Founders
How to build an evaluation dataset for AI analytics copilots
AI analytics copilots: Build an evaluation dataset around the decisions the product must support.
Build an evaluation dataset around the decisions the product must support. Include ordinary cases, difficult cases, and known failure modes with documented provenance. Keep development examples separate from the examples used to judge a change.
Start with the decision
Define the unit being evaluated and the expected result before selecting examples. Record who labeled them, what disagreement means, and how the sample relates to actual use. Revisit coverage when the product or customer workflow changes.
The illustrative scenario
This is an educational workbook. The company or fund and all numerical inputs below are fictional. They are not ADIN results, customer claims, market benchmarks, or a forecast.
The fictional copilot answers business questions over an approved sales dataset. Analysts validate the calculation, while managers use the result in planning. Definitions such as active customer or booked revenue can differ across teams.
A persuasive explanation can conceal the wrong denominator or date range. The reader needs the underlying calculation and data definition, not only a natural-language answer.
An evaluation sample with a denominator
Imagine a permissioned evaluation set with 25 cases and 17 that meet a rubric defined before the test. The resulting percentage describes only this invented sample. Seventy percent is not a recommended target, a safety standard, or evidence about any real system. What matters is the rubric, sample coverage, and consequences of the remaining cases.
A persuasive explanation can conceal the wrong denominator or date range. The reader needs the underlying calculation and data definition, not only a natural-language answer. Break the 8 remaining cases into failure categories before interpreting the aggregate. A recoverable wording issue and an unauthorized action may require different product behavior even when both count as a failed case. Keep reviewer disagreement and missing evidence visible.
The assumed time comparison is 24 minutes before and 15 minutes after, including the stated human step. Validate the same start and end points in a real study, then include retries, repair, and handoffs. Compare time to a verified, decision-ready answer with the current analyst workflow. Treat a quick answer with an unresolved definition as unfinished work. A short average time is not enough if the result cannot be used or trusted.
| Item | Illustration |
|---|---|
| Invented evaluation cases | 25 |
| Cases meeting the fictional rubric | 17 |
| Cases requiring investigation | 8 |
| Observed proportion in this exercise | 68.0% |
| Assumed human minutes before / after | 24 / 15 |
Use this worksheet
Answer the questions with a source, date, owner, and next action. Mark an unknown as unknown. Use public or permissioned information and involve a qualified specialist when a question requires expertise beyond the team’s competence.
- What is the evaluation unit?
- Where did the examples come from?
- Who labeled the expected result?
- Which failures are underrepresented?
- What data must remain held out?
Questions specific to this scenario
- Which metric definition applies?
- Can the analyst reproduce the result?
- When should the tool ask a clarification?
Keep approved metric definitions, query results, data freshness records, and examples where an analyst corrected the answer. Include ambiguous questions that require clarification.
A concrete next test
Give an analyst an ambiguous business question and the proposed answer without hiding the underlying calculation. Ask which definition and date range they would verify first. Record whether clarification changes the result before measuring the time required to use it.
Interpret the result
A larger dataset is not automatically more representative. An evaluation should reveal consequential behavior in the intended workflow. If repeated tuning teaches the team the answers, add independent examples before claiming a general improvement.
Compare time to a verified, decision-ready answer with the current analyst workflow. Treat a quick answer with an unresolved definition as unfinished work.
Sources and method
These primary resources provide background for the concepts. The fictional scenario, calculations, and worksheet are ADIN educational examples; the sources do not report or validate those invented results. Reviewed October 11, 2026.