Dashboard Data

The Dashboard Data API provides pre-aggregated analytics for the platform’s monitoring dashboards. Data is aggregated via SQL to avoid loading large result sets.

GET /api/v1/dashboard-data

Retrieve dashboard data by type.

Query Parameters

Parameter	Type	Required	Default	Description
`type`	string	No	`summary`	Data type to retrieve (see table below)
`days`	integer	No	`7`	Lookback window in days (1-365), used by time-based types
`modelId`	string	No	—	Algorithm model ID (only for `model_metrics_trend`)
`groupBy`	string	No	`channel` for `offer_performance_grouped`, `null` for `score_distribution`	Grouping dimension. Accepts `channel`, `category`, or `segment` for `offer_performance_grouped`; `channel`, `offer`, or `category` for `score_distribution`
`filterId`	string	No	—	Filter by dimension ID (only for `score_distribution`)
`segmentId`	string	No	—	Restrict the result to customers in the given segment. Accepted by `acceptance_rate`, `offer_performance`, `offer_performance_grouped`, `channel_effectiveness`, `daily_trend`, `revenue_trend`, `selection_frequency`, and `why_not_ranked`. Returns `{"data": [], "warning": "..."}` when the segment does not exist or the segment view has not been materialized yet.
`channelId`	string	No	—	Restrict `selection_frequency` to decisions made on a specific channel.
`categoryId`	string	No	—	Restrict `selection_frequency` to offers in a specific category.
`decisionFlowId`	string	No	—	Restrict `selection_frequency` to decisions produced by a specific decision flow.
`offerId`	string	Required for `why_not_ranked`	—	Target offer for the why-not analysis.
`baselineDays`	integer	No	same as `days`	For `anomaly_candidates` — the length of the prior-period window compared against the current window.

Data Types

Type	Description
`summary`	High-level counts: active offers, channels, experiments, journeys, triggers, pending approvals
`acceptance_rate`	Acceptance rate per offer (impressions vs positive outcomes) over `days` window
`offer_performance`	Top 20 active offers with impressions, conversions, revenue, and conversion rate
`channel_effectiveness`	Impressions, conversions, and effectiveness rate per channel
`budget_burn`	Budget allocation utilization: allocated, spent, remaining, burn rate
`funnel`	Offer funnel stages: total offers, active, with creatives, total creatives
`attribution_summary`	Attribution results grouped by model with conversion counts and total value
`daily_trend`	7-day trend of impressions and conversions
`model_metrics_trend`	AUC trend and feature importance from algorithm models (accepts `modelId`)
`latency_trend`	7-day p50/p99 latency percentiles from run execution data
`offer_performance_grouped`	Offer performance grouped by `channel` or `category` (accepts `groupBy`), summed from daily interaction summaries within the `days` window; conversion rate is clamped to 100%
`revenue_trend`	Daily revenue trend from interaction summaries over `days` window
`policy_impact`	Active suppressions grouped by policy type with counts
`score_distribution`	Propensity score histogram from decision traces (accepts `groupBy` and `filterId`)
`degraded_scoring`	Count of degraded scoring events (fallback scoring used due to model errors or timeouts)
`selection_frequency`	Per-offer eligible / scored / selected counts, selection rate, average rank, and rank distribution over the window
`anomaly_candidates`	Core-metric anomalies (acceptance rate, revenue, degraded scoring rate) vs a prior-period baseline — stateless; no alert firing
`why_not_ranked`	For a target `offerId`, breakdown of how often the offer was eligible but not selected and the top reasons

Response — `summary`

{
  "activeOffers": 15,
  "activeChannels": 4,
  "activeExperiments": 2,
  "totalJourneys": 3,
  "activeTriggers": 5,
  "pendingApprovals": 1
}

All entity counts (activeOffers, activeChannels, activeTriggers) exclude soft-deleted rows, so they always match what the corresponding list pages show. Soft-delete only sets a deletedAt timestamp — it does not change status — so these counts filter on deletedAt IS NULL in addition to status.

Response — `acceptance_rate`

{
  "data": [
    {
      "offerId": "off_001",
      "offerName": "Premium Card Upgrade",
      "impressions": 1250,
      "positive": 87,
      "negative": 45,
      "acceptanceRate": 0.0696
    }
  ],
  "total": 15,
  "period": {
    "days": 7,
    "since": "2026-03-23T00:00:00.000Z"
  }
}

Response — `offer_performance`

{
  "data": [
    {
      "offerId": "off_001",
      "name": "Premium Card Upgrade",
      "priority": 90,
      "creativeCount": 3,
      "impressions": 1250,
      "conversions": 87,
      "revenue": 4350.00,
      "conversionRate": "7.0"
    }
  ]
}

Response — `channel_effectiveness`

{
  "data": [
    {
      "channelId": "ch_001",
      "channelName": "Marketing Email",
      "impressions": 5000,
      "conversions": 350,
      "effectiveness": "7.0"
    }
  ]
}

Response — `budget_burn`

{
  "data": [
    {
      "offerId": "off_001",
      "allocated": 10000,
      "spent": 4500,
      "remaining": 5500,
      "burnRate": "45.0"
    }
  ]
}

Response — `funnel`

{
  "stages": [
    { "label": "Total Offers", "count": 25 },
    { "label": "Active Offers", "count": 15 },
    { "label": "With Creatives", "count": 12 },
    { "label": "Creatives", "count": 34 }
  ]
}

Every funnel stage counts live entities only (deletedAt IS NULL). “Total Offers” is all non-deleted offers, not literally every row ever created — soft-deleted offers and creatives are excluded from all four stages so the funnel stays consistent with the offers and creatives list pages. The offer_performance list is filtered the same way.

Response — `attribution_summary`

{
  "data": [
    {
      "model": "last_touch",
      "conversions": 150,
      "totalValue": 45000.00,
      "avgValue": 300.00
    }
  ],
  "totalConversions": 150
}

Response — `daily_trend`

{
  "data": [
    { "name": "Mar 24", "impressions": 340, "conversions": 22 },
    { "name": "Mar 25", "impressions": 420, "conversions": 31 },
    { "name": "Mar 26", "impressions": 380, "conversions": 28 }
  ]
}

Response — `model_metrics_trend`

{
  "aucTrend": [
    { "name": "Mar 15", "auc": 0.82 },
    { "name": "Mar 16", "auc": 0.84 }
  ],
  "featureImportance": [
    { "name": "tenure_months", "value": 0.32 },
    { "name": "monthly_spend", "value": 0.28 }
  ],
  "modelCount": 2,
  "modelName": "Propensity to Buy"
}

Response — `latency_trend`

{
  "data": [
    { "name": "Mon", "p50": 45, "p99": 220 },
    { "name": "Tue", "p50": 42, "p99": 195 }
  ],
  "hasRealData": true
}

Response — `offer_performance_grouped`

{
  "groups": [
    {
      "groupName": "Marketing Email",
      "groupId": "ch_001",
      "offers": [
        {
          "offerId": "off_001",
          "name": "Premium Card Upgrade",
          "impressions": 1250,
          "conversions": 87,
          "revenue": 4350.00,
          "conversionRate": "7.0"
        }
      ]
    }
  ],
  "groupBy": "channel"
}

Response — `revenue_trend`

{
  "data": [
    { "name": "Mar 1", "revenue": 1200.50 },
    { "name": "Mar 2", "revenue": 980.00 }
  ],
  "hasRealData": true
}

Response — `policy_impact`

{
  "data": [
    { "policyType": "frequency_cap", "count": 45 },
    { "policyType": "cooldown", "count": 12 }
  ],
  "totalActive": 57
}

Response — `score_distribution`

{
  "data": [
    { "name": "0.0-0.1", "count": 15 },
    { "name": "0.1-0.2", "count": 22 },
    { "name": "0.2-0.3", "count": 38 },
    { "name": "0.3-0.4", "count": 45 },
    { "name": "0.4-0.5", "count": 52 },
    { "name": "0.5-0.6", "count": 67 },
    { "name": "0.6-0.7", "count": 48 },
    { "name": "0.7-0.8", "count": 35 },
    { "name": "0.8-0.9", "count": 20 },
    { "name": "0.9-1.0", "count": 8 }
  ],
  "totalScores": 350,
  "hasRealData": true,
  "groupBy": null,
  "dimensionBuckets": [
    {
      "name": "ch_001",
      "dimName": "Marketing Email",
      "count": 200,
      "avgScore": 0.542
    }
  ]
}

The score_distribution type accepts groupBy values of channel, offer, or category to break down scores by dimension. Use filterId to show only scores for a specific dimension value.

Response — `degraded_scoring`

{
  "degradedCount": 3,
  "totalCount": 1250,
  "days": 7,
  "hasData": true
}

Returns the count of recommend calls where the scoring engine fell back to degraded mode (e.g., model timeout, missing data). totalCount is the total number of scored requests in the period.

Response — `selection_frequency`

{
  "data": [
    {
      "offerId": "off_premium_card",
      "offerName": "Premium Card Upgrade",
      "eligibleCount": 340,
      "scoredCount": 340,
      "selectedCount": 48,
      "selectionRate": 0.141,
      "avgRank": 2.3,
      "rankDistribution": [48, 112, 85, 42, 30, 15, 6, 2, 0, 0]
    }
  ],
  "totalTraces": 412,
  "period": { "days": 7, "since": "2026-03-23T00:00:00.000Z" }
}

Aggregates DecisionTrace.scoringResults[] and selectedOffers[] across the window. The trace sample is capped at 10,000 rows per call for latency. rankDistribution is a 10-element array covering ranks 1..10; ranks beyond 10 are not tracked. Supports optional channelId, categoryId, decisionFlowId, and segmentId filters. Example:

curl -H "X-API-Key: $API_KEY" -H "X-Tenant-Id: $TENANT_ID" \
  "https://playground.kaireonai.com/api/v1/dashboard-data?type=selection_frequency&days=14&segmentId=seg_vip"

Response — `anomaly_candidates`

{
  "data": [
    {
      "metric": "acceptance_rate",
      "dimension": "offer",
      "dimensionKey": "off_premium_card",
      "dimensionName": "Premium Card Upgrade",
      "baseline": 0.102,
      "current": 0.184,
      "deltaPct": 80.4,
      "zscore": 4.3,
      "severity": "critical"
    }
  ],
  "period": {
    "currentDays": 7,
    "baselineDays": 7,
    "since": "2026-03-23T00:00:00.000Z"
  }
}

Severity is classified by the larger of the z-score magnitude and the absolute percent change: info at |z| >= 2 or |%| >= 15, warning at |z| >= 3 or |%| >= 30, critical at |z| >= 4 or |%| >= 50. Metrics covered: acceptance_rate (overall, per-offer, per-channel), revenue (per-offer), and degraded_scoring_rate (overall). This endpoint is stateless — it does not fire alerts; it surfaces candidates that a scheduler in a later phase may escalate. Example:

curl -H "X-API-Key: $API_KEY" -H "X-Tenant-Id: $TENANT_ID" \
  "https://playground.kaireonai.com/api/v1/dashboard-data?type=anomaly_candidates&days=7&baselineDays=7"

Response — `why_not_ranked`

{
  "offerId": "off_premium_card",
  "offerName": "Premium Card Upgrade",
  "timesEligible": 340,
  "timesSelected": 48,
  "missRate": 0.859,
  "missReasons": {
    "scoredTooLow": 220,
    "filteredByContactPolicy": 45,
    "filteredByQualification": 27,
    "beatenBy": [
      { "offerId": "off_gold_plus", "offerName": "Gold Plus", "count": 102 },
      { "offerId": "off_starter", "offerName": "Starter Card", "count": 48 }
    ]
  },
  "rankDistribution": [48, 112, 85, 42, 30, 15, 6, 2, 0, 0],
  "avgScore": 0.432,
  "period": { "days": 7, "since": "2026-03-23T00:00:00.000Z" },
  "sampleSize": 412,
  "sampleCap": 1000
}

For a given target offer, walks the most recent traces (capped at sampleCap) and attributes every miss to one of: contact-policy block, qualification rejection, or scored-but-outranked. When the offer was scored but not selected, the winning offer in that decision is counted toward the beatenBy list (top 5). The offerId query parameter is required; optional segmentId narrows the analysis to customers in that segment. Example:

curl -H "X-API-Key: $API_KEY" -H "X-Tenant-Id: $TENANT_ID" \
  "https://playground.kaireonai.com/api/v1/dashboard-data?type=why_not_ranked&offerId=off_premium_card&days=14"

Segment dimension on existing performance endpoints

acceptance_rate, offer_performance, offer_performance_grouped, channel_effectiveness, daily_trend, and revenue_trend all accept an optional segmentId query parameter. When provided, the underlying aggregation is restricted to interactions from customers in the segment’s materialized view. offer_performance_grouped&groupBy=segment fans the result out by active segment instead of channel/category; each offer row is decorated with segmentId and segmentName. If the segment is unknown or its view has not been materialized yet, the endpoint returns {"data": [], "warning": "..."} instead of an error, so dashboards can render a friendly notice.

Error codes

Code	Reason
`400`	Unknown `type` value.
`403`	Insufficient role permissions.

Role requirements

Method	Minimum role
GET	`viewer`

Overview

Decisioning

Studio

Data

Algorithms & Models

Content & Metrics

Orchestration

Customers & Traces

Events & Attribution

Analytics

AI

Testing & Simulation

Export & Import

Admin

Integrations

GET /api/v1/dashboard-data

Query Parameters

Data Types

Response — `summary`

Response — `acceptance_rate`

Response — `offer_performance`

Response — `channel_effectiveness`

Response — `budget_burn`

Response — `funnel`

Response — `attribution_summary`

Response — `daily_trend`

Response — `model_metrics_trend`

Response — `latency_trend`

Response — `offer_performance_grouped`

Response — `revenue_trend`

Response — `policy_impact`

Response — `score_distribution`

Response — `degraded_scoring`

Response — `selection_frequency`

Response — `anomaly_candidates`

Response — `why_not_ranked`

Segment dimension on existing performance endpoints

Error codes

Role requirements

​GET /api/v1/dashboard-data

​Query Parameters

​Data Types

​Response — summary

​Response — acceptance_rate

​Response — offer_performance

​Response — channel_effectiveness

​Response — budget_burn

​Response — funnel

​Response — attribution_summary

​Response — daily_trend

​Response — model_metrics_trend

​Response — latency_trend

​Response — offer_performance_grouped

​Response — revenue_trend

​Response — policy_impact

​Response — score_distribution

​Response — degraded_scoring

​Response — selection_frequency

​Response — anomaly_candidates

​Response — why_not_ranked

​Segment dimension on existing performance endpoints

​Error codes

​Role requirements

GET /api/v1/dashboard-data

Query Parameters

Data Types

Response — `summary`

Response — `acceptance_rate`

Response — `offer_performance`

Response — `channel_effectiveness`

Response — `budget_burn`

Response — `funnel`

Response — `attribution_summary`

Response — `daily_trend`

Response — `model_metrics_trend`

Response — `latency_trend`

Response — `offer_performance_grouped`

Response — `revenue_trend`

Response — `policy_impact`

Response — `score_distribution`

Response — `degraded_scoring`

Response — `selection_frequency`

Response — `anomaly_candidates`

Response — `why_not_ranked`

Segment dimension on existing performance endpoints

Error codes

Role requirements