The problem with running a “marketing test” is many tests are flawed.

Either they’re setup wrong, they’re using flawed assumptions (sampling bias), or they’ll never reach statistical significance.