Can Anomaly AI put several findings into one report that still agrees with each answer?
Anomaly AI combined its three earlier answers into one report and a saved dashboard without changing any of them: August revenue of $8,795.45, Outdoor & Patio as the best-selling category at $1,806, and a 21.75% cancellation rate, 67 of 308 orders. The saved dashboard reopens from the Dashboards list into a new chat session.
1 of 1 test case passed
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.
Three findings assembled into one artifactPassEvidence
Input 1: “As of September 2, 2026, what was our total revenue last month?”
Input 2: “What's our best-selling product category?”
Input 3: “What's our current cancellation rate?”
Input 4: “Put all of that together into one report.”
Answer: August revenue
Your total revenue for August 2026 was $8,795.45 USD, as of the end of September 2, 2026 in your local timezone. This is booked revenue, excluding orders cancelled by that cutoff and before deducting refunds. Historical statuses were used; later changes to order amounts cannot be reconstructed from the available records.
Answer: best-selling category
Outdoor & Patio was your best-selling category by revenue, with $1,806 USD, followed by Storage & Organization at $1,736.75 USD. This continues the previous question's scope: August 2026, as of September 2, excluding cancelled orders and before refunds. Categories are based on each order's recorded category; "best-selling" here means revenue rather than units sold.
Answer: cancellation rate
Your current cancellation rate is 21.75% — 67 cancelled orders out of 308 total orders. This covers all recorded orders across all dates, using their current status. Data was last updated today, November 2, 2026, at 21:29 in your local timezone.
Answer: the combined report
Created Revenue, Best-Selling Categories and Cancellations Report, bringing together: August revenue: $8,795.45 USD Best-selling category: Outdoor & Patio, with $1,806 USD Current cancellation rate: 21.75% — 67 of 308 orders The report clearly separates August results as of September 2 from current, all-time cancellation figures.
Transcribed from Proof 1 · Screen recording at 0:34
Every message the tool sent, in order (4)
- Answer: August revenue
Printed above.
- Answer: best-selling category
Printed above.
- Answer: cancellation rate
Printed above.
- Answer: the combined report
The final response, printed above.
| ✓ | Expected: One report holding all three findings, each still agreeing with the answer to the same question asked directly.Found: The report repeats $8,795.45, Outdoor & Patio at $1,806 and 21.75% (67 of 308); the dashboard shows 8,795.45 USD, a 1.81K USD top bar and a 21.75% donut. |
It is a Pass: one artifact had to hold all three findings, each agreeing with its direct answer. The report and dashboard repeat the three figures; the dashboard's 21.8% tile is 67 of 308 shown to one decimal, and its 1.81K bar label is 1,806 rounded.
Best-selling read as revenue. The agent took best-selling to mean August booked revenue rather than units, said so in its answer and in the dashboard narrative, and offered a button to show categories by units; the report itself shows only Outdoor & Patio.
Pro plan on a trial account. The run was made on the Pro plan held on a trial account: the account badge under the avatar in the left rail reads TRIAL throughout.
Dashboard tools offered. The saved dashboard offers refreshing the PostgreSQL data, scheduling, report generation, PDF, sharing, refine, verify and publish, and a per-component Edit with Anomaly dialog, which was opened but not sent.
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 | Reporting → | |
| Scenario | The user wants several findings in one place → | |
| Tool | Anomaly AI → |
Global scenario definition → · Global capability definition → · Anomaly AI 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.