Often goals are: crank out X with programmers Y on deadline Z. I assume the client has no language/framework preference.
If X is a greenfield webapp and Y are not Haskellists yet (language has a steeper learning curve then, for instance, Ruby), and Z is very soon then I will probably choose Rails (given that programmers have some experience with dynamic languages and/MVC framework, and I'm fluent with Rails).
Notice that maintenance costs are not a factor. This is often not accounted for, but a large part of a webapp's costs are maintenance: and this is where Haskell shines: maintenance and refactoring.
Let's say I'm starting up, then low maintenance costs are of PRIME importance. You want to be able to keep adding/improving with a small team, not being bogged down by an endless stream of bugs only found on production -- nothing ruins flow like that.
Does using Yesod mean that you can deliver code that better meets your goals? In a business environment, that might mean easier maintenance, or better performance. Equally, it might mean faster development times, where maintenance is less of a priority. If it meets your goals then awesome! If it doesn't, then there are lots of other things you can use – including Rails.