Comment by zdragnar

I've seen customers be driven away by poorly performing interfaces. I've seen downtime caused by exponentially growing queries. I've seen poorly written queries return such large datasets that they cause servers to run out of memory processing the request.

Unless you're doing stock trades or black Friday sales, though, it can be pretty hard to pin down a specific inefficiency to a concrete measure of list income. Instead, people move on from products for general "we don't like it" vibes.

The most concrete measure, as someone else pointed out, is heavily inflated PAAS spend caused by poorly written code that requires beefier than necessary servers. In theory, moving faster means you're spending less money on developer salaries (the Ruby/rails mantra of old) but there's a distinct tipping point where you have to pony up and invest in performance improvements.