Source: https://www.quantumsteeldesign.com/blog/choosing-steel-data-for-automation/ Field guide · Data # Choosing Trustworthy Steel Data for Automation Reliable automation begins with provenance, identity, units, and explicit failure—not with the first convenient spreadsheet or API response. Q [Quantum Steel Design](https://www.quantumsteeldesign.com/about/)Published August 5, 2026 · Updated September 28, 2026 · About 4 min read Questions answered [What makes a steel dataset authoritative enough for automation?](https://www.quantumsteeldesign.com/blog/choosing-steel-data-for-automation/#authoritative-source) [Why are identifiers and units more important than field names?](https://www.quantumsteeldesign.com/blog/choosing-steel-data-for-automation/#identity-units) [How should missing or non-applicable steel properties be handled?](https://www.quantumsteeldesign.com/blog/choosing-steel-data-for-automation/#missing-values) [How should dataset versions be managed?](https://www.quantumsteeldesign.com/blog/choosing-steel-data-for-automation/#versioning) [What is a useful data contract for steel automation?](https://www.quantumsteeldesign.com/blog/choosing-steel-data-for-automation/#validation-contract) [Practice example](https://www.quantumsteeldesign.com/blog/choosing-steel-data-for-automation/#practice-example) [Interactive learning activity](https://www.quantumsteeldesign.com/blog/choosing-steel-data-for-automation/#interactive-lesson) [Sources and corrections](https://www.quantumsteeldesign.com/blog/choosing-steel-data-for-automation/#sources) Most automation failures blamed on code begin earlier: the source was unclear, units were implicit, identifiers drifted, blank values were guessed, or a new edition entered the workflow without a controlled migration. Direct answer ## What makes a steel dataset authoritative enough for automation? An automation source should identify its publisher, dataset name, edition or version, publication context, units, field definitions, known limitations, and relationship to the governing project requirements. The AISC Shapes Database is appropriate for standardized shape dimensions and properties consistent with its stated Manual edition. It does not by itself define a project's selected grade, connection, availability, mill tolerances, or exact manufactured profile. Direct answer ## Why are identifiers and units more important than field names? Identifiers connect the same object across systems, while explicit units preserve numerical meaning. A familiar field name is unsafe if its object, axis, unit system, or convention is ambiguous. Store both human-readable designations and stable machine identifiers where available. Normalize punctuation for search without discarding the source designation. Keep inch, foot, pound, kip, millimeter, and metric properties labeled in the schema and output. Term explained: Nominal dimension A naming or reference dimension that can differ from a measured or specified dimension. Preserve which kind of value a record contains. [Open the glossary entry](https://www.quantumsteeldesign.com/steel-glossary/#nominal-dimension). Direct answer ## How should missing or non-applicable steel properties be handled? Represent missing, unknown, and not-applicable values distinctly when the source supports that distinction. Do not silently replace blanks with zero or copy a value from another family. A zero can be a real mathematical value, while a blank may mean the property is not defined for that shape family. Validation should check required fields by family and stop or report exceptions when the intended operation needs an unavailable value. Direct answer ## How should dataset versions be managed? Pin the version used by production, record it in outputs, compare changes before upgrading, test every supported family, and make migration an explicit release rather than an invisible file replacement. - Hash or otherwise identify the source file. - Record import date, publisher, edition, and transformation code version. - Diff identifiers, record counts, fields, units, and representative values. - Retain the previous production dataset for reproducibility. - Document the effective date of the new dataset in generated output. Direct answer ## What is a useful data contract for steel automation? A data contract states required inputs, accepted identifiers, units, field meaning, assumptions, output schema, error behavior, version, and verification steps so producers and consumers share the same expectations. Validate record counts, uniqueness, type, ranges, cross-field relationships, and known reference cases. Produce a machine-readable error plus a human-readable explanation. The safest failure is visible, scoped, and recoverable. Practice with a fictional example ## Preserve a blank value through an import A fictional source record has an empty optional property. The receiving spreadsheet normally replaces empty cells with zero. That default would erase the distinction between “not supplied” and an actual numerical zero. - Define the field’s meaning and missing-value rule before importing. - Test an ordinary numeric record, a blank record, and an unexpected text value. - Compare the resulting records and require an explicit response to unsupported values. ### What a useful result looks like A defensible import preserves the source meaning or visibly rejects the record. It does not silently invent a value. Keep source edition, field mapping, and conversion notes alongside the data so future changes can be investigated. [Inspect the published shape reference](https://www.quantumsteeldesign.com/steel-tools/) Try it yourself ## Spot the error: a blank silently becomes zero Fictional learning activity. Choices stay in this page; they are not submitted or saved. The exercise checks information and arithmetic, not a design or release decision. A training data import receives a blank thickness field. It substitutes 0 so every row is numeric, then reports that all records passed. Which correction preserves the meaning of the source? Choose the best answer Keep zero because the file imported successfully. Retain a missing-value state and flag the required field. Copy the thickness from the closest-looking shape. Reveal the answer and explanation Answer: Retain a missing-value state and flag the required field. Keep the original record and its missing-value convention. If a downstream calculation requires thickness, stop that calculation for this record and identify the missing input. An optional field may remain blank if the destination contract permits it. A rejected or incomplete record should remain visible in the result. Continue with the [downloadable fictional practice project](https://www.quantumsteeldesign.com/practice-project/), or keep the related source references below with your review. Primary references ## Sources and governing context These links provide authoritative starting points. The current contract documents, adopted codes, applicable law, and qualified project professionals control a specific project. - [AISC Shapes Database v16.0 ↗](https://www.aisc.org/aisc/publications/steel-construction-manual/aisc-shapes-database-v160/) - [AISC Current Standards ↗](https://www.aisc.org/aisc/publications/current-standards/) - [buildingSMART openBIM Standards ↗](https://www.buildingsmart.org/standards/) Published by Quantum Steel Design. [Read about our use of sources and AI assistance](https://www.quantumsteeldesign.com/about/#editorial-approach). To report an error, [contact QSD](https://www.quantumsteeldesign.com/contact/) with the page address, the passage, and the supporting reference. Please omit confidential project information. Technical boundary This educational material explains workflow and terminology. It is not project-specific engineering advice, a connection design, a safety plan, or authorization to fabricate or erect steel. Continue learning ## Follow the connected steel system. [Shape context Understand the properties first](https://www.quantumsteeldesign.com/learning-hub/steel-shapes-and-properties/) [Digital path BIM, automation, data, and AI](https://www.quantumsteeldesign.com/learning-hub/bim-automation-and-data/) [Working example QSD steel shape data and exports](https://www.quantumsteeldesign.com/steel-tools/)