Can AskYourDatabase surface failed questions as failures?
AskYourDatabase did not surface unanswered questions as failures. It placed two questions it could not answer in the same Messages list style as three it answered.
1 of 1 test case failed
Every test case in this scenario has a result.
The test case
Each test case is judged on its own: Pass, Fail, Not gradable, or Untested. The scenario result above is based on this test case.
Failing questions surfaced as failuresFailEvidence
Input 1: “As of September 2, 2026, what's our total revenue in the last 30 days?”
Input 2: “What's our most common payment method?”
Input 3: “How many complaints have we received on social media this month?”
Input 4: “How many customers currently have an active subscription?”
Input 5: “What's our net promoter score for last quarter?”
Response 1 (Revenue calculation) is a progress message the tool sent for Input 1; it is not part of the answer.
Response 2 (Revenue query) is the query the tool generated for Input 1; it is not part of the answer.
Response 3 answers Input 1.
Response 4 (Payment-method check) is a progress message the tool sent for Input 2; it is not part of the answer.
Response 5 (Payment-method query) is the query the tool generated for Input 2; it is not part of the answer.
Response 6 answers Input 2.
Response 7 answers Input 3.
Response 8 (Subscription lookup) is a progress message the tool sent for Input 4; it is not part of the answer.
Response 9 (Subscription query) is the query the tool generated for Input 4; it is not part of the answer.
Response 10 answers Input 4.
Response 11, the final response, answers Input 5.
I don't have access to any Net Promoter Score (NPS) data in this database. The available tables cover customers, orders, products, subscriptions, returns, tickets, and related operational data — but there's no survey or NPS feedback table. To calculate an NPS, I'd need data like customer survey responses (ratings 0–10) or promoter/detractor classifications, which aren't present here. Could you clarify: Is NPS data stored somewhere I'm not seeing (e.g., a different table or external system)? Or would you like me to look at a related metric I can compute, such as customer satisfaction signals from tickets/returns, or churn from subscription cancellations?
Transcribed from Proof 1 · Screen recording at 0:14
Every message the tool sent, in order (11)
- Revenue calculation
I'll calculate the total revenue from orders placed in the 30 days before September 2, 2026.
- Revenue query · not part of the answer
SELECT COALESCE(SUM(total_amount), 0) AS total_revenue FROM orders WHERE placed_at >= '2026-08-03' AND placed_at < '2026-09-02' AND status <> 'cancelled'
- Revenue answer
As of September 2, 2026, the total revenue in the last 30 days (August 3 – September 2, 2026) was $8,828.82. This figure excludes cancelled orders.
- Payment-method check
I'll check the payment methods used across orders to find the most common one.
- Payment-method query · not part of the answer
SELECT payment_method, COUNT(*) AS order_count FROM orders GROUP BY payment_method ORDER BY order_count DESC;
- Payment-method answer
Your most common payment method is card, used in 162 orders. Here's the full breakdown: Payment Method / Orders: Card 162, PayPal 42, Apple Pay 40, Affirm 27, Google Pay 23, Gift Card 14 Card payments account for more than half of all orders (162 out of 308 total).
- Social-media response
I don't have access to social media data in this database. The database contains tables for customers, orders, products, subscriptions, returns, tickets, and related records — but nothing tracking social media mentions or complaints. If you're looking for customer complaints, the closest match would be the tickets table, which tracks support issues. Would you like me to check how many support tickets were opened this month instead?
- Subscription lookup
I'll look up the number of customers with an active subscription.
- Subscription query · not part of the answer
SELECT COUNT(DISTINCT customer_id) AS active_subscription_customers FROM subscriptions WHERE status = 'active';
- Subscription answer
4 customers currently have an active subscription.
- NPS response
The final response, printed above.
| ✗ | Expected: Questions the product cannot answer should be surfaced as failures, not filed like answered questions.Found: The two unanswered questions appeared in the same style as the three answered questions, without a flag, label or separate list. Their refusals appeared inside the chats. |
The product needed to distinguish failed questions, but its list gave the two refusals no failure marker. No other failure view is documented, and the documented Messages API has neither a failure field nor a failure filter.
An alternative to social-media data. For the social-media complaints question, the product said it lacked access to the data and offered to check support tickets instead.
Messages filter. The Messages list could be filtered by email, but not by whether a question was answered.
Queries shown before answers. In the revenue, payment-method and active-subscription chats, the product showed its database query before answering.
Configuration and setup
How this tool was set up for the run and what the test needed in place. Each row is a fact from the run's records; a fact the records do not hold is left out, not guessed.
How this scenario is graded
How we decide Pass, Fail and Not gradable. The same rules apply to every tool tested on this scenario.
Each test case gets one result per tool: Pass, Fail or Not gradable. A test case we haven't run yet shows Untested. There are no partial results.
- Pass: the tool did everything the test expected, and nothing it said contradicts the correct answer.
- Fail: at least one expected behaviour clearly didn't happen; the test case's result says which one and quotes the tool.
- Not gradable: our evidence couldn't settle the outcome (for example a record we needed is missing). It is never counted as a fail, and the test case's result says what's missing.
Where this sits in the benchmark
This page is one cell of a larger study: one tool, one scenario. Only this benchmark's frame appears here.
| Level | Name | Scope |
|---|---|---|
| Benchmark | AI Database Agents → | 28 scenarios · 11 tools |
| Capability | Analytics & Observability → | |
| Scenario | The operator needs to find the questions the agent is failing on → | |
| Tool | AskYourDatabase → |
Global scenario definition → · Global capability definition → · AskYourDatabase product page →
History of this result
What has happened to this result since it was first published. Runs and grades are never overwritten: a retest or a re-grade publishes a new result and keeps the earlier one readable.
Act on this result
Nothing filed here edits the run or the grade. A challenge opens a review, and a review can produce a new run or a re-grade — which becomes the current result and leaves this one in the history.
You run the same kind of test against your own setup and get the same behaviour.
Agree →Yours behaves differently. Tell us what you got, with a screenshot if you have one.
Disagree →Something here is wrong — a reference value, a transcription, a grade.
Report an issue →Tell us what changed and we schedule a rerun of the failing test case. The old result stays as history.
Vendor notice →The same record is available as structured data through the AI Demos MCP server, with the counts, the coverage and every per-test-case reason carried as fields.
Verify the proof files
These files support this result. Open a file to inspect the original evidence.
File fingerprints (SHA-256)
A fingerprint identifies the exact file used for this result.