Learning delivery workshop / 6 MIN READ
Keep score units consistent from the course to the passing decision
Create a score-representation ledger and boundary-value fixtures for the approved assessment rule.
AI-assisted research and analysis for the SNS Creative Fieldbook
Sources checked 2026-11-07
A score of eight can mean eight points out of ten, eight percent or a value on another rubric. Passing decisions become unreliable when the course, reporting feed and interface use different units or apply rounding before the agreed comparison.
What you will leave with: Create a score-representation ledger and boundary-value fixtures for the approved assessment rule.
Use this when
- A learning feed carries raw and normalized scores.
- A displayed percentage disagrees with the assessment outcome.
Not choosing an educationally valid pass mark or prescribing one normalization formula.
Document the representation before comparing numbers
ADL’s xAPI 1.0.3 Score Object section, checked on 7 November 2026, distinguishes raw from scaled scores and explicitly leaves normalization outside the specification. Its scaled range is minus one to one; raw and scaled values need not have a direct relationship. This lesson uses that version’s data contract, not an assumed percentage conversion or a newer profile’s rules.
Ask the assessment owner for the rubric’s units, possible range and intended transformation, if any. Record the value produced by the course, the representation sent to the learning record store and the value shown to the learner. Name any profile-specific restrictions separately. A valid number within a field’s allowed range is not proof that it represents the intended educational result.
Locate the passing rule and its precision
Write the agreed threshold in the same units as the value against which it is evaluated. Identify which component performs the comparison and whether another component also calculates success. If the course and report each decide independently, compare their configured rule and precision rather than assuming identical labels imply identical logic.
Distinguish a display rule from a decision rule. A rounded percentage can be useful for readability without becoming the authoritative value used for passing. Ask what should happen exactly at the threshold and when a value just below it rounds to the same visible number. Do not invent a universal inclusive comparison or change the threshold during troubleshooting. The responsible assessment and implementation owners must approve the rule being tested.
Use boundary fixtures instead of one comfortable pass
Prepare synthetic values comfortably below and above the threshold, exactly at it and close enough on either side to expose rounding differences. Include a nonzero minimum or another rubric feature if the actual assessment uses one. For every row, write the expected raw value, reported representation, displayed label and outcome under the agreed rule.
Keep missing values separate from zero and invalid values separate from a failed assessment. A feed that omits a score does not necessarily assert a score of zero. Likewise, clipping an invalid value into range can hide a broken transformation. Route unexpected representation to the integration owner with the synthetic fixture rather than silently coercing the data or altering a learner’s result.
Reconcile the result without collapsing other learning evidence
Compare the course output, stored representation and report using the same fixture. Find the first point where units, scale or precision change unexpectedly. Check whether the visible outcome derives from the agreed comparison or from a separate reported success value. Resolve that ownership question before declaring one component wrong.
Retain the ledger with the rubric revision, applicable xAPI profile, producer and consumer versions. Recheck it after a question-weighting change or reporting redesign. Keep completion, multiple-attempt selection and parent-course rollup outside this numeric test; a consistent score does not settle those separate decisions. Acceptance means the representation and approved comparison agree across the tested path. It does not make the pass mark educationally valid, prove workplace competence or justify rewriting historical results without a separate authorized process.
Make the reasoning concrete.
A fictional assessment awards eight points out of ten. Its reporting interface treats the raw eight as eight percent, while the course shows eighty percent, producing an apparent disagreement about the learner’s outcome.
Trust the field label
Opening: Assume every number called score is already a percentage.
Hypothesis: The report needs little transformation logic.
Risk: A valid raw value is interpreted in the wrong units.
Reconcile the score ledger
Opening: Document the rubric and approved representation, then test values around the actual passing boundary.
Hypothesis: The team can isolate unit and rounding mismatches.
Risk: The owner must specify the intended comparison rather than rely on defaults.
The team fixes the representation in a synthetic test and verifies boundary cases without changing the assessment standard or any real learner record.
Take the questions into your next brief.
Use these prompts in your own notes. Nothing is submitted from this page.
Units
What does the raw value measure and what range applies?
Transform
Which approved rule, if any, produces the reported representation?
Decision
Which component compares which value with the threshold?
Boundary
What should happen exactly at and just around the threshold?
Missing
How are absent or invalid values kept distinct from genuine zero scores?
Watch for the wrong conclusion.
- Treating every scaled value as an ordinary percentage.
- Rounding a display before applying an unverified passing rule.
- Replacing missing scores with invented zeros.
Your next action
Create below, exact and above-threshold synthetic rows and reconcile raw, reported and displayed values under the approved comparison rule.
Sources & scope
Workflows, worksheets and fictional examples are original editorial analysis, not supplier implementation instructions or measured business results. Sources support only the specifically attributed platform facts. Supplier controls, contracts and local requirements can change; verify the actual placement and current provider guidance before implementation.
- ADL: xAPI1.0.3 Data specification, Score Object section →Primary technical specification · Supports: Supports raw/scaled distinction, scaled range and absence of a universal normalization formula; threshold ownership and fixtures are editorial.