Collection
Business intelligence for manufacturing: strategy, deployment, open source
Business intelligence for manufacturing is rarely a tooling problem. It is a model problem: who has access, who is accountable for which number, and whether dashboards are placed in the path of decisions or to the side of them.
The two pieces below cover both ends: an Apache Superset deployment that removed a Power BI licensing ceiling for a SE Asia garment manufacturer, and a candid look at the four reasons BI implementations still fail in 60–85% of organisations.
Open-source BI is no longer a compromise. Apache Superset embedded into a custom ERP with ClickHouse underneath delivers role-based dashboards at a cost structure that scales with users, not seats.
But deployment is only half the work. The other half (process clarity, action ownership, training that survives the first quarter) is what determines whether a BI implementation strategy actually changes how the business runs.
The dashboard shows the problem. Take the 3-minute diagnostic to see who should act →
What separates BI that gets used from BI that gets built
Most BI implementations are scoped around tooling. A platform is selected, dashboards are built, and then the questions of who looks at them, when, and what they're supposed to do with what they see, get answered after the fact, if at all. That is why 60–85% of BI projects still fail to meet their objectives.
The work we do starts from the opposite end. We map the decisions first: what does a production manager need to know by 8am Monday, and who owns the action when a number is wrong. Then we build the data pipelines and dashboards around those decision points, not around what the tool makes easy to display.
For manufacturers running on custom ERPs or ageing Dynamics installations, Apache Superset connected to ClickHouse delivers role-based, real-time dashboards without Power BI per-seat licensing constraints. For organisations that already have Power BI but aren't getting value from it, the issue is almost always the model, not the tool. Dashboards placed in the path of real operational decisions get used. Dashboards built to prove a project was completed do not.
Common questions
Frequently asked questions
What is the difference between BI dashboards and decision infrastructure?
BI dashboards display data. Decision infrastructure governs what happens next. A dashboard tells you stock is at 12% of safety level. Decision infrastructure determines who needs to act, surfaces the decision to that person with the data they need, routes the approval, and records the outcome in the ERP. Dashboards produce intelligence. Decision infrastructure produces executed, auditable decisions.
How does real-time BI improve operational decisions?
Real-time BI reduces the gap between when a condition changes and when the right person knows about it. In manufacturing and distribution, decisions made on yesterday's data arrive a day late: after the stockout, after the shipment, after the OTIF window closed. Real-time dashboards connected directly to ERP and CRM give operations teams the same data the system has, as it happens.
What BI platforms work best for mid-market manufacturing?
Power BI works well in Microsoft-native environments already running Dynamics 365 or Azure infrastructure, licensing is often already included. Apache Superset is the right choice for teams that need embedded analytics, want to avoid per-seat licensing, or are on non-Microsoft stacks. Both support direct ERP connections, real-time refresh, and custom dashboards. The right choice depends on your existing stack, not the platform's feature list.
When does a BI dashboard become insufficient for operations?
When the team's bottleneck is no longer information: it's action. If your operations team can see the problem in a dashboard but still needs 2–3 days to get a decision approved and executed, the BI is working but the decision infrastructure is missing. The signal is reaching people; the governance layer for acting on it is not there.