Making a good prediction is hard - even in the short term. Gartner knows this. However, what makes it good is not working out what is going to happen but rather when it will happen. Or even more interesting - when it will happen, why it happens just then - and what someone should or could do about it. That would be a good prediction.
Some examples of bad predictions are for instance saying that newspapers will die. That linear television will have a hard time. That e-books will take over. That iOS can’t maintain its current market position.
All of the above may be true, but without the correct parameter of time they are all fairly useless - and even more so, they are safe. There are tendencies for all of the above. So simply placing a bet saying that they will eventually happen and then take credit for the prediction when it does - well, that is a bit like Nostradamus. Keep it vague enough and sooner or later you’ll be right, to a certain extent.
I tip my hat to people like Horace Dediu and analysts like him. I’ll choose a faulty but honest prediction over a Nostradamus ditto any day. Because these analysts are taking a risk that could prove them wrong in public. I’m impressed by that. And they’re not just waiting long enough to eventually (claim to) be right.