AI + Analytics

Benchmarking an AI Analyst on Your Own 20 Golden Questions

By Chinmay Raibagkar·September 6, 2026·9 min read·Some SQL

The 60-second version

Fluency is not accuracy and demos are not decisions. Building a verified question-SQL-answer suite that scores models, prompts and schema changes.

  • What happened, in one line
  • What to do about it this week
  • What you can safely ignore

"The AI feels smarter than last quarter." Does it actually answer better — on your schema, your definitions, your edge cases? Fluency improves with every model release whether accuracy does or not, and the only way to tell the difference is to stop vibe-checking outputs and start scoring them against questions whose answers you already know.

That scored set is a golden dataset: 20–50 canonical business questions with verified SQL and verified numbers. This post shows how to build one, what to score, and how to run it so model upgrades become measured decisions instead of mood readings.

The short version: an exam the AI cannot bluff

The Golden-Question Loop

Data Journey
Stage 120–50 questions
Collect

Real questions your team asks, covering joins, filters, grains and known trap columns — each with a verified answer.

Your exam paper
Stage 2Number, not prose
Score

Run each question through the tool, compare the returned number and SQL against verified truth. Fluency earns zero points.

% exact + % tolerance
Stage 3Measured upgrades
Decide

New model, new prompt, new schema notes? Re-run the set. Ship what moves the score; revert what does not.

Diff, not demo

A golden dataset is a version-controlled set of question–SQL–answer triples that define correct behaviour for your data. It is the analytics equivalent of a software test suite: nobody would deploy code without tests, but teams deploy new models, prompts and schema changes against production questions daily with no suite at all.

Score the number, review the SQL. A correct number from wrong SQL is luck that will fail next month on different data; wrong SQL with a correct number scores zero. Both artefacts matter — the number for business trust, the SQL for durability.


Building the set: coverage over cleverness

Twenty well-chosen questions beat a hundred variations of "revenue last month". Cover the failure modes from the trust posts:

CategoryExample questionWhat it tests
Grain + dedup"Orders and revenue by day last week"Order-grain counting, no fan-out
Join path"Spend vs delivered revenue by channel"Spend→orders join at right grain
Business definition"Active customers in August"Resolves to the governed definition, not a guess
Filter correctness"Prepaid-only ROAS, Maharashtra"WHERE logic on real dimensions
Date math"Trailing 7-day CAC vs prior 7 days"Window arithmetic, timezone day cuts
Trap column"Net revenue excluding tax and refunds"Picks net over gross, excludes cancelled
Ambiguity"How are we doing on Meta?"Clarifies or states assumptions instead of guessing
Multi-source"Blended MER with COD-adjusted revenue"Cross-source reasoning, RTO awareness

What a 24-question suite revealed

First run
Exact-number pass13 / 24 (54%)Returned number matched verified truth
Tolerance pass (±2%)17 / 24 (71%)FX rounding and window edges explain most of the gap
Failure cluster5 of 7 misses: trap columnsGross-vs-net and prepaid-vs-blended — definitions, not syntax
Fix with highest ROI3 schema notesDocumenting net-revenue and prepaid flags fixed 4 misses at once
The suite's first value is diagnostic, not competitive: it tells you your failures are definitional (fix with schema notes) rather than syntactic (fix with a better model). Most teams discover they never needed a bigger model — they needed three written-down definitions.

Scoring rubric that survives gaming

Three Levels of Correct

Reporting Hierarchy
Tier 1
Exact match

Number equals verified truth to the paisa. Required for definitional questions with deterministic answers.

Score 1.0 — the bar for money questions
Tier 2
Tolerance match

Within a stated band (±1–2%) for questions with legitimate variance: FX dates, window edges, rounding.

Score 0.5 — acceptable with the variance labelled
Tier 3
SQL-correct, number-off

Joins, filters and grain right; number differs for an explainable data reason (fresh partition, late refund).

Score 0.25 — logic sound, data moved

Score zero for: correct-looking number from wrong SQL, unasked-for filters silently added, definitions invented rather than resolved, and confident answers to ambiguous questions with no stated assumption. That last one matters most — an AI that guesses instead of clarifying will eventually guess with your money.


Operating the suite

From Suite to Habit

Process Flow
1

Seed with 20 questions from real history

Support tickets, Slack questions, board asks. Real beats synthetic — synthetic questions test imagination, real ones test the job.

2

Verify answers by hand, once

A senior analyst signs off each triple. The suite is only as golden as its verification; this is the expensive step and it happens once.

3

Run on every model, prompt or schema change

New model version, edited system prompt, new table: re-run before shipping. Regressions announce themselves as diffs.

4

Grow from failures, prune the trivial

Every production wrong-answer becomes a new golden question. Questions every model aces forever get retired.

Refresh answers, not just questions. Verified numbers decay as late-arriving data, refunds and restatements rewrite history. Re-verify the suite's answers quarterly — a suite asserting stale truth punishes correct models for being right.


Frequently Asked Questions

Isn't 20 questions too few to mean anything?

It is too few to rank models on a leaderboard and plenty to govern your deployment. The suite measures fitness for your schema and definitions, where failures cluster in a handful of known traps. Coverage of your traps matters; statistical power over trivia does not.

Who owns maintaining this?

The analyst who verifies the answers — typically whoever currently answers these questions for the business. Budget half a day to seed, an hour a quarter to refresh. The cost of not having it is every model change evaluated by anecdote.

Can DataLens-style tools with visible SQL use the suite directly?

Ideally — visible SQL makes scoring SQL-correctness fast (the 60-second audit becomes a 60-second grade). Black-box tools can only be scored on numbers, which hides the lucky-correct failure mode. Auditability is a prerequisite for meaningful evals, not a nice-to-have.


Summary & Next Steps

Fluency is not accuracy, and demos are not decisions. A golden dataset of verified question–SQL–answer triples converts "feels smarter" into a score, a diff, and a ship/no-ship call.

  • Use a golden dataset to govern every model, prompt and schema change.
  • Use schema mapping notes as the highest-ROI fix the suite will recommend.
  • Use text-to-SQL audits to grade the SQL behind every scored number.
CR

Chinmay Raibagkar

About author →

Founder of DataLens AI. He helps non-technical teams read their ad and database numbers with confidence — which number to trust, what to do next, and what to ignore.