Reading verification analytics and pass rates
The Verification analytics dashboard shows how effectively G8KEPR is inspecting traffic to and from your AI, API, and MCP endpoints. It breaks results down by verification layer so you can see what each stage caught, what it let through, and how those rates trend over time.
Two metrics to know
Most charts on this page are built from two complementary numbers:
- ▸Catch rate — the share of inspected requests or responses that a layer flagged or blocked. A higher catch rate means a layer is actively intervening.
- ▸Pass rate — the share that a layer cleared without action. A high pass rate is normal for healthy traffic; a sudden drop usually signals an incident or a misconfigured policy.
Catch rate and pass rate are complements. Across a single layer they add up to roughly 100% of the traffic that layer evaluated.
Reading the by-layer breakdown
Each request passes through verification layers in sequence, and the dashboard reports each one separately so you can pinpoint where issues are caught:
- ▸Input verification — screens incoming prompts and payloads before they reach your model or tool.
- ▸Output verification — inspects model and API responses before they are returned.
- ▸Policy and threat checks — apply your configured rules and threat-pattern matching.
- 1.Open the Verification section of your dashboard.
- 2.Choose a time range (for example, last 24 hours or last 7 days) to scope every chart on the page.
- 3.Read the per-layer bars or lines to compare catch rates across layers.
- 4.Hover any data point to see the exact counts behind the percentage.
Interpreting trends
- ▸A rising catch rate on one layer often means new threats or a policy you recently tightened.
- ▸A falling pass rate across all layers at once points to a traffic-wide event worth investigating.
- ▸Layers with a near-100% pass rate and near-zero catches are working as expected on clean traffic.
Compare the same time range week over week to separate normal variation from a real shift in behavior.
Next steps
- ▸Drill into flagged requests from the Verification logs.
- ▸Adjust what each layer enforces in Verification policies.
If a rate looks wrong or a chart is empty, confirm the selected time range has traffic, then contact support.