The context quandary: 6 ways to make @computer useful again

Contexts made perfect sense as I read Getting Things Done, but the experience didn’t live up to my expectations. For edge cases like @errands1 and @agendas, they were wonderful. For the bulk of my tasks, contexts added little value. That’s because the situations where I don’t have access to a computer and the internet are as rare as rocking-horse poo. With more than 80% of my tasks landing in the @computer context, it was next to useless as a filter.

The Context Quandary: 6 ways to make @computer useful again

photo credit: adapted from Hanna Wei via unsplash

I experimented with various approaches to making the @computer context as effective as @errands. Let’s do a quick recap on contexts before I get into the detail.

TL;DR: In today’s constantly connected world, @computer holds little value for segmenting tasks. After much experimentation, I now have a useful set of contexts that replaced the traditional @computer setup. The kind of creativity needed and the level of attention required are the variables I use to group computer-based tasks. This approach delivered an unexpected benefit by gathering the work that matters most in a specific context.

A quick recap on contexts

Contexts offer a way of grouping tasks by the tool, location, person or situation that must be present before you can action the task. They look across projects to allow batching of tasks with similar characteristics. Contexts provide another dimension for breaking down long lists.

Contexts look across projects to group tasks with similar characteristics

Contexts look across projects to group tasks with similar characteristics

You know how paralyzing it is when you have too many options. This is why product manufacturers offer a limited choice of colors. Once you start capturing everything that has your attention, scanning the entire list to select your next action quickly becomes overwhelming. Contexts provide a means to filter out the options that are unavailable or inappropriate to act on in your current situation. If you’re out buying groceries, tasks outside of @errands (and perhaps @calls) are irrelevant. There’s no need to see them.

You only need to consider tasks that can be actioned in your current situation

You only need to consider tasks that can be actioned in your current situation

Good task management applications remind you of things when you need to know about them. They should also hide tasks you don’t need to consider in the current situation. Wading through a list of tasks you can’t action is a waste of energy and makes it harder to make a smart choice about what to do next.

Contexts also have an opportunistic function. An agenda context for things you need to discuss with your boss lets you retrieve all the potentially relevant topics during an unexpected meeting.

Contexts as constraints

I think of a task’s context as the most significant constraint preventing me from starting. Common constraints include:

  • availability of a tool or person;
  • being in a specific location;
  • having enough time to make appreciable progress;
  • having sufficient physical and mental energy to tackle the task.

Mobile computers, pervasive WiFi, and sophisticated videoconferencing tools have tamed some of these constraints. Locations, tools, and access to specific people aren’t the challenges they once were. With the external limitations all but gone, internal challenges are the biggest barrier to progress. Making enough time (a priority and scheduling problem) and being able to direct appropriate attention to the task (a focus problem) are now the big constraints.

Context options

I’m not the first person to be challenged by an unwieldy @computer list. Here’s a summary of the approaches I explored along the way.

1. No contexts

Some people have ditched contexts altogether. They advocate contexts are irrelevant when a computer is the means to carry out most tasks. I still find contexts useful, perhaps even more so when all our work looks so similar.

As our work becomes less tangible and takes on a certain sameness, we need to find more nuanced ways to classify it. Throwing the baby out with the bathwater isn’t the right answer.

Eliminating contexts wasn’t something I considered. Splitting the amorphous blob of @computer tasks was my goal. Don’t read this as criticism of the approach if it’s working for you, but it oversimplifies my circumstances.

2. Connectivity

Internet connectivity is another variable people have used to split the list. For frequent travelers, @computer and @computer-offline could make sense. It’s not useful for me since I’m almost always connected.

3. Device

This approach involves tagging different devices for different tasks. This can be an option if you use one computer for work and another for home, or a desktop computer and a laptop.

I experimented with @computer and @computer-mobile (phone or iPad) for a while but found the division too artificial. Most of the time it didn’t matter, though there were tasks I wouldn’t attempt unless I had a physical keyboard to work with. If we’re not there already, it won’t be long before an iPad will be a viable primary computer for many people. Increasingly capable tablets and phones makes this approach less meaningful with each passing day.

4. Tools or activities

This approach groups tasks by the specific application, by the activity, or a hybrid of these two methods. Examples are the easiest way to illustrate this idea.

Tool-based contexts:

  • @email
  • @excel
  • @word
  • @photoshop

Activity-based contexts:

  • @read
  • @write
  • @design
  • @social media
  • @email
  • @review

Activities are a better choice when the process involves many tools. Writing is a good example.

This approach appealed to me but the drawback I found is that it doesn’t take constraints into account. @Excelp could have some mindless data entry tasks sitting alongside a task to create a complex financial model. @Write has a mix of tasks that need sustained periods of focused attention and others I can pull off even when I’m knackered. For each task, I had to ask myself, “Is this something I have the time and energy to tackle?”. This is more work thank it seems so the search continued.

5. Time and energy

With geographical constraints rapidly disappearing, time and energy are the dominant factors. They are the basis of Sven Fechner’s context model.

Sven Fechner's context chart

Sven Fechner’s context chart (photo credit: used with permission from Sven Fechner at Simplicity Bliss)

This approach has become quite popular since Sven floated this idea2. His method appealed to me since it addresses the two real constraints. But I ran into some implementation problems.

  1. My brain interprets @FullFocus as @HardWork creating a subtle but palpable resistance to engaging with this context. This nudged me toward tasks that made me feel productive but didn’t move the ball forward.
  2. I don’t always notice shifts in my energy state because they change gradually. I can spin my wheels for longer than I’d like to admit before I realize the most useful thing to do is to go for a walk.
  3. Sometimes I’ve just got to get things done even if my energy levels are sub-optimal.

Try out Sven’s model if your contexts need a shake-up. Unfortunately it was psychologically incompatible for me.

6. Creative mode

This leaves us with the approach that I have settled on. In his book, The Productivity Habits, Ben Elijah describes contexts as specific configurations of space, time, and thought. Every task requires a certain level of availability, attention, and creativity. Ben’s context triangle model recognizes that our physical environment, our mood, our resources and other factors influence our effectiveness at engaging with tasks. The context triangle helps you establish the compatibility between situations and tasks.

The context triangle

The context triangle (photo credit: used with permission from Ben Elijah)

Technology has (mostly) dealt with availability leaving the creativity and attention dimensions for us to play with.

Creativity can be:

  • Open. Open mode demands divergent thinking to generate new possibilities. This is where we synthesize and generate ideas to create something new. It can be a loose and exploratory process with dead ends. Analog tools and an information-rich environment such as a café, library or getting outdoors can help stimulate the process. Examples include brainstorming ideas to solve a problem or writing a crappy first draft.
  • Closed. Closed mode is about compiling, polishing and refining. Convergent thinking removes the unnecessary to focus on the essential. This mode is about precision and getting the little details right. This usually means digital tools and a distraction-free, bland environment can help. Examples include proofreading a final draft, comparing features to decide between options, or data entry.

Attention can be:

  • Deep. Deep activities demand your full attention and longer blocks of time. This mode typically has a ramp-up time before you get into flow. Researching a new topic or planning a large project are examples. In my experience, these tasks need a block of an hour our more to make any real progress.
  • Shallow. Shallow mode is for activities that need shorter blocks of time to complete (15 minutes to an hour) and aren’t so intellectually taxing. Examples include making an online purchase, compiling monthly expenses, or processing email.

These time boundaries are somewhat arbitrary and flexible, but fit the rhythm of my work. You might need to adjust them for your situation.

Combining these dimensions creates a matrix of four different contexts. Here are a few examples.

Creativity / attention matrix with example tasks

Creativity / attention matrix with example tasks

I’ve augmented Ben’s ideas by adding a context I first saw on Asian Efficiency. My @do context is for tasks that take little effort. The kind I can crank out even when I’m tired (a bit like Sven’s @braindead context). If I estimate a task will take 15 minutes or less to complete, it goes here. Tasks with longer expected durations land in either the @open:shallow or @closed:shallow context. @Do is useful when I have little blocks of time before the next appointment. It makes it easy to pick out some quick things to pick off without scanning through a longer list.


Sometimes several contexts are relevant in a particular situation. OmniFocus only permits a single context for each task but you can aggregate several contexts using custom perspectives3. The nomenclature varies, but most tools support this idea. It’s effectively a saved search that defines a set of contexts you want to see together.

Here are a few examples of custom perspectives:

  • Out and about. @Errands is an obvious choice, but this perspective also includes @calls.
  • Charged. This combines @open:deep and @closed:deep. The things that move the needle often need high energy and attention (the first part of the day for me).
  • Drained. When I’m feeling tired and concentration isn’t at its best. This perspective shows @do, @open:shallow, and @closed:shallow.

Wrap up

@Computer is no longer useful for segmenting tasks in a world where we’re always connected. After trying various approaches, Ben Elijah’s context triangle model is the one that’s been most effective. If you’re not happy with your context setup, pick one of the alternatives and try it for a month. Remember, we all have different circumstances and psychological make-ups. These models are a good starting point but you’ll need to adapt them to suit your own circumstances.

Question: Are you using another approach that I haven’t covered? Let me know about your experiences by leaving a comment below.

  1. I have prefaced specific context names with @ throughout this post. 
  2. I don’t know if this was entirely Sven’s idea, but he was the first person I saw using time and energy to define contexts. 
  3. You need the pro version of OmniFocus to use custom perspectives. 

You may also like...

17 Responses

  1. Marco Navarra says:

    Love it! On point! Thanx for the Creativity/Attention Matrix and the new Contexts I’ll be using!

    • John Scullen says:

      Thanks for the feedback Marco. Glad you found it useful. It’s the first time since I started by GTD journey in 2010 that I’ve felt content with all the contexts I’m using.

  2. This was article spot-on about the @Computer lacking the necessary detail to be meaningful. I like the idea of continuing to develop a system that works for your particular work environment. I don’t think I can effectively employ the Creative-Attention model (though I could see how this would work very well for writers), still, it will lead me to further review my @Computer tag to develop a system that better captures my nuanced ways of operating. Thanks for the thoughtful article!

    • John Scullen says:

      Thanks for stopping by Scott. Glad you found it interesting. You’re right about the need to periodically look at all the parts of your GTD system. As your life changes, the system needs to adapt to support those changes.

      All the best in finding an approach that works for you. Drop me an email if you want to bounce ideas off someone.

  3. Francis Wade says:

    Your post is great! A few comments…

    I think the term context has become used in different ways that are confusing. I recently wrote an article on Quora that tries to address the issue.

    It’s longer than a comment-length but the basic idea is that when we create tasks we unconsciously add attributes at the same time. The attributes added vary per person, but here are some typical ones: urgency, priority, energy, location, duration, due date, equipment, creativity required, etc. We can’t possibly use them all in our task management – what we do instead is use the one or two that represent our most scarce resource.

    Once that’s decided, we make sure to tag each task with an actual value/tag such as Hi, Med, Low in the case of the Priority attribute.

    This attribute is used, as you rightly said, to sort and view tasks so that we only have to focus on a few at a time. For example, when you sort your tasks with the Priority attribute, all the tasks tagged with Hi’s immediately show up.

    In my experience, people use attributes and values interchangeably, and don’t focus on using those tags with the most scarcity for them. The result is a long list of tasks which cannot be properly manipulated.

    I hope this helps. The Quora article is more clear, I think.

    Thanks for your post – again!

    • John Scullen says:


      That’s a thoughtful and comprehensive response you put together in your Quora post! Thanks for your insightful comment here too.

      It got me thinking about two things. The first was the Four-Criteria Model for Choosing Actions in the Moment that David Allen discusses in chapter 9 of Getting Things Done (p. 205 in the 2015 edition). The second was the theory of constraints.

      David Allen’s four criteria are:

      1. Context
      2. Time available
      3. Energy available
      4. Priority

      David uses context to refer to a person, tool or location that must be present to make progress on the task. These are constraints. Time and energy, also constraints.

      While the specifics vary from person to person, there will only be a couple of big constraints that stop you moving on a task. The trick is to understand the internal and external constraints you are most likely to bump into.

      Priority is a function of how well the task lines up with your goals and what you want to achieve. Constraints help dictate a shortlist. Priority helps you pick the best option from reduced set.

      Thanks for commenting. This has helped me to further clarify my thinking.

      • Francis Wade says:

        Sounds like we are both fans of Eli Goldratt – he was right about so many things and I quote him in my book’s introduction. I read The Goal in the first few months of my first full-time job and could not put it down!

        Unfortunately, he was way ahead of his time / the scientists of his time.

        After searching hard, I can’t find anyone who has scientifically studied the issue we are talking about – how we instinctively constrain tasks using attributes/tags that we find most scarce. We may also be ahead of our time…

        Beside identifying constraints, attributes/tags should also help us decide which task to execute next – by giving us a helpful indication.

        Thanks for the comment! It’s great to chat with you – someone who is putting provocative ideas out there. Let’s keep in touch.

        • Joel says:

          If I may, I’d like to jump in on this exchange. I’ve already posted a bit below, so this will be a further elaboration of that, but you’ve both gotten me thinking, and thanks for that!

          It seems to me important to distinguish between two situations in which contexts can be useful. Call them “deliberating” and “engaging”. I think that David Allen used the term “contexts” for both.

          The first is a matter of narrowing down one’s options, and having more factors helps to set constraints. In winnows down the options, so that the final choice is more manageable. There is research on this in decision theory, both Michael Bratman’s theory of planning (at Stanford’s Philosophy Department) is one of the most influential, also in artificial intelligence.

          The other context was the one I was getting at in my last post: it’s a matter of setting yourself up so that, when you find yourself in the context, it’s pretty clear what you need to do. You just engage. This is David Allen’s talk of “cranking widgets”. In this way, contexts can work like notifications, triggering one (or one of several actions).

          Both of these can work. But I find that the “constraint on deliberation” function of contexts doesn’t really help me much. There are just very few situations in which I’m limited in what I could do. Maybe “@rainy-day-at-the-country-house”? But even then, given how wired and mobile we all are, there are a million things most people could do even then. So, it seems more as though the role of contexts part of a set of standing intentions: whenever I’m in this situation, I should engage in doing this. But then, the fact that I’m in a particular situation needs to be something that just hits me in the face.

          If that’s the best purpose for contexts, for you, then the interesting becomes, What makes for good contexts? Locations often work really well for this. If I’ve got reading that I intend to do on the commuter train, then @train is a really good way to ensure that, once I’m at my seat, I just start cranking that widget. I just need a habit of consulting my “@train” list as soon as I sit down (rather than opening Twitter). My sense is that – again, for this application of the idea – a good context is a recurring, distinctive, demarcated situation. It shouldn’t be a one-off event (my cousin’s wedding) but one that recurs (I’m logged into a particular service portal of my employer). It shouldn’t be something that applies all the time (@computer), since that doesn’t add anything. And the distinctive situation should ideally be clearly marked off from other situations, so that it can serve as a trigger. It has to be something that can easily notice (get in the habit of noticing). It has to be naturally salient – the relevance just hits you.

          I should add that most of my GTD actions don’t have a context at all (or they have @any). Usually, it’s the project that provides the structuring focus for deep work.

          • Francis Wade says:

            Thanks for jumping in!

            I struggled with the term “contexts” for a while, until I realized that GTD has chosen it as a label for a small subset of all the attributes we could use to tag tasks. For example, the ones you highlight above: “context-as- location” and “context-as-situation” are useful and I happen to have one for @Miami-shopping. It’s a list of purchases I want to make on my next trip to the city.

            The decision to visit Miami triggers the use of the list. I only visit it beforehand to add a new item.

            This kind of opportunistic tagging is one I have found useful also.

            However, it’s value is limited to me because the dominant constraint I deal with is not context-as-location/situation – it’s time… a la implementation intentions.

            In the examples you gave below your experiments sound interesting! In the absence of sensors to detect when your coffee is finished or when you arrive home, you may get by with a schedule that can be auto-scheduled on the fly. I use SkedPal and while there’s no Android app, it’s possible to reschedule my day once either of the events you mention takes place.

            Not perfect… but a few years ago I realized that GTD and Total Task Scheduling were only possible because they both use tags that are required to use implementation intentions in the practical world.

            Interesting stuff!

          • John Scullen says:

            Hi Joel. Thanks for chiming in on the discussion.

            You’ve put together a handy list of characteristics of effective contexts. If I can summarize:

            1. Contexts should be personally relevant. They must make sense in your current reality. Your circumstances and mental models are unique which explains why there are so many different approaches to contexts.
            2. Contexts should be discrete. When the edges of two or more contexts overlap, choosing the right one becomes so much harder and their usefulness rapidly diminishes.
            3. Contexts should have an obvious trigger. An environmental or habitual cue signals activation of the context. This is what makes location-based contexts so effective and @computer so challenging.
            4. Contexts should represent a recurring circumstance. The trigger occurs at least semi-regularly. If you don’t find yourself in a particular context often enough it won’t be useful. You might have identified a new project instead of a context.

            The new contexts I adopted satisfy the relevance and recurring criteria well. I have a good handle on when I can tackle deep work and when I should do something else. This meets the trigger criterion. The final category, discreteness, is the most ambiguous but I’m in a way better position than having a single @computer context.

            Your comments have helped me nail down exactly why my @computer context failed and why the replacements are more successful. Thanks.

          • Joel Anderson says:

            I couldn’t have put it better myself, John! Thanks. Perhaps if I had your level of self-insight, the context you mention would serve better as discrete and obvious triggers.

  4. It seems to me that the main reason why contexts work, when they do, is when they provide a trigger for a sequence of actions. There is some work on this in psychology, using the concept of implementation intentions. The trick for coming up with good context triggers, it seems to me, is having the trigger be something that you think about a lot or make a decision. This is why location-based contexts work so well. The big problem I have with energy-based contexts is that I first need to stop and figure out whether I am in a full focus state or not. My brain isn’t very good at that. I’m thinking of experimenting with things like “@after my 1st cup of coffee” or “@when I get home work”. Those seem to me to be clear triggers, not something I first have to be self-aware enough to make a judgment about.

  5. John Leahy says:

    Interesting discussion, I like the idea that @Context should have a trigger and this resonates with my view about the systems we employ. If you place your system on the Creativity/Attention matrix at the point of use it should require zero attention and zero creativity. A successful system needs to work like falling off a log. If you need to think too much about it or have to pay it too much creative attention it just will not work at the time of stress, exactly when it is most needed. The creative energy and design needs to be front loaded – I’ve lost count of the ‘clever’ systems I’ve designed only for them to fail at the point needed because they require ‘clever’ input.

    I understand what Joel says about the context needing to be discrete but to me not in the sense that the context simply means one thing rather it conveys a unique set of things e.g. @train is shorthand to convey a particular set of circumstances in a simple format that does not require a complimentary set of tags like @not confidential @no internet @little space @finite time @background noise @outside office hours. And that is the problem with @Computer – it no longer conveys a unique set of circumstances and is not discrete in that sense.

  6. John Scullen says:

    Good points John. If your system still works when you’re too sick to get out of bed or when life gets out of control, you know you’re on a good thing. I too have seen systems fail under stress when I’ve tried to make them too clever.

  7. Chris Mousley Jones says:

    What an excellent discussion! Thanks all for your thoughtful insights. This post has really helped me clarify what wasn’t working with a couple of my key sclerotic contexts (“key” in the sense that there were many next actions on them, and pretty much all of them stuck). I am now reviewing some very long, elderly lists and re-categorising. Thanks!

  1. October 28, 2016

    […] I wrote about applying Ben Elijah’s model to replace my @computer GTD context, I noticed an interesting pattern. Creative1 projects follow a predictable trajectory through the […]

Leave a Reply

Your email address will not be published. Required fields are marked *