Anodot, the business monitoring company, today announced that the United States Patent and Trademark Office has granted the company a patent for a new algorithm that vastly improves the speed and reduces the computational cost of its anomaly detection offering. The algorithm boosts the industry-leading (seasonal) ARIMA model for estimating anomalies in time series data, reducing runtime by 99.98% and slashing the expense of running the model by up to 80%, enabling Anodot to better protect its customers’ bottom lines.
Anodot utilizes a number of data analysis models to effectively monitor the vast amounts of data businesses generate today in order to safeguard customers’ revenue, manage costs, and improve CX. While the ARIMA model is the most preferable due to its high accuracy, it is extremely slow, taking as long as 15 minutes for a single time series, rendering it too expensive to run at large scale. Enterprises implementing the model would be spending at least $125K per week– up to $6M per year. With Anodot’s innovation, the model’s run time is cut to 80 milliseconds – a reduction of nearly 5 orders of magnitude in both time and computational cost, making it realizable for anomaly detection at scale.
The ability to accurately detect anomalies swiftly and economically is critical at a time when companies are feeling the pains of a strained economy and are looking to cut costs and reduce waste – whether across enterprise operations or in the cloud.
“If data scientists tried to run standard anomaly detection on a 10 million time series, they’d be waiting 285 compute years for results,” said Dr. Ira Cohen, Anodot co-founder and Chief Data Scientist. “All businesses today are powered by data, and it’s our job to uncover any anomalies that point to holes that need plugging or issues that need fixing. We knew we had to find a better way to provide anomaly results to the data analysis teams we serve to help them get costs under control and protect their bottom line. This innovation will help them do just that, making the industry’s best model feasible for anomaly detection across the enterprise.”
The new detection tool fits organically into Anodot’s suite of BI tools, including the company’s AI and ML-powered business intelligence monitoring platform, which helps organizations identify revenue-critical incidents.
“This new patent is a milestone for Anodot customers and their frequently overburdened data teams,” said David Drai, Anodot CEO and co-founder. “As businesses of every size generate more and more data, it becomes increasingly difficult to clearly see and understand the impact of every small part making the machine run. We’re eager to see this lightning-fast solution in action.”
Tune in to Martech Cube Podcast for visionary Martech Trends, Martech News, and quick updates by business experts and leaders!