Independent practice · Evidence review

Write a bounded evidence summary

Practice dataset

My saved work →

Your task

Return accuracy_percent (number), median_minutes (number), satisfaction_status (string), source_ids (array).
Use the fixed source IDs. Array order follows the contract.
Do not invent unsupported values. Missing evidence must remain unknown.

Pass every output check to complete this task. Suggested active time: 10 minutes. You can keep working after that.

Required output

Return one JSON object with exactly these fields.

accuracy_percentnumber
median_minutesnumber
satisfaction_statusstring
source_idsarray

Keep strings, numbers, booleans, arrays and null in their required JSON types. Include every field and no extra fields.

Use null only where the evidence and task rules require an unknown value.

Reconciliation checks

Return exactly the named columns in the stated order.
Preserve row multiplicity, numeric types and the declared missing-value policy.

Need a next step?

Hint 1: where to start

Copy the limitation exactly as required by this fixed output contract.

Hint 2: build the method

Check the declared business key, eligibility rules and missing-value policy before editing.

Hint 3: check your work

Compare the exact output columns, row count and each value with the contract; matching a total alone is insufficient.

How your work is checked

Every output check must pass. The score shows how much matched and helps you find mistakes; completion comes from passing all checks. Active time is recorded separately.

These checks compare fixed evidence fields and types. They do not assess arbitrary writing quality.

Passing records a correct practice output on this device. It is not a professional certification.

Source materials

3 original source records.

Data dictionary

sources[].id text: Stable evidence identifier.
sources[].text text: Authoritative fictional evidence for the fixed response fields.

View source data

{
  "sources": [
    {
      "id": "S1",
      "title": "Pilot brief",
      "text": "The fictional Northstar pilot starts 2026-11-02. It covers 12 support agents. The pilot lasts 14 calendar days. Budget is 2400 USD."
    },
    {
      "id": "S2",
      "title": "Review policy",
      "text": "Morgan owns the review. No customer personal data may enter the pilot. No automation may send customer replies without human approval."
    },
    {
      "id": "S3",
      "title": "Results memo",
      "text": "The test processed 80 synthetic tickets. 72 had the correct category. Median handling time was 6 minutes. Customer satisfaction was not measured."
    }
  ],
  "brief": "Return accuracy_percent (number), median_minutes (number), satisfaction_status (string), source_ids (array).",
  "flawed_output": "The system improved accuracy and satisfaction and cut time by half."
}

Your work

Active 0:00 · Paused 0s

JSON answer

Import answer

Keep a handoff you can replay

Check your current answer, then download the output and its report. The report includes the source, answer and each check so you can review what passed and what needs work.

To replay your work, open this task with the same dataset, import your answer and check it again.

Show the reference answer

{
  "accuracy_percent": 90,
  "median_minutes": 6,
  "satisfaction_status": "not measured",
  "source_ids": [
    "S3"
  ]
}

Use the measured results without inventing satisfaction or causal improvements.