Skip to main content
Methodology

How a submission becomes the record.

A data dictionary defines the objects, properties, and templates an organization uses to describe its assets. This page explains how the platform reviews that content and how it protects the quality of the final record.

Three planes, three guarantees.

The platform keeps consultant submissions, review work, and the published record on separate planes. Each plane makes a different promise.

Submission intake

Deliberately forgiving. Real consultant files vary, so the platform absorbs formatting differences and records each one as a finding rather than rejecting the upload.

Review

The truth layer. Every issue the platform finds becomes a check result that a human reviews, marks, and resolves with evidence on screen.

Master record

Strictly guarded. Only content that passed review and a curator's promotion gate can enter. Integrity rules are enforced at the door.

Stage 1 checks structure and integrity.

When a workbook arrives, an automated engine runs dozens of checks against the canonical structure. Reviewers see each result with the evidence behind it.

Structure

Are the required sheets present, are the columns where they should be, and do header rows match the canonical layout?

Identifiers and references

Every object, property, and group carries a unique identifier. Checks confirm the format is valid, nothing is duplicated, and every reference points at something that exists.

Vocabulary and units

Values that must come from a controlled list are validated against it. Units, data types, and physical quantities are checked for consistency.

Datasheets

Sample datasheets show how values will be provided in practice. Checks compare them against the workbook so the two never drift apart.

Stage 2 reviews content, object by object.

Structure alone does not make a good dictionary. In Stage 2 the review team examines each object on its own terms: are the definitions clear, are the right properties present, do the sample values make engineering sense?

Stage 2 runs the same way as Stage 1, with per-object checklists, marks, and evidence. The two stages share one review model, so nothing has to be learned twice.

Review is a team activity with a full memory.

Several reviewers can work the same submission. Every mark is attributed, every edit is kept, and disagreements are visible instead of hidden.

When reviewers disagree, the conflict is flagged and a lead reviewer decides: adopt one reviewer's mark or record an overriding decision with a reason. Each reviewer finalizes their own stage when done, and the submission outcome waits until the whole team is ready.

Every reviewer works from the same evidence, so a disagreement is about judgment rather than missing context.

Feedback loops until the content is right.

A submission that needs work is not a failure. The consultant receives every finding as specific, per-issue feedback, corrects the source files, and uploads the next round. The review picks up where it left off, and earlier rounds stay on record.

Each finding names its check, its severity, and the exact rows behind it. A correction becomes a lookup rather than a hunt through the whole workbook.

Nothing is deleted along the way. The audit log keeps who did what and when, from first upload to final outcome, so any decision can be traced years later.

When a round comes back with nothing left to fix, the loop ends. The submission is accepted and moves toward promotion into the master record.

Promotion is the final gate.

When both stages are accepted, a curator promotes the content into the Master Data Dictionary. The gate enforces integrity rules row by row: anything unresolved is held back rather than promoted quietly. From that point the master record is the reference for every export and every downstream system.

See the standards behind this

See it with your own data.

Sign in to your workspace, or ask your program administrator about access.