quindart
← All articles

Industrial Software Integration Services: What to Specify

Specify industrial software integration around data ownership, interface boundaries, exceptions, acceptance tests and maintainable operation.

An office colleague and a factory operator compare records on a monitor and tablet beside a connected blue machine.

Specify the handoff that needs to work

Industrial software integration services become easier to buy when the request describes a failed handoff rather than a list of applications. An ERP, a scheduling tool and a production record may all work as designed while people still spend hours reconciling what each means. The integration needs to preserve the business meaning of an event as it crosses those boundaries. Begin with the person who notices the failure, the downstream decision it delays and the record that should have informed that decision. Ask the supplier to describe what changes for those people after delivery. A proposal to connect everything leaves too many choices unresolved. A proposal to carry an approved order revision into a defined planning workflow, with visible exceptions and accountable users, gives both parties something concrete to design and test. It also establishes where the supplier's responsibility ends.

Name the authoritative record and the limits of write access

For every shared field, identify the system or person entitled to change its meaning. Customer priority, requested quantity, production release and completion status may belong to different owners. A hypothetical integration that copies a revised order quantity into a production system could overwrite a planner's deliberate hold unless the release decision is separately represented. Ask for a field map that includes identifiers, units, revisions, status definitions and the rules for conflicting updates. Decide which connections are read-only and which may write, under whose approval, and with what audit history. The specification should state how corrections enter the workflow rather than assuming the newest timestamp always wins. Review these rules with business owners and operators before implementation. The purpose is to make authority explicit at the interface, so software does not quietly replace an operational judgment with an accidental synchronization rule.

Specify how the interface behaves through change

A first transfer proves very little about what happens after the tenth revision or an interrupted connection. Require an interface contract that explains the accepted record format, required fields, validation, acknowledgements, permitted timing and handling of repeated messages. In plain language, the team should know whether a retry can create duplicate work, how out-of-order updates are treated and what happens when the receiving system is unavailable. Ask the supplier to show a rejected record and the steps an authorized user takes to resolve it. Include reconciliation that can reveal missing or mismatched records rather than relying only on the absence of error alerts. Agree which vendor-supported access method will be used and which source-system changes need retesting. These are proposed procurement requirements, not a promise that every legacy application supports the same features or that any connection can be built without its vendor's cooperation.

Keep business connectivity distinct from control authority

Connecting industrial information does not automatically justify changing machine control. NIST's operational technology security guidance explicitly addresses performance, reliability and safety alongside security. For procurement, that means an interface near physical operations deserves a clear boundary: what information is read, where it travels and whether any component can issue a command. Ask the supplier and your responsible technical team to document the approved access route, account permissions, monitoring, change procedure and recovery expectations. Do not accept a general integration scope as permission for unsupervised production writes or unreviewed control changes. If controls engineering or specialist safety work is needed, identify the qualified parties and their separate responsibilities before agreeing to delivery. A business-data connection can stand on its own. Expanding it toward equipment should require a justified operating purpose and the relevant engineering review, not happen incidentally during installation. [1]

Use acceptance tests that include disruption and human recovery

Write acceptance around representative work, not only API response codes. A proposed test set could include an approved order, a changed due date, an unsupported status, a repeated update and a temporary source outage. For each, specify the expected destination state, visible evidence and person responsible for any unresolved exception. Ask how the team can pause the connection, determine what has already been processed and resume without silently losing or duplicating work. Test the agreed recovery procedure in an authorized environment before production release. Include a planner or supervisor who will use the result, since technically correct data may still be unusable in their decision context. Record limitations and outstanding issues with an explicit release decision. Acceptance is stronger when the client can reproduce the checks and understand the failure behaviour without watching a supplier conduct a carefully selected demonstration.

Compare handover, change costs and the first useful scope

Before signing, compare more than development effort. Ask about source-system licenses, vendor access, client testing time, hosting, monitoring, documentation, agreed code and configuration rights, and the cost of future changes. Define who maintains mappings when a supplier changes a field or the factory introduces a different workflow. Support hours and response expectations must be agreed explicitly; they should not be inferred from a successful implementation. Quindart's software approach starts with existing business information, validates it against operational reality and builds scoped connections people can inspect and revise. Bring one costly handoff, its relevant systems and the people who own both sides. Request a first working connection with clear acceptance, exclusions and continuing responsibilities. Have the proposed owner demonstrate how to recognize and resolve a rejected update, then show which evidence confirms that both systems agree. Wider integration should follow evidence that the first boundary works as intended.

Sources

  1. NIST SP 800-82 Rev. 3: Guide to Operational Technology (OT) Security