This paper addresses shortcomings in the scholarship about ‘wicked problems’, and suggests ways of tackling them. Firstly, accounts of these problems tend to ‘totalise’, regarding them as intractable masses of complexity, so conflict-prone and/or intractable that they defy definition and solution. By contrast, we put forward a more nuanced analysis, arguing that complex problems vary in the extent of their wickedness, via such dimensions as their cognitive complexity or the diversity and irreconcilability of the actors or institutions involved. We propose a typology of different forms of wicked problems. A second shortcoming, linked to intractability, is that the favoured means of tackling wicked problems has tended towards ‘one best way’ approaches, most commonly collaboration with key stakeholders. Moreover, particular forms of collaboration tend to be routinely applied in ‘one-size-fits-all’ fashion to a variety of situations – notably with a plethora of actors, and a focus on governance rather than implementation management. We put forward a contingency framework, based on our typology, proposing which types of collaboration are suitable for which types of problem. Finally, we argue for a more realistic standard of success in dealing with wicked problems, especially the most difficult ones. To call for the ‘solving’ of these problems is to set up a standard which is not only impossible but also perhaps unnecessary. We argue that we do not so much ‘solve’ wicked problems as make progress towards improvement or towards better managing them. We spell out a more realistic version of ‘progress’.