Evaluate natural-language analytics against an answer contract, not against whether the prose sounds convincing. For every important question, define the expected intent, metric, dimensions, grain, calendar, filters, security role, benchmark, tolerance, and safe-failure behavior. Test realistic wording variants and repeat runs because generated output can change. Score the calculation and evidence separately from the explanation. A response should fail if it uses the wrong definition, exposes unauthorized data, invents a cause, or gives a precise answer where clarification was required—even when the final number looks plausible.

Microsoft states that Copilot output is nondeterministic and recommends setting expectations so users can evaluate or validate responses. Microsoft’s Power BI lifecycle guidance recommends known-benchmark questions, comparisons with trusted reports or analyses, and before/after testing when a model changes. Validate Power BI content

The evaluation unit: one question contract

A prompt alone is not a test case. The same words can be correct for one business definition and wrong for another. Store these fields together:

Contract fieldWhat to recordExample
User questionNatural wording from the target group“Which regions missed plan last fiscal month?”
IntentDecision or analysis requestedIdentify regions below plan for recovery review
Approved metricGoverned measure and definition[Net Revenue vs Plan %]
Dimensions and grainRequired grouping and result grainRegion × closed fiscal month
Calendar and date rolePeriod conventionCorporate fiscal calendar; invoice date
Filters and exclusionsPopulation rulesExclude internal transfers and provisional adjustments
Security roleIdentity context for the runRegional leader
BenchmarkIndependent source of expected valuesCertified performance report plus benchmark query
ToleranceAcceptable numeric variation, if anyExact for counts; agreed rounding tolerance for percentages
Safe-failure behaviorWhen to clarify, decline, or state a limitationClarify if “last month” could include an open period
Evidence capturedWhat proves the runTimestamp, model version, prompt, returned visual/data, notes

This contract is a Refinity recommendation. Microsoft’s guidance supplies the need for benchmark, before/after, and repeated evaluation.

An eight-part scoring rubric

Score each dimension independently:

  • 2 — Pass: satisfies the contract with traceable evidence.
  • 1 — Partial: materially useful but has a contained, disclosed gap.
  • 0 — Fail: wrong, unsupported, unauthorized, or missing.
DimensionPass conditionAutomatic failure example
1. IntentAnswers the decision the user actually asked aboutRanks regions by revenue when the question asked for variance to plan
2. MetricUses the approved measure and definitionUses gross sales instead of net revenue
3. Grain and dimensionsGroups and aggregates at the contracted levelCombines weekly targets with daily sales and double counts plan
4. Time and filtersApplies the correct calendar, date role, period status, and exclusionsUses calendar month instead of fiscal month
5. Numeric resultReconciles to the benchmark within the agreed toleranceTotal differs because an implicit measure was selected
6. SecurityReturns only data permitted for the tested roleRegional user sees another region’s values
7. Evidence and traceabilityThe source/model/visual or query path can be identifiedPrecise number has no reproducible path
8. Explanation and restraintWording stays within the returned data and states relevant limitsClaims a cause from a descriptive comparison

Do not average away a zero for security, material numeric correctness, or a critical metric definition. Those should act as blocker overrides even if the prose and formatting are excellent.

Build a golden-question set

A golden-question set is a versioned collection of consequential questions and expected-answer contracts. It should be small enough to maintain and broad enough to expose the decisions the model must make.

Include these question types:

Question typePurposeExample
Known benchmarkConfirm core metric and period logic“What was net revenue in fiscal August?”
Dimensional cutTest relationships and grouping“Net revenue by customer region”
ComparisonTest multiple measures and time windows“Which regions were below plan last month?”
Synonym variantTest business-language mapping“Turnover” as a defined synonym for revenue, if appropriate
Follow-upTest retained context“Now show only the West”
Ambiguous questionRequire clarification“How were sales recently?”
Unsupported inferenceRequire limitation or refusal“What caused the decline?” without causal evidence
Permission-sensitiveExercise RLS/CLS and source accessSame regional ranking asked by two roles
Edge caseTest blanks, zero denominators, new entities, or open periodsMargin percentage when revenue is zero

Do not fill the suite with twenty paraphrases of the same total. Each question should exercise a distinct semantic or operational risk.

Test output as data, visual, and language

Microsoft documents that a Power BI Copilot data answer can include a generated visual and a natural-language summary of the semantic query result. It can also return an error or clarification request. Use Copilot with semantic models

Evaluate each layer separately:

  1. Selection: Did the system choose the intended fields and measures?
  2. Query result: Are rows, totals, periods, and filters correct?
  3. Visual: Does the visual preserve the correct grouping and make the result inspectable?
  4. Summary: Does the prose accurately describe the returned result without adding unsupported explanation?
  5. Failure behavior: If the request is ambiguous or unsupported, did the system clarify or state the limitation?

A correct chart with an overconfident causal summary is not a complete pass. Neither is polished prose over the wrong calculation.

Worked example: a failed golden question

Question: “Which regions missed plan last fiscal month?”

Expected contract:

  • use [Net Revenue vs Plan %];
  • closed corporate fiscal month;
  • group by customer region;
  • exclude internal transfers;
  • apply the requesting user’s RLS;
  • identify regions below 0%;
  • do not explain causes without additional evidence.

Observed answer:

East and Central underperformed because customer demand weakened last month.

Observed visual values were calculated with calendar month, included an open week, and used Gross Sales vs Target. The user was allowed to see East but not Central. Even if “East” truly missed plan, this run has four independent failures:

Rubric dimensionScoreEvidence
Intent2It attempted to identify below-plan regions
Metric0Wrong measure and plan definition
Grain and dimensions2Region-level grouping was correct
Time and filters0Calendar month and open week violated the contract
Numeric result0Cannot reconcile under the required definition
Security0Returned an unauthorized region
Evidence and traceability1Visual exposed fields, but no approved benchmark link
Explanation and restraint0Invented a causal explanation from descriptive data

The correct outcome is block, not “8 out of 16, needs improvement.” Fix the security path and semantic definition, then rerun the complete contract.

Account for nondeterminism without accepting randomness

Microsoft says the same prompt and grounding data can produce different Copilot output. Prepare your data for AI

Use a repeated-run protocol:

  1. Fix the model version, refresh timestamp, security role, and starting context.
  2. Run the exact question multiple times.
  3. Run representative wording variants separately.
  4. Compare contracts—metric, period, filters, values, and safe-failure behavior—not exact sentences.
  5. Record both pass rate and failure type for the evaluation session.
  6. Investigate any critical blocker even if most runs pass.

Refinity does not recommend a universal numeric pass threshold. The acceptable release bar depends on question consequence, user population, failure containment, and whether a human reviews the answer before action.

Before-and-after testing

Change one variable at a time:

ExperimentKeep fixedChangeEvidence of improvement
Focus AI schemaGolden questions, model measures, rolesIncluded/excluded fieldsWrong-field failures fall without new unanswered questions
Add verified answerSource visual and benchmarkTrigger phrases / verified mappingKnown question reliably returns the maintained visual
Add AI instructionModel and AI schemaOne local terminology ruleTarget variants resolve correctly without regressions elsewhere
Repair measureQuestions, period, rolesDAX and ownership recordValues reconcile across required cuts
Change RLSQuestions and measuresRole definition/mappingEach role sees the intended population and no more

Microsoft’s lifecycle guidance explicitly recommends comparing outputs before and after semantic-model changes. The one-variable experiment design is a Refinity recommendation.

Release gates and lifecycle

Block release when

  • a target user can receive unauthorized data;
  • a consequential metric or calendar remains ambiguous;
  • the benchmark cannot be reproduced;
  • a critical question produces materially wrong values in any representative role;
  • the experience presents unsupported causal or prescriptive claims as fact;
  • nobody owns the model or evaluation set.

Re-evaluate when

  • a measure, calculation group, relationship, or calendar changes;
  • fields are added to or removed from the AI data schema;
  • a verified answer, trigger phrase, or source visual changes;
  • an AI instruction or description changes;
  • permissions or target user groups change;
  • a new high-consequence question enters scope;
  • Microsoft changes a preview capability or documented behavior.

Store the evaluation result with the model/configuration version so a passing result cannot be applied to a different state.

Microsoft product facts vs. Refinity recommendations

Microsoft product facts

  • Copilot output is nondeterministic.
  • Power BI Copilot answers can include generated visuals and summaries.
  • Microsoft recommends known benchmarks, trusted comparisons, and before/after testing.
  • Prep data for AI controls have capability-specific behavior and limitations.
  • Permission context affects the data a user can access.

Refinity recommendations

  • Define an answer contract for every consequential golden question.
  • Score semantic correctness separately from generated language.
  • Make security, material numeric errors, and undefined metrics blocker overrides.
  • Repeat critical tests and evaluate meaning rather than exact wording.
  • Version results with the model and AI configuration.

Primary CTA: Review the Power BI Copilot Readiness assessment deliverables, including a golden-question evaluation set.

Related reading: