Here is my attempt at a general strategy for managing time. I define productivity operationally here by measuring it in terms of publications (of course, this definition may have critics).
The central point is that your time at work can be divided into productive and unproductive time (see graph), and that both are important; however we should try to maximize the productive time.
The graph may be biased towards the kind of work I do (modeling and experimental cognitive science); other disciplines may not have some of the activities, and the partitioning of your time may well be very different, so feel free to make your own graph with relevant tasks.
Of course there are many more tasks that could be included in this graph in all the branches, so feel free to make your own graph. A third major branch would be tasks that are unrelated to academic productivity, but we have to do anyway (to have a copy of life 1.0).
The delegable tasks are very few, this is bad news. Plus, my experience is that most academics have trouble delegating work and are unhappy with the results. This is why I posted about project management and its importance in the academia.
Note that unproductive time is very needed. You cannot just erase it from your life. In fact, if you do not read papers, do your paperwork, and do some kind of service to the profession, your career will be affected.
So what is a good solution for this conundrum? I propose to do some time boxing: set a time quota for productive time a day (or week) and do the same for unproductive time. The trick to end your day with a feeling of having achieved something is to be sure you have fulfilled your time in the productive area. I’d also recommend to do a large chunk of the productive tasks before entering the gray area of unproductive time (because that way you will not feel guilt). For example you could allocate 2 hrs without interruptions first thing in the morning to your most ‘profitable’ activity (e.g., writing the revision of that paper that needs to be resubmitted!), and then indulge in some unproductive, but pleasant aimless browsing.
One important argument is task dependence: sometimes you just need to have one of the tasks living in the unproductive time area done before you can get a major advance in another productive task (e.g., you need to fill some paperwork to get some funding, or to do to some committee reviewing before your PhD student gets some approval). These tasks usually have hard deadlines and always land you in a difficult situation where you find that their urgency eclipses all other priorities. This is a hard decision to make; I’d try to enforce some timeboxing still, while maybe relocating time in a way that temporarily favors the urgent task that has dependencies; but it should be a temporal thing, and you should go back to your scheduled time boxing that allocates most of your resources to your productive areas.
It would be good to distinguish between things that can decrease your productivity before they increase it, and things that just decrease your productivity.
The first group has things like learning math or programming languages, and testing productivity software. The second group has things like paperwork and service to the profession.
Although making this distinction is purely common sense, it is very useful because (for me at least) it is very easy to spend several hours working hard on something that lives in the unproductive time branch, and then be surprised with how little we have to show for it in the end.