If a project isn't going to DeliverValue in excess of costs, it's better to cancel it early than late.
This would argue that a project that won't DeliverValue shouldn't be started.
But sometimes it's impossible to know the true cost (or value) of a project without trying it out first. Experiments are worthwhile. In these cases, EarlyCancellationIsSuccess if the team learns that the project won't be successful. The earlier, the better.
Like folding your hand after the flop in Texas Hold'em. You've saved money by getting out early.
See IsEarlierCancellationFailure (just point to this? It already covers the above points and more) but does this make the point more aggressively?