A Mental Model For Personal Organization (Part V)

Or, how to organize anything with the “three axis” metadata method.

Cameron Flint
14 min readDec 4, 2021
Photo by Reuben Hustler on Unsplash

List of all articles in this series:

  • Part I starts with the core scenarios and requirements, and introduces the original 9-part model. Plus a quick aside on metadata.
  • Part II demonstrates three kinds of hierarchical taxonomies, and offers the “center of gravity” principle for dealing with information scatter.
  • Part III offers tool ideas for fundamental workflows, covering many apps and services. Discusses the intertwining of analog and digital workflows.
  • Part IV reflects on learnings from the original 9-part model, then evolves toward a more nuanced model featuring 5 major divisions and 15 minor areas.
  • Part V dives deep on metadata, providing a “3-axis” rule of thumb for ensuring your content can be rediscovered by you. Discusses a strategy for avoiding metadata rot.

It often happens that when I add something new to my personal organizational system, I struggle for a bit to decide where to put it. Should it go into its own folder? How should I tag it? Should I link to it from an index page? How many keywords should I add?

If you’re already a reader of this blog, you might be familiar with my appreciation for the rule of three when it comes to organizing and navigating information:

  • In Part I of my Mental Model series, I advocated using no more than three levels of nesting when building folder hierarchies.
  • In Part II of the series, I demonstrated a handful of different techniques for structuring a hierarchy of folders that I called the “three taxonomies.”
  • In Part III, I praised Trello’s user interface for what I dubbed its “three-axis navigational system” of Boards, Lists, and Cards.

Here, again, I’d like to invoke this “rule of three” in laying out a mnemonic for how you can take any piece of data and quickly file it within your O.S., in a way that promotes relation, rediscovery, and reuse.

The three-axis metadata method

The central idea behind the three-axis method (called 3XM below for brevity) is to “place” each new item along three primary dimensions, by choosing at least one word from each of three core metadata families:

  1. Dates, times, and locations (the “when” and the “where”).
  2. People, objects, kinds, and groups (the “who” and the “what”).
  3. Projects, areas, and topics (the “why” and the “how”).

These metadata can apply to just about anything you put into your personal O.S.: from notes, to files and documents, to bookmarks, to web clippings, to journal entries, and much more.

The three “axes” of organization under which all essential metadata falls

To help remember the core metadata categories, we can segregate them into three themed groups (or families, or dimensions, or “axes”): (1) time and space, (2) subject and object, and (3) context and relevance.

Aside: If you’re familiar with other organizational methods like Tiago Forte’s PARA method (standing for Projects, Areas, Resources, Archive) or August Bradley’s PPV (Pillars, Pipelines, and Vaults), you might see some parallels to the method described here, though I think 3XM is a little bit more general-purpose than PARA and PPV.

As you pick names from the core categories, it is a good idea to build and maintain a simple index of words you’ve already chosen in the past. This index becomes a “pool” of your personal active metadata, serving to reinforce concepts you’ve already settled on for organizing your data, and preventing confusing overlap of similar terms.

Before we dig deep into examples, let’s review the potential benefits of this system:

  • It helps you more easily navigate back to items later, since there are a small number of predictable paths you can follow back to the target even if you can only think of one “entry point.” This mimics how networks of neurons in the brain are activated to recall a full memory from only one or two associated “trigger” perceptions.
  • It helps you build a compact metadata toolkit by which you can organize all of your information, whether you chose to use tags, labels, folders, or any other affordances to implement it. In contrast to a random ever-growing pile of inconsistent metadata, the metadata toolkit is easy to maintain over time.
  • It helps you rediscover and resurface existing related or recent items that you’ve previously filed. As you “walk” your way along the three core metadata axes to add new stuff, you might stumble upon interesting finds filed previously by your former self under those same categories.

The most important principle to remember when applying any metadata to your system is that metadata should serve you and no-one else. Don’t stick twenty keywords on each and every article that you save, unless doing so satisfies your inner librarian, and/or you truly find it useful.

Use metadata in moderation, like you’d sprinkle seasoning on food.

Lastly, a reminder that the 3XM method — just like all other methods in the Mental Model series — is intended to be compatible with any tool of the trade, whether digital or analog. Although I use #tags and [[page links]] in the examples below (instead of traditional labels, folders, hyperlinks etc.) and I mention certain software products (like Roam Research), there’s no need to switch from something you’re already using as long as it works well for you.

Applying the three-axis method

In order to get a flavor of how 3XM works in practice, let’s study the method using a handful of selected scenarios:

  • Meeting notes
  • Fleeting thoughts
  • Movie recommendations
  • Working drafts
  • Random internet finds
  • Literature notes (reading notes)
  • Journal entries

The scenarios will help us to get a sense of what each of the core metadata categories mean (dates, objects, kinds, groups, topics, areas, etc.).

One major theme you’ll see is the frequent association between dates and other categories. In my organizational method, the timeline is one of the most important and central components of the system, because a linear history provides both sequence and context.

Even though you might not immediately see the value of linking seemingly standalone things (e.g. people, places, entities, topics, etc.) to timelines, I assure you it is one of the most tried-and-true, simple, and effective tricks for staying organized.

3XM applied to meeting notes and 1-on-1's

Meeting notes

When I take a note during a meeting or a 1-on-1, I try to remember to associate:

  • The date and time, including day of the week (first axis).
  • The people participating or attending (second axis).
  • A category label like #Meeting to differentiate meeting notes from other kinds of notes (second axis).
  • An area label like #CompanyName or #TeamName to keep work and life areas separate (third axis).
  • Any current project or topic that the meeting pertains to (third axis).

Note that instead of using an explicit “area” label, I could choose to use an implicit method like a dedicated notebook, file folder, or computer folder that’s clearly labeled by the company I work for.

Each metadata item is listed explicitly for the sake of understanding in the example scenarios, but you can feel free to experiment with other ways to implicitly encode metadata. Doing so can save significant time and effort.

3XM method applied to “fleeting thoughts”

Fleeting thoughts

When I’m struck with a fleeting thought or idea, I usually jot it down even if I’m in a hurry. At a bare minimum, I try to capture one metadata apiece along each axis:

  • The day and time, often automatically recorded (first axis).
  • Any persons, objects, or entities, e.g. #App/Pocket(second axis).
  • If related to a current project or ongoing activity, the unique project label e.g. #MentalModelSeriesPartFive(third axis).
  • If related to a general work/life area, e.g. #Productivity, #NoteTaking, and/or #Apps (third axis).

For example, I once wrote down on my daily notes page:

#PersonalOrganization “I don’t like having my saved articles spread across [[App/Pocket]] and [[App/Raindrop]]. I should only keep articles in Pocket that I haven’t read yet.” — Me, a while ago

In my notes, a major recurring theme is #PersonalOrganization, so in this case I chose it as the leading life area that this thought relates to (third axis).

I don’t need to worry about exhaustively listing all possible relevant areas or keywords for the thought. First, I deliberately limit the number of core life/work areas that I associate notes with, especially quick notes. Second, I know that I can easily tag secondary or tertiary topics upon review (rather than at the time of capture).

The bracketed [[App/Pocket]] and [[App/Raindrop]] words create inline references to those two named entities (apps) thanks to the magic of the software that I’m using, but I could also have just tagged them or explicitly linked to those pages.

The note-taking software just mentioned also takes care of associating each entry on the daily page with that day’s date, such that if I later rediscover it while filtering all notes mentioning [[App/Pocket]], I’ll see which date I wrote that specific entry on.

I’ve found date affiliation for fleeting notes to be more useful than I originally thought, simply because it allows me to go back and see the full context for what I was thinking in a given moment, especially since thoughts and ideas tend to compound incrementally.

3XM applied to movie recommendations

Movie recommendations

Whenever I’m given a recommendation for a movie (or a book, TV show, restaurant, or something else of possible future interest), I add the mention to an appropriate running list. That means capturing:

  • The date of recommendation (first axis).
  • The person who recommended it (second axis).
  • The list or queue it belongs to — books, movies, etc. (second axis).
  • The appropriate genre or subsection, e.g. #Drama (second axis).

For items like this that end up in a specific list, I don’t find it necessary to specify a life or work area, for two reasons:

  1. The context/relevance to me is already embedded in the list name. I.e., “movie recommendations” is plenty specific.
  2. Items in my recommendation lists are in the “outer regions” of my system, in that I don’t know if I’m going to end up keeping any permanent record of it.

Object groups are an important metadata category along the second axis. Lists, collections, and queues are all examples of object groups.

3XM applied to working drafts and documents

Working drafts

I find it useful to keep track of the status of drafts and documents (like this article), alongside where they appear in my notes:

  • The creation and modification timestamps, often automatically captured by the software (first axis).
  • A type label for the #Essay (second axis).
  • The status or process stage of the item, e.g. #Draft or #InProgress (second axis).
  • A relevant life area, e.g. #Writing (third axis).
  • Any relevant topics the essay pertains to, e.g. #Productivity, #PersonalOrganization (third axis).

It may be a surprise that the document’s status metadata (draft/in-progress, scheduled, published, archived, etc.) is considered a subject/object category. But if you think about it, the draft itself is the object, and its status is merely a temporal property of the draft object.

This example illustrates two additional metadata categories within the subject & object axis:

  1. Object identities/kinds (like the title of the essay, and its type “essay”).
  2. Object states (like the status of the document, “draft.” For a person, we might say “moods” instead of “states.”).

If instead of an actual draft this was just an idea for an essay, then I’d use the tag #EssayIdea , which would associate the idea with its corresponding object group (similar to the movie recommendation example).

Object kinds like Essay technically form an object group, too. But I usually find it more useful to browse my collections of things that are grouped on a more intentional/specific criteria than just their type or kind.

3XM applied to a “random internet find”

Random internet finds

When I stumble across something on the internet, such as a tweet, post, meme, comic, quote, etc., I will save it in the following manner:

  • The date of the finding (first axis).
  • The source of the item — #Twitter, #Reddit, etc. (second axis).
  • The type of the item —#Tweet, #Post, #Meme, etc. (second axis).
  • The author, if applicable (second axis).
  • Any meaning the item holds to me personally — is it #Funny? #Inspirational? #Useful? (third axis).
  • The life area it pertains to (third axis).

As you’re noticing by now, sometimes I’ll add more than one piece of metadata along the same axis. This might help with locating the item in the future, but only if the labels I’m sticking are part of my universal metadata index (otherwise I’ll forget they exist).

Regardless of how much or how little metadata I remember to associate with a bit of “treasure” I’ve just found, using the three-axis mnemonic helps to insure that I at least pick one tag in each main dimension.

Another pattern you might have noticed throughout these examples is my loose adherence to the YYYY-DD-MM hh:mm:ss date/time format. Using timestamps like 2021-12-02 15:10 is intentional, because specifying time components in decreasing order of magnitude (year, month, day, …) enables a natural chronological sorting to occur.

Here are a couple additional tips for using dates and times in metadata:

  • If you don’t need to record an exact timestamp, you can freely leave the time components out. E.g. 2021–12–02 for this “random internet find” is fine. I don’t care what time of day I found it.
  • If you’re naming a computer file or an internet resource, consider the path- and URL-friendly format 2021-12-02–15–10 instead to avoid issues with unsupported characters.

Literature notes

I often take reading notes in two stages — first into a small notebook (or onto sticky notes), then later into my digital note-taking system. The analog step helps me stay in the flow while reading the book.

When I’m finished with the book and decide to create a permanent “literature note” within my system, I include the following metadata along with the handwritten excerpts and/or summaries:

  • The publication year, title, and author (sometimes combined into one, e.g. [[2021SmallScottForgettingTheBenefits]]) (first and second axis).
  • The chapter and page number for each quote, excerpt, or summary (first axis).
  • The type of the thing that I read, e.g. #Book (second axis).
  • The category of my life this book is most relevant to, e.g. #MindAndBody (third axis).
  • Optional “keywords” or specific topics beyond the main category: e.g. #Neuroscience #Psychology #Memory etc. (third axis).

There are certain “first-class” object types in my system, of which Books are one. I find it helpful to think of the publication year, title, author, etc. fields as object properties (very much related to the object types and object states mentioned previously).

With something as lengthy as a book, there are easily 10–20+ keywords I can usually think of to classify it under. The problem is that it’s rare that I remember such fine-grained keywords for very long after assigning them.

I’d much rather choose a simple category that’s relevant to me at the top level (like #MindAndBody), than to pretend I need the precision of the Library of Congress. If I do add additional keywords, though, I’ll need to remember to add them to my index.

3XM applied to a journal entry

Journal entry

I consider journaling to be one of my most important habits, even though I’m not 100% consistent about it. When I do practice journaling, the value it brings to my life is multi-fold.

Journal entries are the “source of truth” for how I spend my time — where I went, what I did, what I thought, etc. — so I consider it important to get the details right.

And because it’s easy to accumulate a lot of mundane day-to-day details that I probably won’t ever practically revisit, it’s also important to assign metadata so that I can later see filtered views of my life stream.

Each journal entry includes:

  • The date of the entry — unsurprising by now (first axis).
  • The type of this writing to designate it as an “elevated” note: #Journal (second axis).
  • Any people that I mentioned in the entry — friends, family, colleagues, etc. (second axis).
  • Any places associated with the entry, if special or unusual, like #OregonCoast (second axis).
  • The area of my life the entry most pertains to, e.g. #TravelAndOutdoors, #LoveAndRelationships(third axis).

(Note that in my previous article about Conversational Journaling, I referred to the second-axis items as “Mentions” and the third-axis areas/topics as “Channels.” Use whichever nomenclature you prefer.)

The metadata map

I’ve mentioned a few times that an essential part of being consistent when applying metadata — and choosing it wisely in the first place — is building and pruning an index of the labels used.

Even with the three-axis rule of thumb, it’s altogether too tempting to keep expanding the list of keywords over time. The following is an example from my personal “metadata cheat sheet” that I’ve built as a sort of living map of the labels and tags that I use to organize my stuff.

My metadata “cheat sheet.” View the enlarged version

Remember that while these lists are meant to be comprehensive across the metadata in use inside your system, it doesn’t mean that every new item has to get tagged with one word from every list. I suggest to use the metadata map as a “tickler” for when you know you have a previously-created tag for something, but you just can’t remember what you called it.

Root metadata

With all that metadata, how does one organize a workspace? I like to envision taking a pen or marker and highlighting only the words in my metadata map that I work with most often. Call these metadata the “roots” or “pins” and use them as the primary pivots to input and explore your data.

For example, here is the current crop of terms that I have favorited in my Roam Research sidebar:

My current active crop from the metadata system, pinned to the sidebar in Roam Research

Computer-assisted metadata

Some software products, like DEVONthink for example, or mymind as another, are capable of performing label extraction, AI classification, and automatic tagging of items you put into them.

Feel free to leverage this capability if you find it useful. I tend to be a bit wary of AI-generated metadata, however, because I prefer to build my own mental map of my content. I believe doing so helps aid understanding, synthesis, and recall of the ideas stored there.

Perhaps the best approach is a collaborative one: let the computer suggest, and the human ultimately decide. Evernote’s web clipper is an example of this done right, in how it smartly pre-selects the target notebook based on the page’s content, but allows the user to easily override its suggestion.


Thank you for sticking with me through this whirlwind tour of my “universal metadata system” and three-axis method. If you have feedback, tips, or suggestions from your own experience regarding the effective application of metadata, I’d love to hear from you. You can reach me here in the comments, on Twitter as @camflint, or via email at feedbackforcameron {at} fastmail.com. See you next time.



Cameron Flint

Diving deep on topics related to note-taking, personal information management, and software engineering, with occasional diversions to less nerdy things.