Enterprise AI Administration
Enterprise Variables
Governed variables and features: definition, source, grain, transformation and data quality.
Variables
27
Passing quality checks
24
Quality warnings
2
Source systems
12
Governed variables
| Variable | Source | Type | Grain | Transformation | Used by | Quality | |
|---|---|---|---|---|---|---|---|
| critical_alarm_countCritical alarms raised on a site, last 7 days. | OSS / NOC | integer | Site-day | count(alarm) WHERE severity = CRITICAL | Network Incident Risk Model | PASSED | View details → |
| average_monthly_revenueAverage billed service revenue over the last 3 months. | Billing | numeric | Subscriber-month | avg(service_revenue) over 3 billing periods | ARPU Uplift Model; Revenue & Margin Forecast | PASSED | View details → |
| capacity_utilization_pctBusy-hour capacity utilisation of a site. | OSS / NOC | numeric | Site-hour | busy_hour_traffic ÷ site_capacity × 100 | Network Incident Risk Model; Congestion Risk Model; Energy Cost Forecast | PASSED | View details → |
| churn_propensityChurn Propensity Model score (0–1). | Platform model output | numeric | Subscriber-month | Model score, refreshed weekly | Net Adds Forecast | PASSED | View details → |
| Complaint count 30dComplaints in last 30 days | Trouble Ticketing | integer | Subscriber-day | count of complaint cases in rolling 30 days | Churn Propensity Model; NPS Driver Model | PASSED | View details → |
| Site congestion indexPeak utilisation score per site | OSS / NOC | decimal | Site-hour | peak PRB utilisation normalised 0–1 | Network Incident Risk Model; Congestion Risk Model | PASSED | View details → |
| data_usage_change_30d_pctChange in data MB, last 30 days vs prior 30 days. | Mediation / CDR | numeric | Subscriber-day | (Σ data_mb 30d − Σ prior 30d) ÷ Σ prior 30d × 100 | Churn Propensity Model; ARPU Uplift Model | PASSED | View details → |
| Dropped call rateShare of the subscriber's calls dropped | Mediation / CDR | decimal | Subscriber-month | dropped calls / total calls × 100 | Churn Propensity Model; NPS Driver Model | PASSED | View details → |
| Digital activity scoreApp and wallet engagement score | Digital App / Channels | decimal | Subscriber-month | weighted logins, transactions and self-service events | Digital Adoption Model; ARPU Uplift Model | PASSED | View details → |
| Data session success rateSuccessful data sessions share | Mediation / CDR | decimal | Subscriber-day | successful sessions / total sessions × 100 | NPS Driver Model | WARNING | View details → |
| duplicate_recharge_countRecharges reusing a voucher or repeating within 60 seconds, last 7 days. | Recharge | integer | Subscriber-day | count(recharge) GROUP BY voucher_id HAVING count > 1 | Fraud Risk Classifier; Revenue Leakage Detection | PASSED | View details → |
| energy_cost_variance_pctEnergy cost versus budget for a site or region. | Finance ERP / GL | numeric | Site-day | (energy_actual − energy_budget) ÷ energy_budget × 100 | Energy Cost Forecast; Revenue & Margin Forecast | PASSED | View details → |
| fraud_anomaly_scoreUnsupervised anomaly score on calling and recharge behaviour (0–1). | Fraud Monitoring | numeric | Subscriber-day | Isolation-forest score on usage and recharge features | Fraud Risk Classifier; Revenue Leakage Detection | PASSED | View details → |
| network_quality_scoreExperienced network quality for the subscriber's serving sites (0–100). | OSS / NOC | numeric | Subscriber-day | Traffic-weighted blend of site availability, dropped calls and throughput | Churn Propensity Model; NPS Driver Model | PASSED | View details → |
| offer_response_scoreLikelihood of accepting a retention or upsell offer (0–1). | CRM | numeric | Subscriber-month | Campaign response history weighted by recency | ARPU Uplift Model; Net Adds Forecast | IN REVIEW | View details → |
| Outage countOutages per site in last 30 days | OSS / NOC | integer | Site-hour | count of outage events in rolling 30 days | Network Incident Risk Model | PASSED | View details → |
| Payment timelinessShare of postpaid bills paid on time | Billing | decimal | Account-month | on-time payments / bills issued | Churn Propensity Model | PASSED | View details → |
| Plan typePrepaid, hybrid or postpaid | CRM | category | Subscriber-month | latest plan on record | ARPU Uplift Model; Churn Propensity Model | PASSED | View details → |
| Average recharge amountMean recharge value | Recharge platform | decimal (ZAR) | Subscriber-month | sum recharge amount / recharge count | ARPU Uplift Model | PASSED | View details → |
| Recharge frequencyRecharges per 30 days | Recharge platform | integer | Subscriber-month | count of recharges in rolling 30 days | Churn Propensity Model; ARPU Uplift Model | PASSED | View details → |
| revenue_at_riskMonthly revenue of a subscriber when churn propensity is in the high-risk band. | Billing | numeric | Subscriber-month | average_monthly_revenue × (churn_propensity ≥ 0.65) | Revenue & Margin Forecast | PASSED | View details → |
| Roaming usage flagAny roaming usage in month | Mediation / CDR | boolean | Subscriber-month | 1 if roaming usage > 0 | Roaming Revenue Forecast | PASSED | View details → |
| Customer experience sentiment scoreSentiment of care interactions | Contact Center | decimal | Subscriber-month | average sentiment −1 to +1 | NPS Driver Model | WARNING | View details → |
| Tenure monthsMonths since activation | CRM | integer | Subscriber-month | months between activation date and period end | Churn Propensity Model | PASSED | View details → |
| 30-day usage decline %Change in usage vs prior 30 days | Mediation / CDR | decimal | Subscriber-day | (usage last 30d − prior 30d) / prior 30d × 100 | Churn Propensity Model | PASSED | View details → |
| usage_pattern_deviationDeviation of current usage from the subscriber's 90-day profile (z-score). | Mediation / CDR | numeric | Subscriber-day | (usage_today − mean_90d) ÷ stddev_90d | Fraud Risk Classifier; Revenue Leakage Detection | PASSED | View details → |
| voice_usage_change_30d_pctChange in voice minutes, last 30 days vs prior 30 days. | Mediation / CDR | numeric | Subscriber-day | (Σ voice_minutes 30d − Σ prior 30d) ÷ Σ prior 30d × 100 | Churn Propensity Model | PASSED | View details → |