[quoting Jadav, 3 Oct 96] A few years ago, Michael Allen (of Allen Interactions, Minneapolis) propounded the model of Successive Approximation... And at some point, we freeze the design because there is no more time and money for design!
I'll just note that in HCI design, there is agreement about iterative design (the cycles) and uncertainty how to decide when to stop. I recognize the pragmatics of the indicated approach, but there is another proposal on a more theoretical basis. The proposed approach is to decide as part of the original specification just what are acceptable metrics that would indicate achieving usability goals. These metrics are then used to indicate that the cycle can stop. For instance, you might specify things like: eight out of ten transactions processed error free, a transaction takes no more than 30 seconds, etc.
Translating these into educational products might mean adding some educational goals: learners will successfully complete in 15 minutes, 95% of learners will successfully be able to identify causal factors in the system after completion, etc. (OK, these are off the top of my head, but I hope you get the idea.)
Of course, pragmatics may make this untenable, but it's a nice ideal!