From random alerts → systematic monitoring + action
Overview
Custom + Auto Monitors reimagined anomaly detection in Klaviyo. By introducing user-defined monitors, always-on defaults, and a permanent Alert Log, the system reduced alert times by ~90%, empowered self-service, and lowered support burden.
Context
Some of Klaviyo's biggest clients reached out directly to request better alerts for their metrics. They needed flexibility as well as the ability to diagnose past alerts to stay on top of future problems they may be able to catch early.
Problem statement
Design an anomaly report that showcases various issues for a metric.
Issue 1
Alerts were generic and customers had no control over what was being monitored.
![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 24 34" xmlns="http://www.w3.org/2000/svg"><path d="M 19.729 19.735 L 9.864 29.605 L 0 19.735 M 9.864 0 L 9.864 28.615" fill="transparent" height="29.60500061035156px" id="gFf8LlmTP" stroke-dasharray="" stroke-linecap="round" stroke-linejoin="round" stroke-width="3.947" stroke="rgb(207, 190, 169)" transform="translate(2 2)" width="19.729000000000003px"/></svg>)
Custom Monitors let users choose metrics, thresholds, and conditions — making alerts relevant, actionable, and tailored to their business.
Issue 2
Once dismissed, alerts disappeared with no history → no historical context.
The new Alert Log provided a permanent record of anomalies, so customers could review and act anytime.
Issue 3
Identifying issues was slow and inconsistent.
With Custom + Auto Monitors, anomalies were flagged up to 90% faster, enabling self-service resolution and reducing support tickets.
User Needs
I need flexibility to track the metrics that matter most.
I want thresholds and alerts tailored to my business.
I need a history of triggered anomalies,
not just one-offs.
Process Steps
Step 1
Early Exploration
Started off with anomaly reports per metric that implemented a Root-Cause analysis style of breaking down the anomaly.
Step 2
Pivot to Custom Monitors
Workflow: Create → View → Edit/Delete → Alert Log.
Step 3
Historical Data Viz for Custom Monitors
Trendlines helped set thresholds based on past performance so the user knows what thresholds will work for their business.
Step 4
Table-first Design
Unified Auto + Custom monitors in one page with a view that works for both power + casual users.
Step 5
Alert Log
Permanent anomaly history helped build trust and reliability.
Final Prototype
Impact
90% faster anomaly alert times
Increased self-service option that led to reduced support tickets + escalations
Scaled for both power users and everyday accounts leading to greater adoption by enterprise + mid-market
Recognition
“These real-time alerts allow for swift identification and resolution of issues, helping you minimize any potential impact on revenue, engagement, and deliverability. They act as email automation monitors Klaviyo users can rely on for timely insights.”
Patent filed for customizable anomaly detection system with proprietary monitoring methods powered by the Monitors UX inside Klaviyo.
Learnings
Challenged the brief → from anomaly reports to critical & lifesaving workflows
Drove key partnership between between Product, Engineering, Design and Data Analytics
Key contributions to the design systems via the work on tables
Required deeper understanding of managing Engineering partnerships to ensure the backend is represented faithfully on the frontend