Conversion improvements depend on a lot of factors (amount of visitors you get, number of alternatives under test, differences between the alternatives, and more).
If you are still curious, see some statistics about our customers and the results they got using Nelio A/B Testing.
First experiments on your web may get more than a 10% improvements in conversion. In some cases, you can even obtain a conversion raise of more than 200%.
A/B test results can still be quite unpredictable and best practices may not always work on your site. It is recommended that you test any changes before you make them permanent on your website. Change from an opinion-driven development perspective to a data-driven one.