Manufacturing Intelligence: MI vs. BI, Data Stack & MES

What is Manufacturing Intelligence?

Manufacturing Intelligence (MI) is the discipline of collecting production data from every system on the shop floor — MES, SCADA, PLC, ERP, quality systems — normalising it into a single data model, and transforming it into actionable insight for every level of the organisation: from the operator who needs to see why line 4 stopped 12 minutes ago, to the COO who needs to know whether the plant will hit its weekly output target. MI is not a product category. It is a capability — and the question is not "should we buy an MI tool?" but "does our data infrastructure allow us to answer any production question within 60 seconds?" If the answer is no, that gap is the MI gap.

How does Manufacturing Intelligence differ from Business Intelligence?

The terms are frequently confused — especially by vendors selling BI dashboards to manufacturing companies. The differences are structural:

The critical insight: BI tools can display manufacturing data — but they cannot collect it. A Power BI dashboard showing OEE is only as good as the data that feeds it. If the data comes from manual shift reports entered 8 hours after the fact, the dashboard is a reporting tool, not an intelligence tool. MI starts at the data source — the machine signal, the PLC alarm, the cycle counter — and builds the entire chain from collection through contextualisation to visualisation. The MES is the system that builds that chain.

What are the 4 levels of Manufacturing Intelligence maturity?

MI capability develops in stages. Most manufacturing companies are at Level 1 or 2. The value inflection point — where MI starts delivering ROI that exceeds the investment — is at Level 3.

The practical lesson from SYMESTIC's 15,000+ machine connections: most companies overinvest in Level 3–4 ambitions while underinvesting in Level 1–2 foundations. A plant that does not have reliable automatic downtime capture (Level 1) cannot build predictive maintenance models (Level 3). MI maturity is sequential — and the MES is the system that builds each level.

What is the data architecture behind Manufacturing Intelligence?

MI is an architecture problem before it is an analytics problem. The question is not "which dashboard should we build?" — it is "how do we get clean, contextualised data from 200 machines into a single queryable data model?" The answer follows the ISA-95 layer model:

The MES sits at Level 3 — and it is the layer where raw data becomes intelligence. A PLC knows that a machine stopped at 14:32. The MES knows that machine 5 stopped at 14:32 during order 4712 (product: bracket XY, material lot: B7, operator: shift lead C), that the stop was caused by alarm #3012 (hydraulic pressure), that this alarm has fired 47 times in the last 7 days, and that it correlates with a 3.2 % scrap increase. That contextualisation is the difference between data and intelligence.

Why do most MI initiatives fail — and how does an MES prevent it?

The most common failure modes of MI projects — and the MES-based countermeasure:

• Failure: "We built a data lake but nobody uses it." Companies invest in data infrastructure (Azure, AWS, Snowflake) and collect terabytes of machine data. But nobody on the shop floor can query it. The data scientists who built it left 6 months ago. The production team still uses Excel. MES fix: The MES provides pre-built, manufacturing-specific analytics — OEE, downtime Pareto, cycle time distribution, alarm Pareto — that production teams can use without SQL or Python. At Brita, the SYMESTIC dashboards were used by shift leads from day one — no data engineering degree required.
• Failure: "The data is there but it's wrong." Manual data entry, inconsistent downtime reason codes across shifts, operators entering "other" for 40 % of stops. The MI layer analyses garbage and produces garbage insights. MES fix: Automatic data capture from machine signals. The MES does not ask the operator "was the machine running?" — it reads the signal and knows. At Neoperl, automatic stop-reason capture by PLC alarm eliminated the "other" category entirely — because the machine classified its own stops.
• Failure: "We can't compare across plants." Plant A defines changeover as "last good part to first good part." Plant B defines it as "machine stop to machine start." OEE numbers are incomparable. MES fix: Standardised KPI definitions enforced by the MES across all plants. At Carcoustics (7 plants, 500+ machines), SYMESTIC provides the single source of truth — same definitions, same calculations, same dashboards in Haldensleben, Poland, Slovakia, Czech Republic, Mexico, USA and China.
• Failure: "It took 18 months and we still don't have real-time data." Classic on-premise MI project: server procurement, network configuration, database design, ETL pipeline development, dashboard build, UAT, go-live. 18 months later, the first dashboard shows yesterday's data. MES fix: Cloud-native MES. At Klocke, go-live from first machine connection to production dashboards took 3 weeks. At Meleghy, 6 plants in 6 months. The MI capability is built into the MES — it is not a separate project.

FAQ

Do I need a separate MI platform on top of my MES?
For 80 % of mid-market manufacturers: no. A modern cloud-native MES includes the MI capabilities that matter most — real-time OEE, downtime analysis, alarm correlation, process parameter trending, cross-plant comparison. A separate MI platform (Sight Machine, Uptake, ThingWorx) adds value when you have 50+ plants, multiple MES systems from different vendors, and a dedicated data science team. For a company with 1–6 plants and one MES, the MES is the MI platform. The SYMESTIC production metrics module plus the process data module cover Levels 1–3 of MI maturity out of the box.

What is the relationship between Manufacturing Intelligence and Industry 4.0?
Industry 4.0 is the vision — connected, data-driven, autonomous manufacturing. MI is the foundational capability that makes it real. Without MI (clean data, contextualised information, actionable insight), Industry 4.0 is a PowerPoint presentation. With MI, you have the data infrastructure that enables digital twins, predictive maintenance, AI-driven scheduling and closed-loop quality control. MI is not a subset of Industry 4.0 — it is the prerequisite.

How does Manufacturing Intelligence relate to shopfloor management?
Shopfloor management is the daily discipline of using data to manage production. MI is the capability that provides that data. A shopfloor meeting without MI data is a discussion of opinions. A shopfloor meeting with MI data — yesterday's OEE, top 3 downtime reasons, open actions from last week's root-cause analysis — is a structured problem-solving session. At Schmiedetechnik Plettenberg, the SYMESTIC MES became "the central tool in production control" because it provided the MI layer that shopfloor management requires: real-time transparency across machines, shifts and orders.

Can Manufacturing Intelligence be implemented incrementally?
It should be. The biggest mistake is trying to build a complete MI architecture before capturing a single data point. Start with Level 1: connect 10 machines, capture OEE automatically, display it on a shopfloor screen. That alone changes behaviour — operators see their own performance for the first time. Then add Level 2: downtime reason codes, alarm capture, cycle time analysis. Then Level 3: trending, correlation, prediction. Each level builds on the data quality of the previous level. The SYMESTIC starter package is designed for exactly this: go live with production metrics on 10 machines in under 1 month, then expand.

Related: MES: Definition & Functions · OEE Explained · ISA-95 · Industry 4.0 · Smart Factory · SCADA · SYMESTIC Production Metrics · SYMESTIC Process Data · Shopfloor Management

About the author

Christian Fieg

Head of Sales at SYMESTIC. Six Sigma Black Belt. Built MI-driven production systems at Johnson Controls (900+ machines, 30+ processes) and Visteon (global MES Centre of Excellence). Author of OEE: Eine Zahl, viele Lügen. · LinkedIn