Customer Story
DKMS
Discover how DKMS uses Exasol to unify data across seven countries, standardize KPIs, and reduce international reporting from lengthy manual processes to seconds.
Fortune 50 telecom company transforms spectrum analytics
Network planning with real-time data at a national scale
0
of data
0
records
0
sites
0
active users
Case Study
Executive Summary
On a national scale, network analytics cannot depend on multi-day data pulls or pre-aggregated answers. This leading U.S. telecom needed to work directly with detailed spectrum, geospatial, drive-test, speed-test and network telemetry data (including 27 TB of drive-test and crowdsourced speed-test data annually), while supporting planning and forecasting across 100,000+ sites. By replacing Microsoft SQL Server with Exasol, the company created a single high-performance analytical layer for business intelligence, GIS, and in-database machine learning. More than 140 active users can now access detailed data directly; requests that took 2-3 days now complete in seconds; per-region modeling workloads that took approximately 1 week now finish in hours; and the environment has scaled from 1TB to 43TB without rearchitecture.
Location
United States
Company
Leading U.S. telecommunications provider
Industry
Telecommunications
Use Case
National-scale spectrum and network analytics
The Challenge
Teams across spectrum planning, GIS, network engineering, and finance relied on centrally prepared extracts, turning routine data requests into waits of 2-3 days. Simultaneously, they had to work with highly detailed and diverse data, from FCC auctions and call signs to 280 million geospatial records, 400 shape files, drive tests, speed tests, and network telemetry.
Executive leadership required every result to remain traceable to the underlying records rather than relying on layers of pre-aggregated averages. Therefore, any replacement platform had to deliver responsive analysis on raw fact tables, support concurrent self-service access, and handle geospatial and machine learning workloads without creating another infrastructure silo.
The Solution
Exasol became the shared analytical foundation for 5 mission-critical workloads.
- Spectrum planning and management
Track call signs through FCC auctions, manages spectrum portfolio strategies across frequency bands, and allocate spectrum geographically for 5G, home internet, and millimeter-wave products. - Forecast network capacity 10-12 weeks ahead
Run Prophet models with grid search across 100,000+ cell sites to predict tonnage, congestion, and the impact of home internet users on existing network capacity. - Competitive coverage mapping
Compare coverage with competing carriers using 280 million FCC hexagon records and 400 shape files using UDF-based geometry processing, enabling market-by-market regression testing without losing the national view. - Analyze 30 billion records per quarter
The company ingests 27 TB of drive-test and crowdsourced speed-test data annually and use rolling eight-week windows to analyze recent network performance. - Run AI/ML where the data lives
Execute time-series forecasting and regression models directly inside Exasol using Python UDFs and custom Script Language Containers, reducing per-region processing from one week to hours.
The Results
From 1-2 days to seconds
More than 140 active users across spectrum planning, GIS, network engineering, and finance can access detailed data directly instead of waiting for centralized data pulls.
From 1 week to hours
Per-region time-series forecasting and regression workloads that previously required about one week now complete in hours.
Lower compute cost
For equivalent workloads, the company found Exasol required less compute spend than the hyperscaler alternatives it evaluated.
30 billion records per quarter
Teams can analyze raw fact tables while preserving the record-level detail leadership requires for traceability and explainability.
From 1 TB to 43 TB
The environment scaled without a platform migration or architecture redesign.
Why Exasol
- Lower compute cost than evaluated alternatives
Deliver equivalent workloads and less compute spend than the hyperscaler alternatives evaluated. - Raw-data performance at massive scale
Analyze 30 billion-record fact tables without replacing record-level detail with pre-aggregated “averages of averages.” - In-database AI/ML
Run Python UDFs and custom Script Language Containers directly where the data lives, avoiding a separate machine-learning infrastructure layer. - Scale without disruption
Grow from 1 TB to 43 TB without migrating to another platform or redesigning the architecture. - Enterprise deployment on AWS
Support LDAP integration, service accounts, existing access controls, andc orporate security requirements.
Customer Stories
Discover inspiring success stories
What could your teams do with answers in seconds instead of days?
Bring us your most demanding spectrum, network-capacity, GIS or AI/ML workload. We will help you assess how Exasol could accelerate analysis on detailed data, simplify the analytics architecture, and lower compute cots.