Actionable Alerts: The 'What Now?' Test
You'll learn: How to design alerts that immediately tell responders what action to take, not just what went wrong.
The Problem with Vague Alerts
Many alerts fail because they announce a problem without providing direction. Imagine a car dashboard light that simply says "Problem Detected" instead of "Low Oil Pressure — Pull Over Safely." The first creates panic and confusion; the second prompts immediate, correct action.
An alert that passes the "What Now?" test answers the question: "If I receive this alert at 3 AM, do I know exactly what to do next?"
What Makes an Alert Actionable
Every alert should include or link to:
- The symptom (what's breaking from a user perspective)
- The likely cause (based on the metric that triggered it)
- The immediate action (what to do right now)
- Where to look (relevant dashboards, logs, or runbooks)
Example: Before and After
Bad alert:
ALERT: High error rate on payment-service
Good alert:
ALERT: Payment processing failing (15% error rate)
ACTION: Check payment-service pods in us-east-1
RUNBOOK: https://wiki/runbooks/payment-failures
DASHBOARD: https://monitoring/payment-service
The second alert passes the "What Now?" test because the on-call engineer immediately knows the impact, where to investigate, and where to find detailed procedures.
Connecting to Previous Concepts
This builds on Alert Fundamentals: Symptoms vs. Causes—you're alerting on symptoms, but now you're also providing the investigative path to causes and solutions.
Key Takeaway: An actionable alert tells the responder not just what is wrong, but what to do about it and where to start looking—making 3 AM pages less painful and faster to resolve.