Which Test Cases Should You Automate First? A Practical Guide

The discussion originated from a chat about maximizing the output of test automation, questioning whether coverage, efficiency, or bug‑rate metrics alone are sufficient. Before automating, teams must weigh the time, effort, and resources required against the tangible benefits, recognizing that not every manual test should be automated simply because it is possible.

Tests That Should Be Automated

Business‑critical paths where a failure would damage the business.

Tests that must run for every internal version or release, such as smoke, sanity, and regression tests.

Tests that need to be executed across multiple configurations (different OS and browser combinations).

Tests that repeat the same workflow with varying input data each run (data‑driven testing).

Tests involving large amounts of input data, for example filling lengthy forms.

Tests suitable for performance evaluation, such as stress and load testing.

Tests that take a long time to execute and may need to run overnight or during breaks.

Tests that must capture screenshots to verify application behavior or to compare visual rendering across browsers.

In general, the more repetitive a test is, the better it fits automation.

Automation is not the only way to add value; creating test data for manual exploratory testing can also demonstrate the benefits of automation.

Tests That Should Not Be Automated

Tests that will only run once (except when a very large data set makes a one‑time run worthwhile).

Usability or user‑experience tests that require direct user feedback.

Tests that need rapid execution; new features often require quick manual feedback.

Exploratory tests that rely on domain knowledge or random, ad‑hoc testing.

Intermittent tests with unpredictable results; automation requires predictable, reliable outcomes.

Tests that require visual confirmation; while screenshots can be captured, the final review may remain manual.

Tests that cannot be fully automated should not be automated unless doing so saves a substantial amount of time.

Personal Viewpoint

Simple > Priority > Stability > Repetition.