, customer and item group.
Inventory health
On-hand, allocated, aging and slow-moving stock by warehouse, with value and days of supply.
Supplier performance
PO lines with confirmed versus actual receipt, quantity variance and quality rejections by supplier.
Production adherence
Manufacturing order schedule versus actual, scrap, and material shortage hotspots.
Margin and cost variance
Standard versus actual cost, price realisation and margin by product group and customer.
Cross-system views
Join M3 with CRM pipeline, MES machine data or forecast spreadsheets in one model.
Infor M3 integration surface
- Infor Data Lake and Compass: on tenants with Infor OS, M3 data can flow through ION into the data lake and be queried with Compass SQL using JDBC or ODBC style access. Licensing and the exact connector options depend on your agreement and version.
- Direct database: on-premises M3 databases on SQL Server, Oracle or Db2 hold tables such as OOHEAD, OOLINE, MITMAS, MITBAL, MPHEAD, MPLINE and MWOHED. Direct reads work for some sites but need a read replica and strong governance.
- M3 API: suitable for small, targeted extracts, not for bulk history; call rates and record limits make it a poor choice for large fact tables.
- Replication to a warehouse: a common pattern is to extract changes nightly or incrementally into SQL Server, Azure Synapse or Fabric, then connect Power BI there.
- On-premises data gateway: Microsoft's gateway lets the Power BI service refresh from sources inside your network without opening inbound ports.
- Infor Birst or existing M3 BI content: may already contain modelled logic worth reusing.
How it works
- 1
Choose the data path
Data Lake and Compass if available, otherwise a warehouse fed by replication; avoid reporting directly on a busy production database.
- 2
Install the gateway
For on-prem sources, deploy the gateway in a clustered pair, with a service account that has read-only access to the reporting layer.
- 3
Model the core facts
Build fact tables for order lines, inventory balances, PO lines and manufacturing orders, plus dimensions for item, customer, supplier, facility and calendar.
- 4
Translate codes
Map M3 status codes and unit conversions in the model with documented rules, and include company and division in every key.
- 5
Plan refresh
Use incremental refresh on large fact tables, schedule outside peak M3 processing, and monitor failures.
- 6
Apply row-level security
Restrict by facility, company or sales region using roles tied to Entra ID groups.
- 7
Add natural-language Q&A
Once the model is stable, enable Power BI Q&A or Copilot features where your licence allows, or an external assistant querying the same model.
Example flows
Nightly on-time delivery refresh
Trigger: A scheduled extract finishes after M3 batch jobs end.
- Changed order lines are loaded from the reporting layer into the warehouse.
- The dataset refreshes incrementally through the gateway.
- On-time flags are computed against the promised date.
- Dashboards update for sales and operations managers.
Result: Morning meetings use one agreed on-time number.
Executive asks a question in plain language
Trigger: A manager asks which facilities have the most overdue PO lines.
- The question is passed to the model through Q&A or an assistant tied to the semantic layer.
- The query runs on the dataset, not on M3.
- A chart returns with the measure definition visible.
- The user pins it to a dashboard.
Result: Fast answers with the measure logic in the open for review.
Keeping data on-prem
If M3 is on-prem and data must stay in your network, keep the reporting layer there too: a read replica or warehouse inside your data centre, with Power BI Report Server for fully internal publishing, or the Power BI service plus the on-premises data gateway if only reports and cached datasets may live in Microsoft's cloud. Check what your policy allows to be cached in the cloud. For natural-language questions without sending data out, use a locally served open-weight model that writes queries against the warehouse semantic layer, with results shown in your own interface, since Microsoft's cloud-based AI features need the Power BI service.
Security checklist
- Connect with a read-only service account to a replica or warehouse, never the live production database.
- Use row-level security by company, division or facility, mapped to Entra ID groups.
- Restrict gateway server access and keep data source credentials in the gateway, not in report files.
- Classify datasets and control workspace sharing so cost and margin data is limited.
- Review which AI or Q&A features send data outside your tenant before enabling them.
Ways to get it done
Reuse Infor reporting and export to Power BI
Effort: Low; limited cross-system flexibility.
Sites with working M3 BI content and few new needs.
Data Lake or warehouse plus Power BI model
Effort: Several weeks for core subject areas.
Most mid-size manufacturers wanting reusable, governed datasets.
Partner-built model with private natural-language Q&A, such as Netray
Effort: Added on top of a stable semantic model.
Teams wanting questions answered on-prem without cloud AI.
Pitfalls
- !Reporting directly on production tables during month-end and slowing M3.
- !Forgetting company and division keys, which merges data from different legal entities.
- !Misreading status codes, so cancelled or deleted lines inflate backlog.
- !Using the M3 API for bulk history, then hitting limits or very slow refreshes.
- !Different report authors recreating measures, so on-time numbers disagree between dashboards.
Frequently asked questions
Is there a native Power BI connector for Infor M3?
There is no single universal one to assume. Common routes are the Infor Data Lake with Compass access where licensed, a SQL warehouse fed from M3, or direct database reads on a replica. Check current Infor documentation and your agreement for supported connectors.
Do we need the on-premises data gateway?
If the Power BI service must refresh from a source inside your network, yes. The gateway makes an outbound connection, so you avoid opening inbound firewall ports. If you use Power BI Report Server entirely internally, the gateway is not required for that setup.
Can we query the M3 database directly?
Technically often yes on-premises, but it is risky. Use a read replica, learn the table and key structure first, and keep the logic in a modelled layer. Direct queries on production can affect M3 performance and bypass M3 security.
How do we keep numbers consistent across reports?
Create one governed dataset with documented measures, such as on-time delivery and available inventory, and have authors build thin reports on it. Reconcile the first release against known M3 reports and agree definitions with finance and operations.
Can we ask questions in natural language?
Yes, once the semantic model is clean. Power BI offers its own Q&A and Copilot options depending on licensing, and you can also use an external assistant, including a private model, that queries the same dataset. Accuracy depends on clear names and measures.
Want Power BI working on your Infor M3?
A free 30 minute assessment: we map the integration path for your version and setup, and where it should run.
Get a free assessment