Modern machine and equipment design is increasingly relying on connected controls. According to Deloitte’s 2025 Smart Manufacturing Survey, 46% of manufacturers are already using Industrial IoT solutions at the facility or network level, highlighting the growing role of connected equipment and data in manufacturing operations. By connecting drives, motors, controllers, protection devices and other equipment, engineers can make more information available for monitoring, troubleshooting and optimization.

And yet, connectivity alone does not improve performance. The engineering decisions made before a connected control architecture is implemented determine whether the resulting system delivers actionable information or simply creates another layer of data to manage.

This can be a significant challenge. Engineers must determine where control decisions should occur, how legacy equipment will be integrated, which communications architecture makes sense, what data is actually useful, and how cybersecurity and future expansion plans should be addressed.

These seven best practices can help ensure that any connectivity decisions are aligned with overall operational goals:

1. Start with the Control Architecture

The most successful connected control systems begin with the machine or process requirements, not which communication protocol or cloud platform to implement. To start, divide control and information processing into three levels: primary control, secondary control and higher-level systems. 

Primary control should remain at the device level for functions requiring rapid response, such as speed control, interlocks, trips and other time-sensitive operations. Secondary control can coordinate equipment at the machine or cell level. Higher-level systems can then be used for fleet monitoring, trend analysis, energy management, remote diagnostics and other applications that do not require millisecond-level responses.

Mapping out this distinction is important because sending every decision to a supervisory or cloud system can introduce latency and create additional points of failure. At the same time, keeping all intelligence at the device level will make it difficult to identify trends across machines or production lines. The goal is to determine where each decision belongs based on response time, safety, reliability and the value of the information.

2. Identify What the Existing Architecture Can Realistically Support

For a new machine, engineers can specify the controls, drives, communications modules and other components as an integrated architecture from the beginning. Retrofitting an existing machine is much more nuanced.

Before replacing or adding a VFD, PLC or other connected device, document the existing architecture and determine how the new component will interact with it. This includes reviewing available I/O, communication protocols, controller capabilities, motor characteristics, network infrastructure and existing safety functions.

A replacement drive may have considerably more diagnostic and communication capability than its predecessor, but those capabilities have little value if the existing PLC cannot receive the information or the network lacks the necessary infrastructure to support it. Mapping out the various distinctions can help establish a practical path forward and ensure the new architecture can function as needed.

READ MORE: Cutting Through the Complexity: Getting Real on Incorporating Robotics into Your Operation & Proving ROI

And not every machine needs to be replaced. Gateways, communication modules and edge devices can often bridge newer equipment and existing controls. The goal is to review the existing architecture and identify where modernization will produce measurable operational value while allowing reliable legacy equipment to remain in service.

3. Choose a Communication Architecture Based on the Application

Simply selecting the newest protocol may not effectively modernize the communication architecture. Keeping the existing plant infrastructure in mind, consider key factors such as response time and latency, the amount of data being exchanged, diagnostic requirements, network redundancy and even future expansions. 

For instance, a relatively simple machine may require a straightforward communications structure. However, a complex production line may need Ethernet-based communications to move information among drives, PLCs, HMIs and supervisory systems. The goal is not to connect every device to every other device across the plant floor; it is to establish the communication paths necessary to support the machine’s control, maintenance and operational objectives.

4. Determine which Data is Worth Collecting

More data does not automatically mean better control. Modern VFDs are a good example. Many drives can provide additional information about a variety of operating conditions. Depending on the application, these measurements can help further determine whether equipment is operating as intended and, in some cases, identify developing problems without requiring additional sensors. But engineers should first identify the decisions they want the data to inform. 

If the objective is to identify an overloaded motor, for example, current and torque information may be more useful than other unrelated parameters.

A useful design question is: What action will we take if this value changes? If there is no clear-cut answer, the data may not need to be collected after all.

5. Match the Motor, VFD and Control Strategy to the Specific Application

Motor and VFD selection is a critical part of the connected controls ecosystem. To ensure these elements can deliver improved machine performance, start with the requirements of the application, as the motor, drive and control strategy must all work together to deliver the required speed, torque and efficiency. 

For applications that require high torque at low speeds or fast, dynamic response, vector control can provide more precise control of motor torque. This can be important for things such as conveyors, extruders and positive-displacement pumps. Scalar control may be sufficient for applications such as fans and pumps where precise torque control and rapid response are less critical, and simplicity and cost are priorities.

The motor itself also needs to be matched to those requirements. Synchronous motors can provide higher power density and efficiency, while induction motors are generally easier to replace and repair. The trade-off between upfront cost, operating efficiency, maintenance and expected service life should be evaluated as part of the overall system design.

READ MORE: Robotics Readiness Checklist: 6 Things to Ask Before Implementation

This same application-based approach applies to energy performance. A VFD can reduce energy consumption by operating a motor at the application’s optimal speed, rather than running continuously at full speed. In appropriate variable-torque applications, even relatively modest speed reductions can produce substantial energy savings. Engineers should therefore evaluate the motor, VFD and load profile together to deliver meaningful efficiency gains without compromising process performance.

6. Ensure Cybersecurity is an Engineering Requirement

Connectivity can deliver significant operational value, but as connected devices become part of a plant’s digital environment, the risk for potential cybersecurity attacks increases exponentially. Cybersecurity requirements now need to be considered during equipment selection and architecture development, not after the system has been commissioned. 

Consider the overall network. For many organizations, the Ethernet network was not intentionally designed—it simply grew as operations expanded. Understanding your overall network infrastructure, from cabling and routing to firewalls and switches, can aid cybersecurity efforts.

The European Union’s recent Cyber Resilience Act reinforces this shift, mandating reporting requirements for actively exploited vulnerabilities and serious incidents. This includes hardware and software products with digital elements that connect directly or indirectly to devices or networks. Although the regulation currently applies to products placed on the EU market, global manufacturers may need to consider these requirements when designing products and managing product lifecycles.

For engineers in decision-making roles, cybersecurity should be treated as another system performance requirement. Asking practical questions during specification is critical. How is the device authenticated? How are credentials managed? How are firmware and security updates handled? What happens when a vulnerability is discovered? The answers to these questions will help identify and mitigate potential risks. 

7. Design for the Machine You Will Build Next

The final consideration is scalability. A connected control architecture should solve today’s problem without making tomorrow’s expansion problematic. Can additional drives, motors, sensors, machines or production cells be incorporated without redesigning the entire system?

This is particularly important for OEMs building equipment that may be replicated across multiple installations. Standardized architectures, consistent device configurations, repeatable communications structures and reusable parameter sets can reduce engineering and commissioning effort while making future troubleshooting much easier to execute. The same principle applies to data. A system should be capable of adding useful information over time without requiring every connected device to be replaced.

The ideal connected control architecture is not the one with the most devices, the most data or the newest technology. It is the one in which control, communications, data, electrical infrastructure and cybersecurity have been engineered together around a defined operational objective.

Connectivity should make the machine easier to control, maintain and improve. Achieving that outcome requires engineering decisions at every layer, from the motor and VFD to the network and higher-level analytics. When those decisions are made deliberately, connected controls become more than a way to collect information; they become a practical foundation for improving machine performance, reliability and long-term scalability.