Leaders Insights
Leaders Insights

Rester au meilleur niveau, un peu chaque jour.

DomainesMarketingDataFinanceIA
RessourcesApprendreTestOutilsBlogGlossaire
© 2026 Leaders Insights — Tous droits réservés.
Formations/Data in telecom/Data landscape, quality and metrics/Mapping the telecom data landscape beyond the CDR
1/5+150 XP

Data landscape, quality and metrics

5Mapping the telecom data landscape beyond the CDR+1506Data quality dimensions for subscriber and network records+1507Master data management across OSS, BSS, and network inventory+1508Benchmarking network and service KPIs that leadership tracks+1509Data lineage and pipeline reliability for telecom reporting+150

Mapping the telecom data landscape beyond the CDR

# Mapping the telecom data landscape beyond the CDR

A single 4G voice call can generate more than a dozen separate data records across a carrier's systems: a Call Detail Record (CDR) in billing, a session log on the Mobility Management Entity, a QoS (Quality of Service) trace on the radio controller, a roaming record if the subscriber is abroad, and a customer care note if the call dropped and the subscriber complained. Most telecom analytics training stops at the CDR. That is like studying a bank by only reading receipts. This lesson maps the fuller data estate: where it lives, who owns it, and where it quietly breaks.

Why the CDR is not enough

The CDR (Call Detail Record) captures the metadatametadataDonnées sur les données, informations décrivant le contexte, la structure, la provenance et les caractéristiques d'un asset de données (auteur, date, format, source, définition). of a voice call or SMS: caller, callee, duration, timestamp, cell tower. It is the historical backbone of telecom billing and fraud detection. For data (packet) sessions, the equivalent is the xDR or IPDR (Internet Protocol Detail Record).

But CDRs answer only "what happened on the network." They cannot tell you why a customer churned, whether a SIM was ever activated, or if a roaming partner is overbilling you. For that, you need the rest of the stack.

The core data estate

OSS: the network's operational truth

OSS (Operations Support Systems) manage the physical and logical network: inventory of base stations, fiber routes, spectrum allocation, fault and performance management. Key datasets:

  • Network inventory: every cell site, antenna, and piece of transmission equipment, with location and configuration.
  • Performance Management (PM) counters: aggregated KPIs like dropped call rate or Physical Resource Block utilization, collected every 15 to 60 minutes per cell.
  • Fault/alarm logs: hardware and software failures, often the first signal of a coming outage.

These feed capacity planning, capital expenditurecapital expenditureCapital Expenditure (CapEx) is money spent to acquire, upgrade, or extend long-lived assets like equipment, property, or software that deliver value over multiple years.Voir la définition complète → prioritization, and network health dashboards. Hidden gap: PM counters are vendor-specific (Ericsson, Nokia, Huawei all format differently), so consolidating a multi-vendor network requires a normalization layer, often the single biggest hidden cost in a network data warehousedata warehouseA central repository that consolidates data from many source systems into a structured, query-optimized store designed for analytics, reporting, and business intelligence.Voir la définition complète → project.

BSS: the customer and money truth

BSS (Business Support Systems) run the commercial side: customer relationship managementcustomer relationship managementCustomer Relationship Management: software and strategy to manage and analyse customer interactions throughout their lifecycle.Voir la définition complète →, billing, order management, product catalog. Key datasets:

  • Subscriber master records: identity, plan, tenure, contract status.
  • Billing and mediation data: usage rated into charges, tied to the CDR/xDR feed.
  • Order and provisioning logs: what was requested, activated, or churned, and when.

This is the primary feed for churn models, average revenue per user (ARPU) reporting, and customer lifetime valuecustomer lifetime valueLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.Voir la définition complète → work. Hidden duplication: the same subscriber often exists as multiple records across prepaid, postpaid, and IoT (Internet of Things) systems inherited from mergers, a classic source of double-counted subscriber bases in KPIKPIKey Performance Indicator, a measurable value that shows how effectively you're achieving a specific objective, tracked over time against a target.Voir la définition complète → reporting.

Network element logs: the granular layer

Below OSS aggregates sit raw logs from routers, switches, and radio equipment: signaling protocol traces (e.g., Diameter, SIP), deep packet inspection samples, and security event logs. These are voluminous (often petabytes per month for a national carrier) and rarely stored longer than days or weeks due to cost. They matter for root-cause analysis of outages and for training network AI/ML models (e.g., anomaly detection for self-organizing networks). Most of this data is generated and discarded before anyone outside network engineering ever sees it, a genuine visibility gap for commercial teams.

Roaming data: TAP and NRTRDE

When a subscriber uses a network abroad, the visited network sends usage data back to the home carrier for billing and fraud checks, via two GSMA (GSM Association) standards:

  • TAP (Transferred Account Procedure): batch files of rated roaming usage sent by the visited operator to the home operator, typically daily.
  • NRTRDE (Near Real-Time Roaming Data Exchange): a faster, unrated feed (within about 4 hours of usage) designed to catch high-usage fraud before it becomes a large unpaid bill.

Why this matters: roaming fraud and billing disputes between carriers run into real money industry-wide, and TAP/NRTRDE reconciliation is one of the more manual, error-prone data processes in telecom because it depends on dozens of bilateral partner formats. Reference: GSMA's public roaming and fraud resources outline the operational standards.

Device and SIM registries

  • EIR (Equipment Identity Register): tracks device IMEIs (International Mobile Equipment Identity), used to blacklist stolen phones.
  • HLR/HSS (Home Location Register / Home Subscriber Server): the authoritative record of which SIM is active, its permissions, and current location area.
  • eSIM provisioning platforms: increasingly the record of truth as physical SIM issuance declines.

Mismatches between the SIM registry and the BSS subscriber master are a classic audit finding: a SIM can be technically active on the network while billing systems mark the account as churned, inflating "active subscriber" counts if not reconciled.

Customer care transcripts and unstructured data

Call center transcripts, chat logs, and social media mentions are unstructured but increasingly mined for churn prediction, sentiment analysis, and Net Promoter ScoreNet Promoter ScoreNet Promoter Score (NPS) measures customer loyalty by asking how likely customers are to recommend a brand, then subtracting detractors from promoters.Voir la définition complète → (NPSNPSNet Promoter Score (NPS) measures customer loyalty by asking how likely customers are to recommend a brand, then subtracting detractors from promoters.Voir la définition complète →) diagnostics. This is the fastest-growing data category in telecom because of generative AI use cases, but it carries the highest privacy sensitivity and typically the least mature governance.

Where duplication and gaps actually happen

| Failure pattern | Typical cause | Business impact |

|---|---|---|

| Duplicate subscriber records | M&A system migrations, prepaid/postpaid silos | Overstated subscriber KPIs |

| Orphaned SIMs | SIM active in HLR, absent from billing | Revenue leakage |

| Roaming file mismatches | Partner format drift, timezone errors | Billing disputes, fraud blind spots |

| PM counter gaps | Vendor outage, clock drift | Understated network problems |

| Unlinked care transcripts | No common customer ID across channels | Missed churn signals |

A simple worked example: spotting subscriber duplication

Suppose a carrier's BSS reports 22.4 million active subscribers, but the HLR/HSS shows 21.1 million unique active SIM identities (IMSIs) over the same billing cycle (illustrative figures for this exercise, not a real carrier's reported numbers).

Duplication estimate = (BSS active count − HLR unique IMSI count) / BSS active count
                      = (22.4M − 21.1M) / 22.4M
                      = 1.3M / 22.4M
                      ≈ 5.8%

A gap this size, roughly 5.8%, is a governance red flag worth investigating before the number reaches an investor or regulatory report. It usually traces back to churned accounts not deprovisioned in the network, or duplicate accounts from a system migration.

Vérification des acquis

1. Why does the lesson argue that CDRs alone are insufficient for a full telecom analytics picture?

2. A single voice call can generate over a dozen separate records across different systems. What is the main analytical implication of this?

3. What is the primary role of OSS (Operations Support Systems) data in the telecom data landscape?

CHOIX MULTIPLES

4. Select ALL correct answers about Performance Management (PM) counters in OSS.

Sélectionnez toutes les réponses correctes.

CHOIX MULTIPLES

5. Select ALL correct answers about records generated beyond the CDR for a single call.

Sélectionnez toutes les réponses correctes.

Governance: who is accountable for what

European carriers operate under GDPR (General Data Protection Regulation), which mandates data minimization and a lawful basis for processing customer and location data; the relevant enforcement bodies are national Data Protection Authorities (e.g., CNIL in France, ICO in the UK, though the UK runs a parallel UK GDPR post-Brexit). In the US, telecom customer data (CPNI, Customer Proprietary Network Information) is governed by FCC (Federal Communications Commission) rules under the Communications Act, with state-level privacy laws (like California's CCPA/CPRA) adding further obligations.

Good practice, seen at leading operators, assigns a data owner per domain (network, billing, customer care) plus a cross-functional data governancedata governanceData governance is the set of policies, roles, and processes that ensure data is accurate, secure, well-defined, and used responsibly across an organization.Voir la définition complète → council that owns the master customer ID linking all systems together. Without that master ID, every cross-domain analysis (like the subscriber duplication check above) requires a manual, ad hoc reconciliation.

🎬 [VIDEO: "What is a CDR (Call Detail Record)?" - youtube.com - search for TechTarget or telecom-vendor explainer channels covering CDR structure and use in billing and fraud detection]

Key Takeaways

  • Telecom data lives in at least five distinct estates: OSS (network operations), BSS (commercial/billing), raw network element logs, roaming exchanges (TAP/NRTRDE), and device/SIM registries, plus fast-growing unstructured customer care data.
  • Each dataset feeds different decisions: OSS drives capexcapexCapital Expenditure (CapEx) is money spent to acquire, upgrade, or extend long-lived assets like equipment, property, or software that deliver value over multiple years.Voir la définition complète → and network health, BSS drives churn and ARPU, roaming data drives fraud and interconnect billing, and care transcripts increasingly drive retention models.
  • The most common hidden data-quality failure is subscriber or SIM duplication across systems inherited from M&A or channel silos; reconcile BSS active counts against HLR/HSS unique IMSI counts regularly.
  • Roaming data (TAP, NRTRDE, both GSMA standards) is uniquely fragile because it depends on bilateral file exchange with external partners, not internal systems.
  • Governance requires a named data owner per domain and a shared master customer ID; without it, cross-domain KPIs cannot be trusted, and GDPR or CPNI compliance becomes harder to demonstrate.

Suivant

Data quality dimensions for subscriber and network records