MTN EVA 3.0 on Azure Databricks: Africa Telco Cloud Modernisation

MTN’s migration of its Enterprise Value Analytics platform to Microsoft Azure — rolled out as EVA 3.0 — marks one of the most ambitious telco cloud modernisations in Africa, promising faster analytics, tighter security, and a reusable blueprint for data-led operations across the MTN Group.

Background / Overview​

MTN and Microsoft have been publicly aligned on a wide-ranging cloud partnership for several years, positioning Azure as a primary cloud target for both core network proofs of concept and enterprise analytics modernisation. The recent announcement of EVA 3.0 describes a cloud‑native rebuild of MTN’s Enterprise Value Analytics stack on Azure Databricks, secured by Microsoft Defender, and framed as the largest telco cloud implementation across the Middle East and Africa.
Public statements and industry reports attribute several headline-scale metrics to the new platform: processing in the order of tens of billions of records daily, running hundreds of analytics workflows, and ingesting more than a thousand distinct data feeds. These claims have been repeated across multiple trade outlets and company briefings; where the exact figures are cited, they originate from MTN’s announcement and partner summaries. Those numbers strongly indicate a lakehouse-scale deployment designed for continuous ingestion, real‑time analytics, and AI-enabled operations — but they remain company‑provided operational metrics that are not independently audited in the public domain.

---

What MTN says EVA 3.0 delivers​

• Near real‑time visibility into service performance and customer behaviour.
• Faster processing of large volumes of telco telemetry to reduce detection-to-remediation times.
• A consolidated analytics backbone that supports personalised offers, predictive maintenance, and operational automation.
• A hardened security posture using Azure-native tooling to support compliance and responsible AI deployment.
• A reference architecture that can be adapted and rolled out across MTN Group markets, reducing duplicate engineering effort.

Multiple industry writeups and MTN‑led communications consistently describe the platform as built on Azure Databricks with security integration via Microsoft Defender, and positioned as the central analytics backbone for MTN South Africa’s operations. The company also frames EVA 3.0 as a repeatable blueprint for other operating companies within the MTN Group.

Architecture and technology choices: what sits under EVA 3.0​

Core building blocks​

The public description of EVA 3.0 maps to a modern data lakehouse architecture composed of the following elements:

• Data ingestion layer: Hundreds to thousands of source feeds (network telemetry, OSS/BSS events, customer interactions, device telemetry) ingesting both streaming and batch records.
• Storage: Azure Data Lake Storage (ADLS Gen2) holding Delta Lake‑formatted tables to enable ACID semantics and efficient incremental processing.
• Compute: Azure Databricks providing scaled Apache Spark clusters for both streaming and batch ETL/ELT and for model training and scoring.
• Orchestration: Databricks Jobs, possibly integrated with Azure Data Factory or other workflow managers for dependency handling and scheduling.
• Security and governance: Azure Active Directory for IAM, Microsoft Defender for threat protection, and cataloging/lineage via Unity Catalog or equivalent metadata services.
• Consumption and productisation: Power BI, APIs, and downstream systems for operational dashboards, customer experience tooling, and automated remediation pipelines.
• Responsible AI controls: Model registries, versioning, bias checks, and monitoring to ensure models deployed against customer data comply with governance policies.

This stack aligns with commonly adopted enterprise patterns for lakehouse deployments on Azure: Databricks for scale and Spark processing, Delta Lake for durability and schema evolution, and Azure security services for a centralised control plane.

Why those choices make sense for telcos​

Telco workloads are characterised by extremely high event velocities, strict SLAs, and sensitive customer metadata. Azure Databricks is a logical choice because it provides:

• Elastic Spark compute for bursty workloads (e.g., congestion events, billing runs).
• Delta Lake semantics for reliable incremental updates and time-travel queries.
• Native integrations with Azure identity and storage services to reduce integration friction.
• An ecosystem that supports ML lifecycle management and model deployment at scale.

Microsoft Defender and Azure's broader security stack allow for consistent threat detection and compliance controls across platform layers. For MTN, the combination of Databricks and Azure services streamlines governance, observability, and the operational tooling that telcos need to maintain uptime and regulatory compliance.

---

Verified claims and numbers — what is corroborated​

• MTN and Microsoft’s partnership on Azure and joint cloud initiatives is a documented, multi-year strategic relationship that includes skilling programmes, proofs of concept for cloud‑native network functions, and enterprise migrations.
• Public materials from MTN and partner outlets confirm that MTN has invested heavily in Azure skills and established a Cloud Centre of Excellence to drive consistent adoption.
• Multiple industry sources repeat the high-level technical stack for EVA 3.0 — Azure Databricks and Microsoft security tooling — and describe the project as intended to deliver real‑time analytics and faster operational insight.

Where the public record is consistent across sources, the architecture and business rationale for the migration are well supported by both vendor documentation and industry reporting.

Claims that should be treated with caution​

• The headline metric attributed to EVA 3.0 — figures like processing ~22 billion records per day, running 800+ analytics workflows, and ingesting over 1,700 data feeds — are presented in MTN’s announcement and reported across media outlets. These are plausible for a telco of MTN’s scale, but such operational tallies are company‑reported and have not been independently audited in the public domain. They should therefore be considered credible company metrics rather than independently verified facts.
• The count of internal Microsoft Azure certifications reported for MTN employees is cited at varying levels in public materials (figures in the low‑to‑mid thousands across different announcements). While these numbers demonstrate a substantial skilling effort, exact tallies may differ between releases and across dates; treat them as indicative of a major training push rather than an immutable statistic.

---

Strengths: why EVA 3.0 is a meaningful step for MTN and African telcos​

• Operational velocity and customer experience
• Shifting analytics into a cloud-native lakehouse enables quicker detection of network incidents and behavioural trends. This lowers mean time to detect/repair and creates the data foundation for personalised customer engagement.
• Scale and elasticity
• The ability to autoscale compute for peak telemetry ingestion (e.g., mass events, network incidents) reduces the need to overprovision on-premises infrastructure and allows a pay-for-what-you-use model.
• Centralised governance and security
• Using Azure native services for identity, access, and threat detection simplifies compliance and enables consistent policy application across datasets and models.
• Repeatable blueprint across markets
• A standardized architecture fosters reusability across MTN’s markets, accelerating rollouts and lowering integration costs for subsequent deployments.
• Skills uplift
• A substantial certification programme and a Cloud Centre of Excellence increase internal capability to run and evolve the platform without over‑reliance on external contractors.
• Enabler for AI monetisation
• With governed data and scalable compute, MTN can develop new AI‑driven products (predictive maintenance, personalised offers, enterprise analytics) and potentially monetise data in compliant ways.

---

Risks, trade‑offs, and things to watch​

1. Vendor lock‑in and portability​

Deep dependence on Azure-managed services (Databricks as a managed layer, Unity Catalog, Defender) increases switching cost. Over time, tight coupling to platform‑specific APIs and tooling can make multi‑cloud migration expensive and operationally disruptive.
Recommendation: adopt open formats (Delta Lake), enforce clear API boundaries, and maintain a documented export strategy for critical datasets.

2. Data sovereignty and regulatory exposure​

Telco telemetry often contains personally identifiable metadata subject to sectoral and national regulations. Centralising analytics in a public cloud requires contractual clarity about data residency, auditing, and law‑enforcement access.
Recommendation: define region-aware architectures, use private peering options (ExpressRoute/Private Link), and ensure contractual SLAs address data residency and government access scenarios.

3. Exploding operational complexity at scale​

Processing billions of records daily with hundreds of workflows demands mature SRE practices: observability, alerting, back‑pressure controls, and chaos testing. Without these capabilities, pipelines fail in ways that are hard to diagnose.
Recommendation: invest in observability (metrics, tracing, logs), automated testing for data pipelines, and runbooks for common failure modes.

4. Cost governance​

Cloud compute and storage can scale faster than budgets. If not managed, costs can outstrip the business value delivered.
Recommendation: implement tagging, budget alerts, resource quotas, cost-aware autoscaling, and finite retention policies for high-volume intermediate data.

5. Security surface expansion​

A larger cloud footprint increases attack vectors. Defender and other tools help, but they must be tuned and staffed.
Recommendation: maintain an active threat‑hunting program, continuous compliance checks, and regular red-team exercises targeted at data plane and identity flows.

6. People risk​

Certifications are useful but are a proxy for real operational experience. Staff churn or insufficient on‑the‑job training can erode gains.
Recommendation: combine certifications with rotational on-call, SRE apprenticeships, and platform engineering communities of practice.

---

Practical steps and a checklist for telcos planning similar migrations​

• Define the business outcomes first (MTTR reduction, churn reduction, new AI revenue) and map them to measurable KPIs.
• Establish a Cloud Centre of Excellence to set standards for architecture, security, and governance.
• Build an ingestion strategy that handles both streaming and batch sources, and validates schemas early.
• Adopt open data formats (e.g., Delta Lake) and ensure metadata and lineage visibility for compliance.
• Implement a robust orchestration layer with retry semantics, dependency graphs, and back‑pressure handling.
• Bake in cost governance from day one: tagging, budgets, and automation for idle resource reclamation.
• Formalise responsible AI controls: model approval gates, bias tests, and production monitoring.
• Run proof‑of-value pilots with measurable SLAs before a full‑scale migration to validate cost, latency, and operational practices.

---

Governance and responsible AI: more than PR​

MTN explicitly framed EVA 3.0 as supporting responsible AI. That claim needs to be tied to concrete controls:

• Clear model governance (registries, lineage, versioning).
• Data minimisation and anonymisation for model training.
• Bias detection and fairness checks embedded into model evaluation.
• Production monitoring for drift, performance, and unintended consequences.
• Human-in-the-loop controls for high-impact customer actions.

Responsible AI is a continuous process, not a one-time checklist. The platform must provide instrumentation for operationalising AI governance: feature stores with access controls, model explainability tooling, and incident workflows that assign accountability when models misbehave.

---

Business impact: what MTN and its customers stand to gain​

• Faster incident detection reduces downtime and improves customer satisfaction.
• More relevant offers and personalised experiences can increase average revenue per user (ARPU) and reduce churn.
• Operational efficiency from automation lowers OPEX and allows reallocation of staff to higher-value activities.
• New services for enterprise customers (analytics-as-a-service, network insights) create non‑connectivity revenue streams.

These gains are realistic but require disciplined measurement and a feedback loop linking platform teams to product and commercial teams.

---

Open questions and where transparency would help​

• Independent verification of the headline operational metrics (daily record counts, workflow counts, number of feeds) would strengthen external confidence in EVA 3.0’s scale claims.
• Greater detail on data residency and cross-border data flow controls would clarify regulatory risk posture in jurisdictions with strict teleco data laws.
• Cost transparency or typical run‑rate scenarios for a deployment at this scale would be valuable to other operators contemplating a similar migration.
• Details on the exact governance tooling used for AI lifecycle management (model catalog, explainability tools, and CI/CD pipelines) would help teams evaluate replicability.

Where possible, customers and partners should request proof-of-value demonstrations and transparent KPIs (latency, cost per TB processed, detection-to-remediation times) before committing to large rollouts.

---

Verdict: pragmatic optimism with guarded oversight​

EVA 3.0 represents a credible, high‑impact step for MTN toward a cloud‑native, data-driven future. The architectural choices — Azure Databricks, Delta Lake semantics, and Azure security integration — are consistent with industry best practices for large analytical workloads and AI productisation. MTN’s emphasis on skilling and establishing a Cloud Centre of Excellence addresses one of the most persistent failure modes in platform modernisation: people and process.
At the same time, the move highlights the usual cloud transition trade‑offs: vendor dependency risk, regulatory complexity, potential for runaway costs, and the need for mature operational practices. The headline operational metrics are impressive, but should be understood as company‑reported figures that would benefit from independent corroboration for third‑party analysis and procurement decisions.
For telcos and enterprise CIOs, the lessons are clear: adopt repeatable architectures, invest in skills and governance early, and insist on measurable proof-of-value. For vendors and hyperscalers, the test is to provide transparent migration paths, open data export options, and demonstrable commitments to compliance controls in diverse regulatory environments.

---

Practical recommendations for IT leaders considering a similar initiative​

• Start with an MVP that proves value on a single, high-impact use case (e.g., proactive fault detection) and iterate.
• Require vendors to demonstrate data exportability and an exit plan as part of procurement.
• Build strong cost observability from day one and treat cloud spend as a first-class operational metric.
• Institutionalise SRE practices for data pipelines — on-call rotations, playbooks, and automated remediation.
• Enforce a staged approach to AI deployment: pilot → controlled rollout → wide release, with governance gates at each stage.
• Keep legal, compliance, and network engineering teams in the core decision group to address sovereignty, egress, and latency constraints.
• Use the Cloud Centre of Excellence to codify best practices and accelerate adoption across countries while allowing local customisation.

---

EVA 3.0 is a significant milestone in African telco digitalisation: a practical demonstration of how hyperscaler capabilities, when combined with an operator’s domain expertise and a focused skilling programme, can create a modern analytics backbone. The long-term value will depend less on the initial technology choice and more on disciplined governance, cost control, and the sustained ability to operationalise the platform into reliable, customer‑facing services.

---

Source: TechAfrica News MTN Completes Major Cloud Modernisation with Microsoft, Unveiling EVA 3.0 in South Africa - TechAfrica News