Tags and categories: what's the difference?

Arseni Mourzenko
Founder and lead developer
November 10, 2014
Tags: user-experience 10 tagging 4

This ar­ti­cle was orig­i­nal­ly pub­lished on Stack Ex­change.

Is there a dif­fer­ence be­tween tags and cat­e­gories? Is this dif­fer­ence clear for peo­ple with­out tech­ni­cal back­ground, i.e. or­di­nary users?

Tags are in­her­ent­ly dif­fer­ent from cat­e­gories and es­pe­cial­ly al­low to solve the unique­ness of at­tri­bu­tion prob­lem I've al­ready il­lus­trat­ed in the black sleep­ing cat ex­am­ple. Users, on the oth­er hand, are not al­ways aware of this dif­fer­ence. When I asked peo­ple what is the dif­fer­ence be­tween cat­e­gories and tags, there was a lot of dif­fer­ent an­swers:

  1. “Cat­e­gories are usu­al­ly pre­sent­ed in a form of a tree; tags are nev­er pre­sent­ed this way.”

    Wrong. As cat­e­gories may be flat, tags may be con­tained in larg­er tags which are them­selves in oth­er tags, etc. The sim­i­lar ap­proach is used by Adobe Light­room with its key­words: while a pho­to may have sev­er­al key­words (tags), those key­words may be or­ga­nized in a form of a tree. As­sign­ing a child key­word usu­al­ly (de­pends on the set­tings) as­signs the par­ent key­words as well.

  2. “Cat­e­gories are more gen­er­al than tags.”

    Wrong. Cat­e­gories are ex­act­ly the same as tags; the only dif­fer­ence is that the same el­e­ment may have mul­ti­ple tags, but the same el­e­ment can­not be in sev­er­al cat­e­gories at a time.

  3. “Cat­e­gories are used to group con­tent, where­as tags are used to quick­ly find the con­tent lat­er.”

    Wrong. Both are used to group con­tent. If I tag some of my pho­tos my cat, it means that I have a group of pho­tos with my cat on them.

    Both are used to quick­ly find the con­tent lat­er. When I put a file in G:\De­ve­lop­ment\<Pro­ject name> and an­oth­er file in E:\Misc\Funny pic­tures, it's ex­act­ly for the pur­pose of find­ing the con­tent eas­i­er lat­er.

  4. “Tags are used to in­di­cate some­thing about the tagged el­e­ment; cat­e­gories show that the el­e­ment be­longs to some­thing.”

    Wrong, or at least dif­fi­cult to un­der­stand. When I tag a pho­to as Ni­a­gara Falls, it means ex­act­ly the same thing as if I were putting this pho­to in G:\Pho­tos\Nia­ga­ra Falls\ di­rec­to­ry, i.e. that the pho­to has Ni­a­gara Falls on it, and that it be­longs to the set of pho­tos of Ni­a­gara Falls.

  5. “Cat­e­gories are mu­tu­al­ly ex­clu­sive. Tags are not.”


Con­clu­sion: So, what do we get? We get that users don't re­al­ly un­der­stand the dif­fer­ence be­tween cat­e­gories and tags. The worst of all is that they are ex­pect­ing dif­fer­ences which don't ex­ist, and that the dif­fer­ences they imag­ine may turn them against the us­age of the tags.

How does it ap­ply to search and fil­ter­ing?

Search and fil­ter­ing

There are ba­si­cal­ly three ways to search for con­tent:

  1. Cat­e­go­ry-based search­ing.

    This is the ba­sic form of search where the user re­mains pas­sive. Tree-ori­ent­ed struc­tures are the most di­rect il­lus­tra­tion of cat­e­go­ry-based search­ing. When I want to buy a new Xeon E5-2620 CPU on my fa­vorite web­site, I go to:

    Hardware › Components › CPU › Socket 2011
  2. Meta-based and/or as­sist­ed search­ing.

    This search still as­sists the user, but en­ables the user to be more ac­tive. For ex­am­ple, when I want to buy the AF-S NIKKOR 70-200mm f/2.8G ED VR II lens, I may fil­ter the list of every lens by spec­i­fy­ing that I want to see only the lens­es pro­duced by Nikon, which are from 70 to 200mm and have vi­bra­tion re­duc­tion.

  3. Free text search­ing.

    This search is the most per­mis­sive, since the user writes what­ev­er she want. This is also the most pow­er­ful one when the user knows ex­act­ly what she's search­ing for and when search ac­tu­al­ly works (most of the time, it doesn't).

    For ex­am­ple, if I want to read the specs of Ford Fu­sion Hy­brid SE, by typ­ing Fu­sion Hy­brid SE on a web­site which pub­lish­es ve­hi­cle spec­i­fi­ca­tions, I ex­pect to see ex­act­ly the page cor­re­spond­ing to this mod­el.

Many web­sites and ap­pli­ca­tions al­low sev­er­al of those three search mod­els. Of­ten, tree-form cat­e­gories-based con­tent can also be found through a tex­tu­al search, or there are tags and tex­tu­al search at the same time, or cat­e­go­rized con­tent can ad­di­tion­al­ly be fil­tered with meta-based fil­ters, etc.

This is done be­cause of the sim­ple ob­ser­va­tion: peo­ple are us­ing the type of search they need in a spe­cif­ic cir­cum­stance. Tak­ing again the black sleep­ing cat ex­am­ple:

The place of tags in search and fil­ter­ing

Tags are weird, since they re­place cat­e­gories, but also be­long to the sec­ond type of search: the meta-based one.

For the sake of sim­plic­i­ty, we can as­sume that there might be a dif­fer­ence be­tween tags and meta­da­ta:¹ meta­da­ta would be pre­sent­ed more as a pure­ly fil­ter­ing tech­nique, where­as tags would be pri­ma­ry used for search.

For ex­am­ple, the size of the pho­to would be pure meta, used to fil­ter pho­tos to show only the large ones.

In this case, tags would pre­sent them­selves as a search el­e­ment which is used when the user wants to re­main pas­sive. Near­ly iden­ti­cal to cat­e­gories, es­pe­cial­ly when placed in a form of a tree, tags would still re­main dif­fer­ent from cat­e­gories be­cause of their non-ex­clu­siv­i­ty.

When cat­e­gories are re­placed by tags, users may not re­al­ly un­der­stand that they should in­clude mul­ti­ple tags in or­der to fo­cus their search to what they re­al­ly need. This is ex­act­ly as the is­sue of some peo­ple when it comes to us­ing tex­tu­al search. For ex­am­ple when search­ing for the tick­ets price they need to pay to go on a trip to Switzer­land, they may start by try­ing to type “trip” alone, or “Switzer­land” alone.

Con­clu­sion: cat­e­gories, de­spite be­ing ter­ri­ble as a way to or­ga­nize in­for­ma­tion, would be more in­tu­itive for be­gin­ners in or­der to use as­sist­ed, pas­sive tree-based search. On the oth­er hand, a well-im­ple­ment­ed tag­ging sys­tem should help the user to un­der­stand both:

Am­bigu­ous ter­mi­nol­o­gy

  1. While tags them­selves are used more and more, the term tag may not be un­der­stand clear­ly. GMail, for ex­am­ple, uses la­bels which are ex­act­ly the same as tags.

  2. The word tag has also a dif­fer­ent mean­ing, clos­er to iden­ti­ty, like in an­i­mal tag­ging.

  3. Fi­nal­ly, in web com­mu­ni­ties where tags are as­signed by mod­er­a­tors, tags may be per­ceived more like an ap­proval (ex­am­ple: tag­ging a mes­sage to ap­pear on a home page) or a dis­ap­proval (ex­am­ple: tag­ging a post as off-top­ic).

Those three points make it more dif­fi­cult for be­gin­ners to un­der­stand what tags are in an un­am­bigu­ous way. When there is a risk of mis­un­der­stand­ing, de­sign­ers should:

1 Even if the as­sump­tion that there is a dif­fer­ence be­tween pure meta­da­ta and tags makes things sim­pler, this as­sump­tion is high­ly ques­tion­able. For ex­am­ple a date would be pure meta­da­ta, but still, it may be used as a first-class search el­e­ment, like when some­body search­es for pho­tos of 9/11 at­tacks. In the same way, tags are of­ten used to ac­tu­al­ly fil­ter the con­tent, in­stead of search­ing.