Operational data
Operational data records the activity and state of a functioning business. It can include transactions, status changes, messages, measurements, and documents generated during work. Unlike a dataset designed specifically for research, its original purpose is usually to run or document operations.
How it works
Applications record activity according to their own schemas and workflows. An event may represent an actual action, a scheduled intention, a system update, or a later correction. Understanding field definitions, time zones, identifiers, and process rules is therefore essential.
Operational records can be combined into reports or task histories, but those transformations add assumptions. Documentation should distinguish source facts from inferred relationships or labels.
Why it matters for licensing
Such data may offer context about business tasks and outcomes, subject to a specific recipient’s needs. A licensing assessment must consider coverage, quality, permissions, confidentiality, and personal information. The existence of records does not establish that they are eligible or useful for AI.
Example
Fictional example: A service business records appointment requests, technician assignments, completion events, and follow-up visits. Before evaluating a proposed dataset, it documents whether “completed” means work finished, paperwork submitted, or an invoice posted.
Limitations and misconceptions
Operational systems may omit unsuccessful attempts or work performed offline. Business rules change over time, and records can reflect local practices rather than a general process. A large archive may require substantial interpretation and preparation.
Questions to ask
- What business event does each field or status actually represent?
- Which time periods, locations, and outcomes are missing?
- What permissions and exclusions apply to reuse outside operations?
Sources
- IBM — System of record · Accessed
- Gebru et al. — Datasheets for Datasets · Accessed
- OECD — Enhancing Access to and Sharing of Data · Accessed
Explore whether your business data could be a fit.
Start with a description of your systems—not a data upload.