Skip to main content
The Live Console is KaireonAI’s real-time monitoring dashboard for day-to-day decisioning operations. It brings together live decision metrics, customer-level diagnostics, and action analysis (Why-Not) in a single interface.
Access the Live Console at Dashboards > Live Console or navigate directly to /operations/live in the platform.

Business Value

PersonaHow the Ops Manager Helps
Operations analystMonitor decision throughput, error rates, and latency in real time
Marketing managerInvestigate why specific campaigns are under-delivering against targets
Support agentLook up a customer and see exactly which offers were shown (or suppressed) and why
Data scientistValidate that model scores are flowing through the pipeline and producing expected rankings

Key Capabilities

Live Decision Metrics

The dashboard shows real-time aggregates for the current tenant:
MetricDescription
Decisions/minNumber of Recommend API calls processed per minute
Avg latencyMean decision pipeline response time in milliseconds
Error ratePercentage of decisions that returned a 5xx error
Suppression ratePercentage of candidate offers suppressed by contact policies
Budget utilizationCurrent spend vs. allocated budget across active campaigns

Customer Lookup

Enter a customer ID to view their complete decisioning context:
  • Unified Profile — demographics, interaction history, CLV score, active suppressions
  • Recent Decisions — the last 20 decision traces with timestamps and outcomes
  • Active Journeys — journey enrollments and current step
  • Segment Memberships — which segments the customer belongs to

Action Analysis (Why-Not)

From the customer lookup, select any offer to run a Why-Not analysis. The Ops Manager renders the results inline:
  • Qualification rule pass/fail breakdown with reasons
  • Contact policy evaluation with usage counts vs. caps
  • Customer attribute values used in each evaluation
  • Interaction history summary (impressions, last contact)
Why-Not analysis runs in real time against current rules and data. If you changed a qualification rule 5 minutes ago, the Why-Not result reflects the new configuration.

Cross-Channel View

The operations console shows decisions across all channels:
ViewWhat It Shows
By channelDecision volume and conversion rate per delivery channel (email, push, SMS, in-app)
By offerTop offers by impression count and conversion rate
By flowDecision flow execution counts and average latency
SuppressionsOffers suppressed by policy type (frequency cap, cooldown, budget exhausted)

How It Works

The Operations Manager pulls data from several platform subsystems:
1

Decision traces

The Recommend API records sampled traces (configurable sample rate per tenant) that capture the full pipeline execution, including which candidates entered, which were filtered, and the final ranking.
2

Interaction summaries

Aggregated interaction data (impressions, conversions, value) per customer-offer pair, updated by the Respond API.
3

Real-time metrics

Prometheus-compatible metrics exposed by the API instrumentation layer (withMetrics wrapper on every route handler).
4

Why-Not API

On-demand diagnostic endpoint that replays the decision pipeline for a specific customer-offer pair.

Configuration

Enable Decision Tracing

Decision tracing must be enabled for the Ops Manager to show pipeline details:
  1. Go to Settings > Tenant Settings
  2. Enable Decision Trace
  3. Set the Sample Rate (e.g., 0.1 for 10% of decisions)
Setting the sample rate to 1.0 (100%) captures every decision. This is useful for debugging but increases storage usage and may impact performance at high volumes. For production, 5-10% is recommended.

Dashboard Refresh

The live dashboard auto-refreshes every 10 seconds. Metrics are computed from the most recent 5-minute window for decisions/min and latency, and from rolling 24-hour windows for utilization and suppression rates.

Supported Workflows

Investigating Under-Delivery

1

Check the offer

Look up the offer in the Ops Manager. Check its status (active/paused), budget remaining, and total impressions.
2

Run Why-Not for sample customers

Pick 3-5 customers from the target audience and run Why-Not analysis for each. Look for patterns in blocking rules.
3

Review qualification rules

If a specific rule is blocking most customers, check whether the rule’s condition is too restrictive or the customer data is missing.
4

Check contact policies

If frequency caps are the primary blocker, consider adjusting the cap or switching to a different channel.

Real-Time Incident Response

When decision latency spikes or error rates increase:
  1. Open the Ops Manager live dashboard
  2. Check the Error rate and Avg latency tiles for anomalies
  3. Navigate to Decision Traces to find failing requests
  4. Examine the trace detail for the specific pipeline stage that failed
  5. Check the Dashboards for correlated infrastructure metrics

The Operations Manager UI is backed by these APIs:
APIPurpose
Decision TracesQuery and view pipeline execution traces
Why-NotOn-demand action analysis for customer-offer pairs
Unified ProfileCustomer lookup with all data sources merged
Metrics SummaryAggregated operational metrics
Dashboard DataPre-computed dashboard tile data

Next Steps

Why-Not Analysis

Deep dive into action analysis diagnostics.

Dashboards

Business and operational dashboards for reporting.

Decision Traces

Query raw decision trace records.

Unified Profile

Aggregated customer data from all sources.