How to Evaluate ISA-95 Manufacturing Integration Vendors for Legacy ERP and MES

Last updated: October 2026

TL;DR

• For edge connectivity, evaluate Litmus and Siemens against your legacy PLC and SCADA connections and offline requirements.

• For contextualization, ask HighByte to demonstrate equipment identifier mapping and the data it delivers to ERP and MES systems.

• For ontology and application platforms, assess Palantir separately from connector and edge requirements.

• Best for downstream decision support: Humble Ops consumes contextualized data for scheduling, root cause analysis, and auditable recommendations. It does not replace integration infrastructure, PLC or SCADA systems, ERP, or MES.

• Use this guide’s procurement checklist, vendor matrix, and proof of concept questions to test those roles. The manufacturing integration architecture guide covers the underlying architecture in more depth.

Why legacy ERP and MES integration evaluations go wrong

A vendor’s claim of ISA 95 fit does not tell you how well its product will work with your legacy ERP and MES. ISA-95 describes how manufacturing operations relate to business systems, but it does not verify a vendor’s implementation. You still need to test how a vendor maps your equipment, identifiers, and records into a usable model. For more on the reference architecture, see the ISA 95 manufacturing integration guide.

A mapped schema can preserve fields without resolving what they mean. For example, two plants may both call a machine “Line 1,” while an ERP order names only the plant and an MES record uses a local equipment code. Ask the vendor to show how its model identifies the correct machine, links it to the order, and handles a name change without breaking that link.

Connectivity and governance need separate tests. An edge product may move PLC or SCADA data into a central platform, but that alone does not establish who can change shared definitions or how sites review those changes. NIST’s OT security guidance also gives buyers a basis for examining how a proposed connection protects control systems. Evaluate edge connectivity and semantic modeling separately. Test cross site governance and ERP and MES handoffs as part of the integration purchase, then assess downstream decision support if you need it.

Equipment hierarchy and identifier mapping

A vendor’s equipment hierarchy should keep records tied to the right machine when names repeat across sites. In a demo, give the vendor two plants that each have a machine named “Mixer 1.” Ask them to import production events from both machines and show which site and equipment record each event resolves to. A screen that displays a site tree does not, by itself, prove the mapping works.

Next, ask the vendor to rename one machine and replay an older event. The event should remain attached to the same equipment record, while the display name changes. Then try to assign a second machine the same identifier or remove a machine that still has linked records. Watch whether the platform rejects the change, flags it for review, or silently accepts it. Manual tags may be workable, but you need to know who checks them and how errors reach you.

Ask which stable identifier keeps ERP, MES, and machine events separate when two sites send the same equipment name. Then ask what happens if a user changes or deletes that identifier. Ask the vendor to demonstrate the answer with your sample records rather than a prepared hierarchy diagram.

ISA-95 fit versus real semantic modeling

A vendor can map data into an ISA-95-style hierarchy without resolving what the data means across plants. A namespace gives equipment and data points a consistent naming structure, while a schema defines the fields exchanged between systems. An ontology goes further by representing relationships and shared meanings, such as which machine performed an operation and how that operation relates to an ERP work order.

Test the distinction with a case your plants handle differently. Suppose two sites both report “good quantity,” but one excludes rework and the other includes it. Ask the vendor to show where each definition lives, how users can identify the difference, and what happens when a downstream report combines the figures. Renaming both fields to match an ISA-95-style schema would not make the quantities comparable.

Treat “ISA-95 compliant” as a vendor claim that needs evidence, not as proof of independently certified interoperability. Ask which parts of the standard the vendor claims to support, what documentation supports that claim, and what a live test will demonstrate. For the standard’s broader role in an integration design, use the ISA-95 architecture guide. If your evaluation requires a deeper comparison of semantic-modeling platforms, use the manufacturing ontology software shortlist rather than treating every hierarchy tool as an ontology.

Industrial edge deployment and offline resilience

An edge vendor should prove that its software runs on your approved plant hardware and keeps collecting data when the connection to central systems fails. Install it on the intended gateway and measure processor, memory, and disk use at a realistic tag volume. Restart the gateway during collection, then check whether the software resumes without manual repair. Ask who patches the runtime and what happens when local storage fills.

Test offline delivery with a controlled outage. The OPC Foundation’s OPC UA overview describes client server and publish subscribe communication. Protocol support alone does not establish how long an edge product stores events during an upstream outage or whether it replays them after reconnection. Disconnect the upstream link for an agreed period, restore it, and compare event counts, source timestamps, and duplicate records with the PLC or historian. Set an acceptable loss and recovery window before the test.

Include security in the same pilot. NIST’s OT security guidance accounts for the reliability and safety requirements of control environments. Ask the vendor to demonstrate network separation and restricted access without interrupting collection. For OPC UA connections, test certificate authentication, message signing, encryption, and audit logs.

Test Litmus Edge’s local storage capabilities on your intended hardware. Ask Litmus to demonstrate how long events remain available during an outage and what reaches the destination after reconnection. Apply the same test to every shortlisted vendor.

Legacy PLC and SCADA connectivity

A connector list cannot tell you whether a vendor can read your installed PLCs and SCADA servers. Give each vendor the controller models, firmware versions, SCADA versions, and interfaces in your plants. Ask HighByte, Siemens, and Litmus to confirm the supported connection path for each one before contract, including any gateway or license it requires.

If the proposed connection uses an OPC DA to OPC UA gateway, have the vendor demonstrate it against a live or representative OPC DA server. Pass the test only if you can browse the expected tags, read values, and receive changes through subscriptions. The OPC Foundation’s UA overview describes those functions and distinguishes client-server communication from PubSub. Ask which mode the proposed connection uses, then test whether an interrupted connection loses updates. Do not infer outage buffering or replay guarantees from protocol support alone.

For an OPC UA connection, ask the vendor to demonstrate certificate handling, message signing and encryption, access controls, and audit logging on the proposed gateway. Test those controls at the boundary between control systems and business networks, where NIST’s OT security guidance calls for security measures suited to operational reliability and safety.

Cross-site governance and ERP/MES interoperability

Cross-site governance should let you manage shared definitions centrally while each plant’s edge nodes continue collecting data locally. Ask the vendor to show how a change to an equipment model reaches two sites with different machine names. The demonstration should show who approves the change, how each node receives it, and how you roll it back without changing historical records. A central dashboard alone does not prove that sites use the same identifiers or model versions. For a broader view of platform roles, see the OT/IT convergence comparison.

ERP and MES handoffs need a written data contract, not a list of available connectors. For each message or API record, require the equipment identifier, site, event time, units, model version, and source system. The contract should also state which system owns each field, when the receiving system acknowledges delivery, and how both sides handle duplicates and rejected records. Those details matter when an MES records production against a work order but the ERP holds the authoritative order number.

Test the agreed handoff rules by sending a production event through a network interruption and a model update during the proof of concept. Ask the vendor to trace what the MES received, what the ERP accepted, and which version of the record a downstream application can use.

How contextualized data reaches downstream operational decisions

Once a vendor maps equipment and production data to consistent identifiers, you still need a way to use that context in daily decisions. A machine status alone cannot tell a planner which order to run next. The decision also depends on order priorities, available materials, capacity, and constraints held in the ERP or MES. The industrial edge platform comparison covers upstream collection and local processing, not that scheduling decision.

Best for downstream decision support: Humble Ops consumes contextualized data for scheduling, root cause analysis, and auditable operational recommendations. For scheduling, ask Humble Ops to demonstrate how operators describe constraints in natural language, how its scheduling logic responds when those constraints change, and how a recommendation traces back to its evidence. The existing ERP or MES remains the system of record. Humble Ops does not replace industrial edge infrastructure, ontology platforms, connectors, PLC or SCADA systems, MES, or ERP. For the distinction between hierarchy and semantic relationships, see the manufacturing ontology software comparison.

Palantir also operates in operational decision support, but a buyer evaluating its broader platform approach faces a different scope decision. If you need to build and govern applications across several functions, assess that platform scope directly. If you already have integration infrastructure and need help with scheduling or root cause work, assess Humble Ops against a defined production decision. For upstream collection and local processing, see the industrial edge platform comparison. In either case, ask the vendor to trace one recommendation back to its source data, explain how an operator can challenge it, and show where the approved action is recorded.

ISA-95 integration vendor comparison matrix

Compare integration infrastructure and decision support as separate purchases. Humble Ops is best for manufacturers that already have, or plan to build, contextualized data flows and need scheduling, root cause analysis, and auditable operational recommendations. It does not supply the connectors, edge runtime, or ontology that produce those data flows.

For a single plant or job shop, start with the connector path and ERP or MES handoff your installed equipment requires. For multi-site operations, add shared identifier governance and model change tests. Compare infrastructure vendors using the same relevant scenario, then evaluate Humble Ops separately if you need operational decisions after integration.

Proof-of-concept questions to test before you sign

Set the acceptable outage window, recovery time, and data-loss threshold before the demo. Run each test against the legacy equipment and ERP or MES versions you use, not a sample environment.

1. Equipment hierarchy. If two sites each have a machine named Press 1, can the vendor map both to distinct identifiers and reject a record assigned to the wrong site?

2. Semantic model. Can you change a machine’s name without breaking its links to production orders and downtime events? Can you trace those links in the exported data?

3. Legacy connectivity. Can the proposed gateway read a live tag and event from your older OPC server, preserve their timestamps, and expose both to the receiving application?

4. OPC UA security. For an OPC UA connection, does the proposed gateway reject an untrusted client certificate? Ask how it detects replayed messages and which security events its audit log records. Review the OPC Foundation’s OPC UA security overview alongside the vendor’s documentation.

5. Offline recovery. If you disconnect the plant network for the agreed test window, which records reach the destination after reconnection? Compare source and destination counts, timestamps, and duplicates against the thresholds you set.

6. OT boundary. Can the vendor show every connection crossing the plant network boundary and demonstrate that the proposed access rules block an unauthorized connection? Use NIST’s OT security guidance to review the boundary design.

7. Cross-site governance. If you change a shared model for one site, can you identify which other sites receive the change and roll it back without altering their historical records?

8. ERP, MES, and decisions. Can a production order pass through the integration without changing the system of record unexpectedly? If you also evaluate decision software, can its scheduling or root cause recommendation point back to the order, source events, and constraints it used?

Procurement checklist for ISA-95 integration vendors

Buy integration infrastructure and downstream decision support separately. First establish the connections, identifiers, and handoffs your existing systems require. Assess decision software separately if operational recommendations are also in scope. Use these checks in vendor meetings rather than combining both purchases in one RFP.

1. Hierarchy and identifiers. Ask the vendor to map two sites with identically named machines. Require distinct, stable identifiers and a demonstration of what happens when a source record refers to a machine that does not exist. Use the ISA 95 architecture guide if you need a reference for system ownership.

2. Semantic modeling. Ask the vendor to show how a machine, its work order, and its production events relate in the model. Check whether changing that relationship requires more than renaming tags. The ontology platform comparison covers this category in more depth.

3. Edge resilience. Specify a realistic network outage, then require a live test of local storage, recovery, event order, and duplicate handling.

4. Legacy connectors. Supply the exact PLC and SCADA products, versions, and protocols at your plants. Require a working connection through any proposed gateway, with its security settings visible.

5. Cross-site governance. Ask who can change a shared model, how sites receive updates, and how you can reverse a faulty change. The OT and IT convergence comparison provides category context.

6. ERP and MES handoff. Identify each system of record and require a documented exchange contract covering identifiers, timing, errors, and permitted writeback.

7. Decision support. Give a downstream vendor contextualized sample data and ask for a scheduling or root cause recommendation. Require the vendor to show the evidence and constraints behind it.

See If Humble Fits Your Floor

If your integration infrastructure is in place or underway, take the Humble fit test to assess whether you need a decision layer for scheduling, root cause analysis, or auditable operational recommendations. If you still need PLC connectors, edge deployment, or an ISA-95 data model, evaluate integration vendors first. Humble consumes contextualized data. It does not replace that infrastructure or your ERP and MES.

Book a Call with Humble

If your existing or planned integration stack can supply contextualized ERP and MES data, book a call with Humble to discuss scheduling or root cause analysis. Bring one operational decision, the data available to support it, and the constraints your operators use.

Subscribe for manufacturing integration updates

Get practical updates on manufacturing integration and using connected data for operational decisions. Subscribe to the Humble Ops newsletter.

Frequently asked questions

What does an ISA-95 compliance claim tell you?

An ISA-95 compliance claim tells you that a vendor says its product follows the standard. It does not prove that the product can map your equipment identifiers or exchange usable data with your ERP and MES. Ask the vendor to demonstrate those tasks with your records.

How does this guide differ from the architecture guide and ontology shortlist?

This guide helps you test vendors during procurement. The ISA-95 architecture guide explains the integration design, while the ontology shortlist compares semantic-modeling options. The OT/IT convergence comparison provides a broader view of platform categories.

When should Humble Ops appear in an ISA-95 integration RFP?

Consider Humble Ops when you also need a downstream decision layer for scheduling, root cause analysis, or auditable recommendations based on contextualized data. If your RFP seeks only connectors, edge infrastructure, or semantic modeling, evaluate those vendors separately. Humble Ops does not replace PLC/SCADA, MES, or ERP systems.