Prelert launched a new Retail Order Analytics solution, which helps online and multichannel retailers identify technical and operational issues as they’re happening in order to stem losses and protect revenue streams.

The technology is already being used by a number of retailers to improve digital commerce efficiency, including one of the world’s largest multichannel brands and one of the world’s largest pure-play ecommerce sites.

Prelert’s Retail Order Analytics solution can be broadly applied to analyze real-time transaction metrics such as orders per minute, carts created per minute, invoices per hour, or deposits per hour, so that revenue-impacting events can be found and fixed quickly. For example, after automatically learning what normal behavior looks like within any given metric, it can identify issues such as an unusually high number of abandoned carts, an unusually low number of completed checkouts, or even an invoice brown-out, so the root cause can be identified and addressed in near real time.

Accurately modeling periodicity – also known as seasonality – is a difficult data problem for retailers to solve. As a result, automated data analysis is becoming necessary for retailers to identify critical problems and avoid drowning in false positive alerts. Due to the varying nature of periodic data, writing rules that can accurately monitor constantly changing behaviors is nearly impossible. Employing humans to watch dashboards and graphs is expensive and subject to human error. Even using supervised or trained machine learning is a poor solution because it can generate a stream of false alerts as data patterns change.

Built with unsupervised machine learning technology, Prelert’s solution automates data analysis and automatically detects the periodicity of daily and weekly order cycles. It adapts to changing data patterns that may result over time due to factors such as a new product becoming available or current events that cause a spike in product interest, and constantly updates its self-generated models of normal baselines. As a result, it accurately finds deviations in expected behavior that can indicate costly problems.

“A significant drop in the number of orders taken by an e-commerce site during a particular day might be obvious in retrospect, but can be very difficult to catch in near real time without automated machine learning. Static thresholds and even moving averages can’t reliably identify issues,” said Mark Jaffe, CEO of Prelert. “Our anomaly detection algorithms have been proven to work and provide significant ROI within hundreds of progressive IT organizations around the globe. We can provide the same value now for retail and ecommerce organizations, with a solution tailored specifically for them.”

Prelert is easy to deploy, bringing analytics to where an organization’s data already resides to analyze it in near real time. In addition, an open API allows developers to use Prelert in their own products or environments.