Review candidate evidence for Ofsted Essex delay slice

[happygeneralist]

Scope

Add the Stage C candidate evidence review for issue #91. This PR contains no structured evidence or knowledge-object changes.

This draft depends on the proposed slice plan in PR #93. It should not be merged ahead of that plan without reconciling the boundary first.

Candidate-set result

The review identifies:

• six proposed new evidence records;
• one existing evidence record to reuse rather than duplicate;
• one source-attributed causal contributor to retain as context unless later analysis needs a separate record.

The proposed selected candidates cover:

1. assessment-of-needs delay and the inspection's asserted inconsistency consequence;
2. poor EHC plan timeliness;
3. one bundled finding about waiting for appropriate provision and support;
4. family dissatisfaction attributed to the paragraph's delay cluster;
5. too few EHC plans within the statutory timeframe;
6. inconsistency in support attributed to EHC plan delays.

Material findings

• Page 2 says assessment of needs; it should not be silently narrowed to EHC needs assessment.
• The explicit page 5 EHC needs-assessment claim is already preserved in EVID_003, so a duplicate record should not be created.
• provision and support should remain one evidence unit because the source does not support a clean split.
• The different delay conditions should not be collapsed into one broad SEND delays pain point.
• The selected evidence does not directly support PP_002, UN_015, UN_007 or BEH_009.
• No user need, civic need, insight or pain point should be created in the evidence PR.

What changed

• Adds one candidate evidence table with source locations, claim types, actor and context, direct support, non-support, nearby-object fit, recommendation, confidence, uncertainty and operating questions.
• Records two public modelling findings for Define evidence-context-object and non-linear journey modelling guardrails #87: source-led granularity and reuse of multi-claim evidence across slices.

What did not change

• No evidence records were created.
• No existing evidence or knowledge object was edited.
• No source evidence map was changed.
• No schema, template, validator, maturity, confidence or review state changed.
• No protected rubric was added or exposed.

Expected validation effect

A later selected-evidence PR should be able to create only the six strongest non-duplicative records, preserve claim-level boundaries and avoid premature object formation.

Related work

• Run bounded Ofsted Essex delay ingestion slice #91
• Develop safer repeatable non-linear ingestion workflow #86
• Define evidence-context-object and non-linear journey modelling guardrails #87
• PR Plan bounded Ofsted Essex delay slice #93
• happygeneralist/design-intelligence-orchestration#10