# Compare a task classifier with a simple base-rate rule

Illustrative planning brief; no automatic product import.

Synthetic planning case: A ten-case evaluation contains eight not-ready outcomes and two ready outcomes. Rule A always predicts not ready and advertises 80% accuracy as evidence that it detects readiness.

## Decision

Rule A matches eight of ten labels but detects zero of two ready cases. Report the majority-outcome baseline and the required ready-case result before treating overall accuracy as useful.

## Owned work

- Preserve class counts
  - Owner role: Evaluation owner
  - Acceptance evidence: The eight/two split is retained with the ten-case cohort.
- Write the constant reference rule
  - Owner role: Reviewer
  - Acceptance evidence: Always-not-ready is an explicit comparator rather than hidden in an accuracy claim.
- Check the decision-relevant class
  - Owner role: Planning lead
  - Acceptance evidence: The two required ready detections remain visible before selecting the rule.

## Workflow

1. Preserve class counts. Check: The eight/two split is retained with the ten-case cohort.
2. Write the constant reference rule. Check: Always-not-ready is an explicit comparator rather than hidden in an accuracy claim.
3. Check the decision-relevant class. Check: The two required ready detections remain visible before selecting the rule.

## Judgment

This tiny synthetic cohort does not validate a classifier, class distribution or live task readiness model.


## Filled manual planning note

Rule A matches eight of ten labels but detects zero of two ready cases. Report the majority-outcome baseline and the required ready-case result before treating overall accuracy as useful. Overall accuracy is 8/10=80%. Ready-case detection is 0/2=0%. The high accuracy follows the observed class mix and the constant rule, not evidence that A recognizes a ready case. This tiny synthetic cohort does not validate a classifier, class distribution or live task readiness model.


## Workflow questions

### Is 80% accuracy necessarily bad?

Its usefulness depends on the actual decision and error consequences; it does not alone demonstrate readiness detection.

### Does the eight/two split apply to future tasks?

No. It describes only these invented evaluation cases.

## Product connection

Use the owned checks and downloaded brief to discuss this planning decision alongside your TeamBoostAI tasks. Confirm available fields, roles and account features separately. The example is manual; it does not calculate live analytics, create work or run an experiment in the product.

Confirm account availability before adopting this manual outline.

## Original worked case

Synthetic records, manual planning only. No account import or live analytics.

### Inspect the invented case records

Observed class | Cases | A predicts | Correct | Ready cases detected
--- | --- | --- | --- | ---
Not ready | 8 | Not ready | 8 | Not applicable
Ready | 2 | Not ready | 0 | 0 of 2

### Reasoning

Overall accuracy is 8/10=80%. Ready-case detection is 0/2=0%. The high accuracy follows the observed class mix and the constant rule, not evidence that A recognizes a ready case.

### Bounded result

Rule A matches eight of ten labels but detects zero of two ready cases. Report the majority-outcome baseline and the required ready-case result before treating overall accuracy as useful.

### Distinct decision

The artifact tests an apparently strong accuracy claim against an explicit majority-class reference.

### Limits

This tiny synthetic cohort does not validate a classifier, class distribution or live task readiness model.

### Definitions and method context

- Original worked-case definitions — Definitions, policy choices, records and calculations are authored for this worksheet. No external standard, statistical validation or live product measurement is claimed.