Systems have to be good enough, not perfect

One of our bedrock beliefs around here is that humans are fallible, so therefore will be any system for anything built, run and maintained by humans. It's one of the reasons we're so in favour of market solutions, markets containing the feedback mechanisms to kill off the truly dreadful mistakes in a manner that planning - with its fondness for reinforcing failure - just doesn't.

We're thus entirely fine with systems which are just good enough, not perfect. Sure, we struggle more, search for ever less tolerance of failure the greater the stakes at issue but still, we've got to accept that we should only struggle so far.

Tens of thousands of EU migrants could lose their right to be in the UK after Brexit - and the authorities will not know who they are, a new report warns.

EU citizens must register using an online system to secure "settled status" when the UK leaves next March.

The government has said it expects about 3.5m applications.

But the Migration Observatory said ministers had no precise figures for how many EU citizens were living in the UK and how many plan to stay.

The think tank, which is based at Oxford University, has previously warned that thousands of EU citizens could inadvertently become illegal residents in the UK after Brexit despite meeting the required criteria to stay.

The claim is that we must now go and find who these tens of thousands are and, well, do something. Our reaction is rather different. Objectively, asking people to use an app, upload a photo and pay a reasonable processing fee seems reasonable. And an error rate of tens of thousands among 3.5 million sounds remarkably close to perfection for government work. We'd suggest that this is about as good as it gets, better left well alone in fact.

That idea that the perfect is the enemy of the good applies here. For how intrusive would the State have to become to approach solving the last 0.8% or so of the problem? We don't even manage that for investigating murders, do we?