Noisy alerts
Replace static thresholds with anomaly detection.
AI-powered Analytics.
Anomaly detection, forecasting and analysis on your operational data, built into the tools you already work with.
The problem we solve.
Static thresholds do not know what time it is: they alert on every normal peak in activity and stay silent when unusual behaviour remains below the limit. And the historical data that could anticipate problems goes unused.
What’s included.
- Anomaly detection on metrics and KPIs
- Capacity and demand forecasting
- Models trained on your historical data
- Integration into alerts and dashboards
- Measurement of false alarms and correct detections
How we work.
Five phases, always in the same order. Select each one to see what happens in it. In full projects they map onto the stages of our method.
We identify the use case, the data available, the risks and how success will be measured.
We design the solution: models, data, integrations, security controls and traceability.
We train the models on your historical data and integrate them into alerts and dashboards.
We compare false alarms and detected anomalies against the current thresholds over a trial period.
We tune the system based on real usage and monitor the system like any other production service.
Technical capabilities.
- Splunk Machine Learning Toolkit
- ITSI adaptive thresholds
- Splunk App for Data Science and Deep Learning
- Time series forecasting
- Outlier detection
- Natural language queries
A fixed threshold doesn't know what time it is
The same metric over two days, watched two different ways. Move the fixed threshold and the model's sensitivity, and compare how many alerts each one fires and how many of the two real anomalies it catches.
Synthetic example series. The orange dots above are fixed-threshold alerts; the pink ones below are from the adaptive model.
Fixed threshold
- Alerts
- 0
- False alarms
- 0
- Real anomalies detected
- 0 of 2
Adaptive
- Alerts
- 0
- False alarms
- 0
- Real anomalies detected
- 0 of 2
In Splunk we solve this with adaptive thresholds in ITSI or with Machine Learning Toolkit models trained on your own history.
Use cases.
Capacity
Forecast when a resource will run out.
Fraud or abuse
Detect behaviour that is out of the ordinary.
Benefits for your organisation.
- Fewer false alarms
- Anomalies detected sooner
- Capacity decisions made ahead of time
- Models that run on your platform
Deliverables.
- Trained and documented models
- Integrated alerts and dashboards
- Results report
- Retraining procedure
Frequently asked questions.
Do we need a data science team?
Not to get started. We use tools built into Splunk and document how to maintain the models.
Does our data leave our platform?
With Machine Learning Toolkit and ITSI adaptive thresholds, the models run inside your Splunk environment.
How much historical data is needed?
It depends on the seasonality of the metric; usually several weeks to capture the weekly cycle.
Other Artificial Intelligence services.
Shall we talk about AI-powered Analytics?
Tell us about your situation. If this service is not what you need, we will tell you; if it is, we will propose a concrete first step.
Request this service