Anomaly Detection

Introduction

Distributed systems are hard. Infrastructure SLAs, sluggish database queries, and microservices latencies must be stitched together to form an accurate view for any eCommerce business. Add agile CI/CD with a Kubernetes-esque control plane, and the fragmented observability toolchain becomes a tax. How does one answer a simple question: are the customer experience (CX) and business KPIs healthy?  Well, that’s what bi(OS) empowers eCommerce teams to do.

Unified Observability across Storage, Queries, and Pixels

For over two years, Isima has used bi(OS) as its observability tool for internal operations across the big 3 public clouds. bi(OS)’s SDKs record query latencies on the application side within bi(OS) for analytics by Isima’s DevOps teams. Our customers use bi(OS) across multiple SaaS platforms to obtain real-time observability as their consumers interact with their business websites. This observability capability was achieved within days. 

This unified capability has allowed bi(OS) to support the entire data lifecycle, from pixels to alerts. We gained invaluable insights across the big 3 clouds while our customers have used the same capabilities across their business KPIs.

From Pixel to Slack

One of Isima’s customers has invested significantly in the operational reliability of its tech systems over the years. However, like any micro-service architecture, the sprawl is ever-expanding. It mandated an objective means of quickly detecting and alerting on the degradation of customer experience (CX) impacting business KPIs, e.g., click-throughs, product availability, and order conversions, so site reliability engineers (SRE)/Operations teams can initiate corrective actions.

An aspiring analyst used bi(OS) to build an NRT (near-real-time) anomaly detection framework using bi(OS). This module scanned the last 15 minutes of product views, clicks, inventory availability, and conversion events and compared it against same-time day-over-day and week-over-week metrics over the past 60 days. Upon detecting anomalous patterns, it fired a Slack alert that their product and leadership teams monitored. They are expanding the KPIs tracked in NRT and capabilities that would also help pinpoint issues across various tech systems faster. This dual approach allowed them to harden their DevOps capabilities as they grew with multiple acquisitions.

Conclusion

Unified observability is business critical for all eCommerce teams. Fragmented observability systems hinder this, and hence CX suffers. bi(OS) provides a unified single-pane-of-glass across systems, clouds, and business metrics. Data analysts can stitch together insights to build data apps such as forecasting, preemptive alerting, and auto-scale-out/in capabilities within hours.

1 Hour to BI, 1 Day to AI, and 1 Week to API. To get started for free, click here.