Anomaly detection
Every KPI watched against its forecast, with seasonality and planned works discounted, and the worst cells first, clustered by common cause.
Results
Effort to explain network anomalies.
Anomalies grouped by common cause, worst cells first.
Video recap
At a glance
For
Network operations and performance teams
Replaces
Static KPI thresholds and manual alarm triage
Data used
PM counters, alarms, configuration changes and the planned works calendar
Runs on
Your cloud or on-prem, every major vendor, 2G to 5G
The problem
Static thresholds either fire too often or too late. Engineers spend their day on alerts that turn out to be planned works or normal seasonal changes.
Real anomalies hide among them and are often found when customers start to complain.
How it works
Follow up
The agent keeps each cluster under watch until KPIs return to forecast, and closes it when they do.
What changed in the network overnight that is not explained by planned works?
The anomalies of the night, grouped by common cause and ranked by impact, with the explanation and the cells affected.
Frequently asked questions
How is it different from threshold alarms?
Each KPI is compared with its own forecast, so normal daily and seasonal changes and planned works do not raise alerts.
Which vendors and technologies does it support?
All major vendors and technologies, from 2G to 5G. The data from every vendor in your network is brought into one model before the agents work on it.
Does the platform change anything on the network by itself?
No. The agents detect, explain and recommend. Your engineers decide, and changes go through your own systems.
Where has it been used?
With Telefónica, where the effort to explain network anomalies dropped by 75%.
“O2 climbed one position in independent rankings after deploying our anomaly detection and root cause analysis engines.”
See what your thresholds are missing.
Bring one region. We run the agents on your data and show you the anomalies, grouped by cause.
