Your First Assessment

Goal: evidence in, report out, findings you can read in fifteen minutes.

1. Get evidence in

Runtime mode — clients already point at the gateway (Deployment model): let real traffic flow. A few minutes of a busy workflow is enough for a first pass.

Assessment mode — no production rollout required: Blackridge connects the runtime evidence store to the investigation API so the assessment can be read through the same GraphQL-backed surface used by the UI.

2. Generate the report

Generate the assessment from the investigation API over the captured evidence window. Treat the assessment as an evidence artifact: inspect data quality before acting on findings, and use the linked request evidence when a number needs to be defended.

3. Read it

The report leads with a headline and top action, then breaks down data quality and coverage, waste by category (retry amplification, fallback tax, failed waste, abandoned branches, cache-miss opportunity, unknown/unattributed), workflow lineage, and fix-first recommendations with request-level evidence. Read data quality first — it tells you how much to trust everything below it. Customer docs give the exact field names for scripting against the report.

How each category is computed, and its limits: Finding methodology.

4. Investigate in the UI

The investigation UI gives you the assessment spine, ranked findings with evidence drawers, workflow replay, and the request ledger. These are the surfaces used during a guided assessment and in customer deployments.

If the report is empty

• The evidence store contains captured request records.
• The selected tenant matches the captured request attribution.
• The investigation API is healthy and can read the evidence store.
• The investigation API is pointed at the same evidence store as the runtime path.
• The selected tenant and time window contain captured request evidence.
• Still stuck: Empty reports.

Make the second assessment better

First assessments usually show attribution gaps — that's the product working. Each gap names its fix: enable a zero-code resolver, declare workflow context via an SDK scope, or add pricing for unpriced models. Coverage goes up; evidence grades go up with it.