How to Set Up the Smart Manufacturing Process Correctly

Smart manufacturing projects often look straightforward on a presentation slide:

Real factories are rarely that simple.

A machine may be 15 years old and have no modern communication interface. Different departments may use different names for the same equipment. Production data may be available, but nobody is sure whether it is accurate. Maintenance may have its own system, while production planning works from spreadsheets.

This is why setting up smart manufacturing correctly is more about building the right foundation than buying the newest technology.

A practical implementation should normally follow this path:

What Does Setting Up Smart Manufacturing Actually Mean?

Smart manufacturing connects physical production with digital information so manufacturers can understand what is happening, make better decisions and improve operations.

That usually involves several layers.

ISA-95 provides a widely used framework for organizing these manufacturing and enterprise layers and defining information exchange between them. The latest Part 1 revision was published in 2025.

But manufacturers do not need to implement everything at once.

Implementation Steps and Best Practices

Step 1: Choose the Problem Before the Technology

One of the easiest ways to waste money on smart manufacturing is to start with:

“We need AI.”

or:

“We should connect every machine.”

Start instead with an operational problem.

For example:

• excessive unplanned downtime;
• high rejection rates;
• poor production visibility;
• long changeovers;
• energy consumption;
• machine bottlenecks;
• unreliable maintenance information;
• lack of product traceability.

Then define what improvement would look like.

Instead of:

Install machine monitoring.

Use:

Reduce unexpected stoppages on Packaging Line 2 and understand their main causes.

The technology now has a clear job to do.

Step 2: Assess What You Already Have

Before buying anything, walk through the actual production process.

Document:

• machines and controllers;
• PLC brands and communication interfaces;
• sensors already installed;
• SCADA systems;
• MES or production software;
• ERP system;
• maintenance software;
• network infrastructure;
• available production data.

Older equipment should not automatically be considered useless.

Some machines may provide data directly through their controllers. Others may require gateways, additional sensors or external energy and condition-monitoring devices.

The objective is to understand the current factory before designing the future one.

Step 3: Decide What Data You Really Need

A smart factory does not need every possible data point.

It needs the right data for the decision being made.

Suppose the goal is to reduce CNC downtime.

Useful information might include:

• machine running status;
• alarm codes;
• cycle time;
• spindle load;
• tool usage;
• downtime reason;
• production quantity.

Collecting hundreds of unrelated variables may simply create a larger database.

A useful rule is:

Ask:

What decision are we trying to improve, and what information is required to make it?

That simple question can prevent a surprisingly large amount of unnecessary complexity.

Step 4: Standardize Machines and Data

This step is easy to underestimate.

Imagine one system identifies a machine as:

CNC-01

Another calls it:

Machine 17

And maintenance records it as:

Mazak-A

Technically, the machines may be connected. Operationally, the data is still messy.

Create standards for:

• equipment names;
• production lines;
• product codes;
• units of measurement;
• machine states;
• downtime categories;
• timestamps;
• alarm terminology.

Modern smart manufacturing increasingly depends on contextualized information rather than isolated data values. ISA’s 2025 update to ISA-95 specifically highlights the growing importance of data-centric architectures and standards-based contextualization.

Step 5: Build Reliable Machine Connectivity

Once the information requirements are clear, machines can be connected.

Depending on the factory, connectivity may involve:

• industrial Ethernet;
• PLC communication;
• OPC UA;
• MQTT;
• industrial gateways;
• APIs;
• edge devices.

OPC UA is widely used because it provides a platform-independent infrastructure for industrial information exchange. Its scope extends from sensors and control systems through MES and ERP to IIoT and cloud environments.

But choosing a protocol is only part of the job.

The real question is:

Can the data move reliably from the machine to the system that needs it while retaining the correct meaning?

Step 6: Integrate Production and Business Systems

Avoid creating another isolated dashboard.

Smart manufacturing becomes more valuable when information flows between production systems.

For example:

ERP

receives actual production information.

ISA-95 was specifically developed to help structure the integration between manufacturing-control and enterprise systems, reducing risk, errors and confusion around information exchange.

Not every application needs access to every data point. Integration should follow actual business and manufacturing requirements.

Step 7: Design Cybersecurity From the Beginning

Connecting factory equipment changes the cybersecurity risk.

An old machine that once operated almost independently may now communicate with servers, engineering computers or external systems.

Security therefore cannot be added at the end.

Important measures include:

• inventorying connected assets;
• network segmentation;
• controlling user access;
• removing or changing default passwords;
• securing remote access;
• backups and recovery plans;
• system monitoring;
• controlling vendor access;
• incident-response planning.

NIST SP 800-82 Rev. 3 provides guidance specifically for securing operational technology while recognizing the safety, reliability and availability requirements of industrial systems.

CISA also recommends prioritized cybersecurity practices for both IT and OT environments, particularly as a practical baseline for organizations beginning to improve cybersecurity maturity.

Step 8: Start With One Pilot

Trying to transform the entire factory in one project creates unnecessary risk.

Choose one process where:

• the problem is clear;
• data is accessible;
• management supports the project;
• operators understand the process;
• improvement can be measured.

For example:

Pilot Project

Problem: Frequent conveyor stoppages.

Data: PLC states, motor current, alarms and downtime.

Solution: Connect the conveyor data to a monitoring dashboard.

Action: Categorize stoppages and notify maintenance when abnormal conditions occur.

KPI: Unplanned downtime.

A successful pilot teaches the organization how machines, data, software and people actually work together.

Step 9: Measure Whether It Worked

Do not judge the project by the dashboard.

Judge it by manufacturing performance.

Useful KPIs might include:

• downtime;
• throughput;
• scrap;
• cycle time;
• availability;
• changeover time;
• energy consumption;
• quality rate.

ISO 22400 provides an industry-neutral framework for defining and using manufacturing operations management KPIs.

Compare:

If the project does not improve the targeted problem, investigate why before expanding it.

Step 10: Train People and Scale Gradually

Technology alone will not transform a factory.

Operators need to understand what the new information means. Maintenance teams need to know how alerts should be handled. Production managers need to trust the dashboard before using it for decisions.

This human part is sometimes treated as secondary. It should not be.

Once the pilot works:

Expand the same architecture to another machine, line or plant instead of building every project differently.

Common Smart Manufacturing Setup Mistakes

Avoid these common problems:

• buying technology before defining the problem;
• connecting everything without a data strategy;
• ignoring legacy equipment;
• inconsistent equipment naming;
• building isolated dashboards;
• introducing AI too early;
• ignoring OT cybersecurity;
• excluding operators and maintenance teams;
• launching a plant-wide project before proving a pilot;
• measuring technology implementation instead of manufacturing improvement.

A factory with thousands of connected tags but no better decisions is not necessarily a smart factory.

Conclusion

Setting up smart manufacturing correctly is not about creating the most technologically advanced factory possible.

It is about creating a factory where useful information reaches the right people and systems at the right time.

The practical sequence is:

And perhaps the most important point is this:

Do not begin by asking what technology you can install. Begin by asking what manufacturing problem you need to solve.

Once that answer is clear, choosing sensors, connectivity, MES, analytics, cloud systems or AI becomes much easier.