Finder tags: how to decide what you need
Advice about Finder tags almost always starts with the palette. Assign red to urgent, yellow to waiting, green to done, and get to work. That is the wrong end of the problem. The palette question only matters once a harder one has been answered: does this particular classification belong in a tag at all, rather than in a folder name, a filename, or a saved search. Getting that wrong is what produces the familiar half finished state, where a few hundred files carry colors, several thousand do not, and clicking a tag in the sidebar returns a slice of the disk that nobody trusts. What follows is a way to sort each candidate classification into its correct home before any color is chosen.
Start from how the file will be asked for, not from the file
The usual design method is to open a folder, look at the files, and ask what to call them. That method inflates. Any given file has a dozen properties worth naming: the client, the project, the document type, the deadline, the stage it is at, the person who sent it. Name all of them and the result is a list of twenty five tags that nobody can hold in their head.
Working backwards produces a much shorter list. The question is not what this file is, but what sentence will be used to ask for it later. The invoices from last quarter. The version that goes to the client. The ones that came back with comments. Say those sentences out loud and something useful happens: most people run out at three or four. The properties a file has are many. The handles anyone actually reaches for are few.
Properties that never appear in one of those sentences should not be promoted into a classification at all. Anything visible from outside the file, such as its extension, its modification date or its enclosing folder, is already searchable without marking anything. What deserves a mark is the small set of facts that cannot be seen without opening the file and that show up in the way it gets requested.
Four questions that place a classification
Take each candidate from that shortlist and run it through four questions. The answers determine where it lives.
Does it change over time? A marker that moves from unreviewed to reviewed to shipped behaves differently from a permanent attribute like which client a file belongs to.
Does it cross folder boundaries? If every file with this property already sits in one folder, the folder is the classification and nothing more is needed.
Can a machine express it as a condition? Extension, date, location and words inside the document can all be written as search conditions. A judgement about quality or intent cannot.
Does the file leave the Mac? Being handed to someone, copied to a stick, or shared with a Windows machine changes which storage is safe.
| Changes over time | Crosses folders | Expressible as a condition | Where it belongs |
|---|---|---|---|
| Yes | No | Either | A folder per state |
| Yes | Yes | No | A tag, used sparingly |
| No | No | Either | The folder structure itself |
| No | Yes | Yes | A saved search |
| No | Yes | No | A tag, the intended case |
The fourth question does not place anything. It vetoes. A classification that the table sends to a tag but whose files will travel should be written into the filename instead, because tags are stored in an attribute attached to the file rather than in its contents, and Apple's own description limits the guarantee to two places.
Tags work with all your files and folders, whether you store them on your Mac or keep them in iCloud Drive. Source: support.apple.com
Read the other way, that sentence makes no claim at all about network shares, third party sync clients, USB drives formatted for cross platform use, or the person on the other end who is not running macOS. Classifications attached to outbound files are better carried by names, which survive every filesystem and every archive format.
Why conditions beat tags whenever both would work
When the third question comes back yes, the saved search wins, and the reason is not convenience. It is coverage.
A tag applies to the files that existed and were selected at the moment someone applied it. It is an operation on the past. A saved search stores conditions, so it keeps matching files that arrive afterwards. It is an operation on the future.
Whenever you add, change, or delete a file on your Mac that meets the criteria, the list of files in the Smart Folder is automatically updated. Source: support.apple.com
The gap shows up wherever the population grows. Take a classification like PDFs received since the start of the quarter. As a tag, it generates a small chore every time one arrives, and the chore gets skipped on busy days, which is exactly how the trusted set turns into the half tagged set. As a condition, the maintenance cost is zero and there is nothing to skip.
The Finder can build these from any search by adding criteria and saving the result, a process Apple documents in Narrow your search results in Finder on Mac. Adopting one habit, checking whether a classification can be written as a condition before reaching for a color, tends to cut a proposed tag list roughly in half. Do the cutting first. Assigning colors and then discovering that four of them were conditions all along means unpicking work that has already spread across thousands of files.
The keyboard sets the ceiling on how many tags survive
With the placements decided, the remaining question is how many tags the scheme can carry. That number is not a matter of taste. It comes from the hardware.
The Finder keeps a set of favorite slots. Reading FavoriteTagNames out of the Finder's preferences returns eight entries: one empty entry followed by the seven default color names. The empty one at the front maps to Control-0, which strips every tag from a selection, so seven slots are actually available for assignment.
Seven is therefore the ceiling for tags that can be applied without looking. Anything beyond the slots has to be picked from a menu or typed, which costs a few seconds per file. A few seconds sounds trivial and is not, because it is precisely the amount of friction that gets skipped when the day is busy. Skipping produces gaps, and gaps destroy the value of the whole set: once tagged and untagged files are mixed, a sidebar click no longer answers a question, it produces a sample.
So a scheme meant for constant use should fit inside seven. Seven slots cannot host three separate axes at once. Pick the single axis with the highest rewrite rate, give it the slots, and push the remaining axes into folders and filenames where no such limit applies. Classifications that come up once a month do not belong in the slots at all, and usually do not belong in tags either. They belong in saved searches.
Naming rules decide whether the set is still legible in six months
Once the count is fixed, the names matter more than the colors, and for a reason that is easy to miss. Spotlight stores only the tag names in its index. Reading the indexed value back with mdls -name kMDItemUserTags returns the names with the color numbers stripped away. The color a person sees in a list view is the Finder matching a name against its own palette afterwards.
That has a direct consequence for design. Color is presentation. Name is data. A scheme where the names are careless and the meaning lives in the colors has, from the perspective of search, classified nothing at all.
Two naming habits carry most of the benefit. The first is prefixing the axis, as in state/waiting and state/shipped, so the list groups itself and a second axis added later does not interleave with the first. The second is refusing near duplicates. Pairs like reviewed and to review, or done and complete, will both end up in use within a few months, and after that any filter returns half the files it should. Resolve those collisions while the list is still short, because merging them later means touching every affected file.
Color assignment works better when it follows the order of a process rather than a meaning. Left to right in workflow order means Control-1 through Control-4 walk a file forward through its stages, and nobody has to remember what purple stands for. Attaching fixed meanings to specific colors breaks as soon as a fifth meaning is needed and the palette has run out.
Test the design on two weeks before applying it to everything
A finished design should not be applied to thousands of files immediately. Run it on a sample first, because reversing a tag scheme is far more work than creating one.
A reasonable sample is whatever was touched in the last month, run for two weeks. Three things get measured at the end. How often did a file fail to fit the chosen axis, which shows whether the axis matches reality. What proportion of the sample ended up unmarked, which shows whether the scheme is too heavy to keep up with. How many times did something outside the seven slots get used, which shows whether the count is right.
A fourth measurement is worth taking if the scheme will ever be handed to someone else. Ask a colleague to place five files using the written rules and compare the result with where those files would have gone otherwise. Disagreement is not a sign that the colleague misread anything. It is a sign that a rule relies on context that only exists in one head, and rules like that quietly stop being applied the moment the original author is busy.
When the test fails, the correction is almost always to remove tags rather than add them. The feeling that a tag is missing usually turns out, on inspection, to be a classification that folders or conditions should have absorbed in the first place. Adding one more tag relieves the symptom for a week and makes the gap problem worse.
Pruning belongs to the same exercise. The Finder's settings hold the list of every tag that exists on the machine, which is the inventory rather than any individual file. Removing an entry there is a single operation that detaches that tag from every file carrying it, so it is worth confirming that nothing outside the test sample is still using the name before clearing it out.
What to decide first
Before choosing any colors, write down the three or four sentences that will be used to ask for files later, and run each one through the four questions. Conditions go to saved searches, states go to folders, and only cross folder judgement calls that no condition can express stay as tags, capped at seven. That exercise leaves a short list of rules and queries to write, and doing that work in a file manager that keeps folders, a terminal and an assistant in one window removes the window switching it otherwise requires. How that arrangement compares with the Finder is set out by Atriens.
Frequently asked questions
How many Finder tags is too many?
There is no technical limit, but there is a practical one at seven, because that is how many keyboard slots exist. Tags outside the slots require a menu or typing, which is slow enough that people skip them under pressure, and skipped tags are what turn a reliable set into an unreliable one. Classifications used less than weekly are usually better served by a saved search than by an eighth tag.
Should tags be nouns like client names, or states like in progress?
Both work, but not in the same seven slots. States change constantly and benefit most from being one keystroke away, while client names never change and are often better expressed as folders or as a saved search. Pick whichever axis has the highest rewrite rate for the slots, and place the other axis somewhere without a limit.
Is it worth tagging files that will be sent to other people?
Generally no. Apple's documentation covers files stored on the Mac and in iCloud Drive, which leaves out the routes most outbound files take. A classification that needs to arrive with the file should be part of the filename, since names survive every filesystem, archive format and operating system the file might pass through.
What is the fastest way to tell whether an existing tag scheme is working?
Look at the proportion of files in its scope that carry no tag at all. A scheme in good health covers nearly everything in its own territory, so a large untagged remainder is the signal, regardless of how sensible the tag names look. The usual cause is too many tags rather than too few, and the repair is to merge or remove rather than to add.