Manual code reviews are one of the oldest and best-understood code quality assurance techniques. In our group, we have employed lightweight, continuous code reviews successfully for the last five years on all changes to the program analysis toolkit ConQAT. In our experience, it is the most important factor for its continued maintainability and success.

However, continuous code reviews are hardly used in practice. While some companies use them sporadically, few employ them on a large scale. This seems to be one root cause for many real-world software maintenance prob-lems. For instance, inconcise/inconsistent naming or missed reuse opportuni-ties, which hamper software maintenance, could be detected early by contin-uous code reviews.
But: is our assumption correct? How much benefit do we really derive from the code reviews we perform? An analysis of the data we gathered in 5 years with over 50 developers could deliver important insights, which can help to make a stronger case for (or against) continuous code reviews in practice. It can also uncover improvement potential for our own review process.

The goal of this thesis is to analyze the thousands of code reviews performed by the ConQAT (www.conqat.org) development team in our group, to quantify the costs and benefits of lightweight, continuous code reviews.