Cost of bad architecture

From a recent discussion with former colleagues of mine:
What are the consequential costs of a bad architecture, a wrong decision, a bad design?
(Design as in Joel's upcoming opus magnum).
About 5 years ago the company (my former employer) had to decide between 2 data models: one (everyone thought) was perfect, one was less costly to implement.
Guess which road they took.
Less costly - or as we now know: cheap(er).

Everyone knew that the other model was better. Everyone said so. In the end, the lesser cost of (the first) implementation ruled out everything else.
Now, 5 years later, we might even have an estimate on the actual cost of this model... every odd workaround, every quirk in other applications necessary, ... We don't have the exact figures, but everyone agrees, that all this (over the last 5 years) turned out probably more then twice as expensive as the "nice" model.
I'm pretty sure we could even prove it (if only anyone was listening).

So the big question for me is: could we have known (or calculated) this upfront ? At the time of the decision?
Not now... now is cheating.
post festum is for wise-guys.

Back Then, upfront, a priori, ante festum... That's what I'm interested in!
Can anyone estimate what a bad design will cost?
The actual cost difference between design A and B over the next - say - 5-10 year?