Evidence, Quality Control & Completion Standard
Evidence standard
Every completed lab must leave enough evidence to answer three questions:
- What existed before?
- What did you do or decide?
- What changed, or what did you learn?
Evidence should be lightweight but inspectable. A screenshot alone is rarely sufficient because it can hide the exact logic. Prefer source scripts, structured registers, before/after outputs, and decision records; use screenshots as visual support.
Evidence package
| Evidence type | Example | What it proves |
|---|---|---|
| Executable logic | sql/ch10/10_03_match_candidates.sql, SQL/Python/DDL/validation queries |
Exactly what operation or test was performed. |
| Configuration/object evidence | Schema/table definitions, roles, view DDL | What the environment actually contains. |
| Business artifact | Policy, decision log, RACI, glossary, retention rule | How a Data Management decision was operationalized. |
| Result evidence | Query output, Data Quality score, reconciliation, dashboard | What happened or what was discovered. |
| Interpretation | Lab journal / DMBOK debrief | That the learner understands meaning rather than only procedure. |
Naming examples and rules
- SQL:
sql/ch10/10_03_match_candidates.sql— readable, commented where logic is non-obvious, rerunnable. - Python:
python/quality/profile_customers.py— declares input/output and preserves raw input. - Result:
evidence/ch13/dq_profile_before.csv— enough context to identify the producing script/run. - Diagram:
diagrams/ch08_customer_lineage.png— every arrow has a defined meaning. - Decision:
docs/DEC-014_customer_survivorship.md— issue, options, decision, rationale, owner/date. - Screenshot:
evidence/ch07/access_denied.png— supportive evidence only; never a substitute for role/GRANT logic. - Lab journal:
evidence/ch10/lab_debrief.md— DMBOK connection, surprises, errors, and independent rebuild result.
Technical QC
Before declaring a lab technically complete:
- Run SQL/Python from a clean or rebuilt environment where practical; remove hidden dependencies on manual state.
- Validate expected row counts, keys, and important constraints before and after the exercise.
- Use deterministic synthetic data when the answer key depends on exact results.
- For destructive SQL, test rollback/rebuild paths.
- Avoid hard-coded personal paths when relative/project paths work.
- Never place credentials or secrets in scripts, screenshots, or Git.
- If the lab depends materially on a current software feature, verify that behavior against current official documentation.
DMBOK / source QC
- Read the full controlling chapter before final lab writing.
- Every DAMA-specific claim, role, process, metric, model, or terminology statement must be supported by that chapter or a clearly identified earlier DMBOK prerequisite.
- Never present a Meridian schema, role name, tool feature, or implementation order as a DAMA requirement unless the source actually says so.
- Apply the same question-resolution standard as the Mastery Lab: teaching questions receive immediate answer/explanation/source; true retrieval may hide answers only with a complete key.
- Include an explicit DMBOK Debrief so the learner can articulate conceptual meaning after the tool work.
Independence QC
Pass test: If the learner can make the screenshot but cannot explain the query, table grain, business rule, source-to-target meaning, or governance decision, the lab is not complete.
| Check | Pass condition |
|---|---|
| Explain | Can explain the concept and technical action without reading the step. |
| Modify | Can change an input/condition and predict the effect. |
| Rebuild | Can recreate the learning-critical portion from a blank file/editor. |
| Diagnose | Can interpret a common failure or unexpected output. |
| Transfer | Can choose the right approach in a changed Meridian scenario. |
Visual / PDF QC
- Use the established navy/teal/orange CDMP visual system and semantic emoji cues.
- Use diagrams/screenshots only when they teach or verify something.
- Keep commands/code in readable blocks with explanations adjacent.
- Do not strand learner questions away from their answers in teaching sections.
- Create the companion PDF only after content/technical QC.
- Inspect PDFs for clipped code, broken tables, missing glyphs, orphan pages, and misplaced images.
Completion rubric
| Dimension | Not yet | Working | Mastered for this lab |
|---|---|---|---|
| DMBOK explanation | Cannot explain without notes. | Explains main concept with some prompting. | Explains accurately, uses DAMA terminology, distinguishes nearby concepts. |
| Technical execution | Copies without understanding. | Can modify and rerun. | Can rebuild core work from blank / troubleshoot common failure. |
| Interpretation | Reports output only. | Explains what output shows. | Connects output to business impact, control, and DMBOK meaning. |
| Evidence | Missing or screenshot-only. | Basic saved artifacts. | Reproducible logic + result + decision/context evidence. |
| Transfer | Only recognizes exact tutorial. | Handles changed input. | Handles changed fact/scenario and selects the right approach. |
Lab completion record
Record at minimum:
| Field | Record |
|---|---|
| Lab ID | CH##_LAB_## |
| DMBOK chapter/source | Chapter + section/page anchors |
| Technology blocks used | T# list |
| Date completed | Date |
| Scaffold level | Guided / Faded / Diagnostic / Independent |
| Independent rebuild result | Pass / Partial / Fail + note |
| DMBOK debrief score | 0–3 |
| Evidence folder/link | Link/path |
| Mistakes / repair | Notes |
| Next revisit | Date/trigger |
Source: Applied Laboratory Master Program Guide Sections 7 and 10; Applied Lab Execution, Evidence & QC Standard Sections 2–7.