We commonly think that promising creates reason for us to act. If I promise to have lunch with you, there will then be reason in favour of me doing so, and there will also be reason against me doing anything else. But this seemingly plausible view generates a puzzle: if making a promise gives us reason to do something, moral agents can sometimes promise to fulfil a less weighty obligation, and thereby ensure that they will do all they ought to do by fulfilling this obligation. Call this the puzzle of suboptimal promises. In this paper, I examine and reject two proposals for responding to this puzzle, before presenting my own solution. I argue that we should conceptualise promising with reference to what Joshua Gert calls the ‘justifying/requiring strength’ of reason. In particular, when an agent S promises to do something, the requiring strength of doing that thing increases in virtue of the promise.