Summary  
This chapter covers the concept of guardrail metrics: auxiliary measurements monitored alongside primary metrics to catch unintended negative effects and ensure system stability.

General domain of usage  
A/B testing and experimentation in software products.

When running A/B tests, you track **primary** and **secondary metrics** to measure the impact of your changes. However, relying only on these outcome metrics can put your experiment at risk. This is where **guardrail metrics** become essential.

## What Are Guardrail Metrics?

Guardrail metrics are extra measurements you monitor during experiments to catch and prevent unintended negative effects. Think of them as safety checks: they make sure that, while you aim to improve key metrics, you do not accidentally damage other critical aspects of your product or user experience.

- **Guardrail metrics are not your main success criteria;**
- They are crucial for protecting users, business goals, and the integrity of your experiment.

**Example:**
You might test a new feature that increases user engagement. But if it also causes a spike in errors or slows down your website, the experiment could be too risky to launch—even if your primary metric looks positive. Guardrail metrics help you spot these issues early, so you can make smart decisions and prevent costly mistakes.


What is a likely consequence of ignoring guardrail metrics in an A/B experiment?

Master the principles and practical application of hypothesis testing and A/B testing. Learn statistical tests, experiment design, data preparation, and analysis techniques using Python to make data-driven decisions in real-world scenarios.

Explore the core concepts of hypothesis testing, including t-tests, hypothesis formulation, and statistical assumptions, with practical Python examples.

Learn about z-tests for means and proportions, chi-square tests, and how to select the appropriate statistical test for different data types.

Build a strong foundation in A/B testing by understanding experiment structure, metric selection, randomization, and sample size considerations.

Learn how to structure, clean, and validate experimental data, including outlier detection and metric construction for robust analysis.

Apply statistical tests to A/B test data, interpret results, and understand key concepts like confidence intervals, uplift, and p-values.

Understand and implement guardrail metrics, monitor variance, and apply stopping rules to ensure experiment integrity and reliability.