This is a very important point though personally I formulate it the other way around:<p><pre><code> you only (proactively) fix what you measure, and you only measure what you think is important.
</code></pre>
This is of course sensible (you know weaknesses/risks/scale points in your design and so you monitor them).<p>But end users do unexpected things and it's worth looking, since at the end of the day it's all about the end users, else you don't have revenue! Sometimes, as the article notes, there's nothing you can do. But trust end users to find your blind spots for you.
> email Jeff Bezos with a streaming video problem. He'd add a single '?' in the body<p>"A good driver should know what's wrong."<p>P.S.
not accurately quoted. Better: <a href="http://mtchndrn.blogspot.pt/2007/12/when-i-first-started-using-unix-when-i.html" rel="nofollow">http://mtchndrn.blogspot.pt/2007/12/when-i-first-started-usi...</a>