About

Fourier is a standalone desktop application, built with data security and compliance in mind, for identifying and reviewing outliers and anomalous behavior in numerical, real-valued time-series data using advanced statistical methods and explainable AI. It runs entirely on the user's local machine and requires no internet connection to use; just a valid license file.

Fourier was designed to help automate and solve a persistent problem that many institutions and businesses who manage large volumes of time series data face: Identifying anomalies in newly reported values as data updates over time. It acts as a virtual analyst or analyst assistant  that efficiently scans all of your data to surface outlier behavior for human review,.

Fourier is a truly plug-and-play solution with no manual tuning required, eliminating learning curves and enabling effortless use even for non-technical users . With clean, intuitive interfaces and fast local processing, analysts can run anomaly checks across hundreds of thousands of time series in seconds and quickly prioritize their reviews using Fourier’s ranked anomaly scores. 

Flagged results can be exported for downstream use or examined immediately within Fourier’s integrated Review Dashboard, where analysts can filter results, rapidly generate plots, define custom columns for deeper investigation, and more. This streamlined workflow saves time, improves data quality, and eliminates the need to juggle spreadsheets with external BI tools, allowing analysts to move seamlessly from detection to decision within a single platform.  Users can also manually load their own data (via CSV or Excel) to the integrated dashboard, without needing to first run it through Fourier’s analysis.

Fourier is an all-in-one anomaly detection and review platform that streamlines the entire workflow—from detection to validation—allowing your analysts to focus on interpreting results and making informed decisions.

At its core, Fourier uses efficient yet powerful mathematical techniques to model trends in your data and automatically detect if the latest entries deviate from their historical norms. 

Unlike traditional rules-based methods which are often ineffective when the underlying data is highly variable or complex, this dynamic approach allows Fourier to work on a huge variety of datasets without the need for any tuning. The only technical requirements are that the data must be real-valued,  1-dimensional, ordered left to right from oldest to newest, and have at least 5 data points. Data must also be provided in the form of rows in an Excel or CSV file.

Fourier also possesses advanced and  toggleable feature detection and mitigation systems, which are explained in more detail below.