That’s correct — you cap it at the maximum value that each bracket allows. In the case you mention it is indeed 3. Another example — if you had three pieces of anecdotal evidence, 0.5 each, and one piece of market evidence(1), you have a confidence level of 0.5 x 3 + 1 = 2.5. If you found another market evidence item you’ll have theoretically 3.5, but the max is 3 so your confidence level is 3. This is not exact science of course, my goal is just to prevent people feeling too confident about having many low-confidence tests (many people think this is a good idea, we sent multiple surveys… ).