The principle is actually very sound - a simplified system that makes sure any amount of work pays (in contrast to partial return to work leading to outweighed cuts to benefits). It's been the implementation that has been a massive balls up.
Like most big projects like this, I imagine it is the army of consultants who have been brought in who have ground everything into such a mess.
You can't build anything without consultants. They may be coders on day rates or they may be shopped in by a systems integrator, but without consultants there aren't any IT projects. Not in Whitehall, anyway. There's no such thing as turnkey COTS for national benefits administration.
UC as a concept failed because it started off with the sensible aim of making work pay, and was then recast as a means of aggressively reducing the overall benefits bill. Tougher scrutiny, complex hurdles and slower payment are features rather than bugs.
UC as a software project failed because a) the scope and mission of the project veered crazily over time, as above, b) it was cast as an agile project for reasons of political IT fashion, but the customer wasn't agile and some fundamental architecture questions were simply ignored - two years into coding, there still wasn't clarity on whether an enterprise service bus would be deployed - and c) again for reasons of political fashion, the procurement was multi-supplier, which has meant that it has never been obvious which arse to kick, and who is ultimately responsible for making the system work.
And the constant change of both political and Whitehall masters has screwed things, as well.
If there's a good side to this, it's absolute, solid proof that a replacement customs declaration system or a magic NI border solution will take ten times as long and ten times as much money as originally planned, to achieve a tenth of the required functionality. The UC debacle is an irrefutable knock-down to Legatum technology hand-waving.