[Theforum] Re: Sub-categories

[Madhu Menon]

(when I've been quiet for a while, you know something's brewing. :D This a 
fairly long message that you shouldn't skim. So put it aside until you have 
to time to read it all.)

At 10:12 AM 12/16/2001, Chris Spruck typed these words:
>My position on sub-categories is now completely unsure. :) I have the 
>O'Reilly IA book - I'll see what it says about this type of issue and see 
>if I can reach a conclusion with further IA research. The former scientist 
>in me was always a "splitter" instead of a "lumper" when it came to 
>classification.

I've written something on my weblog about categories and IA.
"Do users appreciate good IA on large sites?" - 
http://madman.weblogs.com/2001/11/22

(It was just a thought. Don't take it as an article on IA.)

Typically, classification ala Yahoo becomes necessary when you have far too 
much content to place in one category. Putting 300 articles into one 
category becomes unwieldy.

There are also problems with classifying everything in detail:

1) It can actually make something *harder* to find.

Yes, you read that right. The problem arises when people's own mental 
models don't match that of the Information Architect. For example, if you 
wanted to find articles on ColdFusion on evolt, would you go to "Backend" 
or "Code"?

Think about that from the view of a non-admin or new user of the site 
(actually, even I'm not sure). ColdFusion is clearly "code" (you write it, 
don't you?) and it *does* run at the "backend". If you look at it closely, 
it could even come under "Site development" AND "Software"!

Oh dear, what an array of choices!
So you go to "Backend" and don't find *any* articles on ColdFusion (wtf?). 
You then check out the vague "Code" category and find one article in the 
list, along with articles on CSS formatting, javascript window-targetting, 
and IE6 centering table layouts.

Is this the only article on ColdFusion on the site? Clicking to the next 
page of search results seems to convey that impression. But what the heck, 
I'm a persistent user, so I do a search to make sure:

http://www.evolt.org/search/index.html?keywords=ColdFusion

Sure enough, there *are* more CF articles, but some of them are in 
"backend", some in "code", and even one in "site development", but the 
categories don't match my mental model. As you have seen, the search has 
been my best shot at finding articles on CF.

Apply the same logic to taxonomies that are more than two levels deep and 
you could have a sticky mess where people resort to using the search 
instead of the categories themselves. If you have any subcategories under 
"Backend > ColdFusion", for example, you may be defeating the purpose of 
good classification. If the user isn't an advanced programmer, he or she 
may not know where to go next.

Let me illustrate that with another example in which I use ASP (with which 
I'm more comfortable).
First, visit www.asptoday.com and check out their categories (it's a 
dropdown in the top left).

Now, assume that evolt had the same subcategories under "Backend > ASP". So 
we have:

Backend > ASP >  .NET Framework
Backend > ASP >  Components
Backend > ASP >  Data Access
Backend > ASP >  DNA 2000
Backend > ASP >  E-commerce
Backend > ASP >  Performance
Backend > ASP >  Security/Admin
Backend > ASP >  Site Design
Backend > ASP >  Scripting
Backend > ASP >  XML/Data Transfer
Backend > ASP >  Other Technologies

Your user is Jack Brown, an ASP programmer who's been working on ASP for a 
year and has just been put on a team developing an e-commerce site for a 
medium-sized company. (I'm intentionally restricting the persona to minimum 
detail.)

Jack's not happy with the performance of his ASP pages. On analysing his 
code, he finds that his database queries are taking too much time to 
execute. He wants to optimise this. He wants to find some resources on the 
Web, and what better place to look than ASPToday, one of the best ASP sites 
out there. (Jack remembers someone on a mailing list mentioning something 
about "using native OLEDB drivers instead of ODBC for database access".)

He checks out the categories and gets confused. Let's see...

1) He wants to optimise his database queries so articles about that must be 
under "data access"
2) On the other hand, he's developing an e-commerce site, so could 
performance improvement be in *that* category?
3) Oh gosh, there's yet another category called "Performance". But does it 
include databases?

Jack has no idea what to do next. The detailed classification has him stumped.

Of course, the information architect would say that you should put an 
article into more than one category to prevent this problem. Indeed, that's 
what ASPToday has done. But our CMS won't permit that.

<aside>

Check out their site map at: 
<http://www.asptoday.com/information/information.asp?infoid=11&WROXEMPTOKEN=560325ZmO3gkfhwWK8wu3Ijivs>

They have another method of classification - by article type:

Article Types
   Cutting Edge
   In Depth
   Overview
   Tutorial

Interesting.

And their search is bloody interesting too. I'll leave you to explore it.

</aside>

In my weblog entry linked above, I've shown how it takes seven clicks to 
get to the Yahoo category for "Weblogs". How many people will do that? More 
importantly, how many people will do it *correctly*. Many will click on at 
least one wrong subcategory along the way, and they will then reach nowhere.


2) You need to have content for all categories.

If you create subcategories, you also need to have content to put into 
those categories. On a volunteer site with no editorial calendar and no 
schedule, this may be difficult. If you don't have content for a particular 
sub-category, you either have to leave it out of the site structure (ugh! 
bad idea) or put a "there are no articles in this category" message, which 
disappoints users.


I'm just putting forward some of the IA issues we have to consider. I'm not 
making recommendations yet, but my gut tells me that:
a) Any more than one subcategory will not work (i.e., stop at "Backend > 
ASP"), unless evolt gets acquired by someone like Internet.com tomorrow ;)
b) We need to identify articles by target audience i.e., 
Beginner/Intermediate/Advanced/Expert
c) We *may* need to add an "article type" like ASPToday has (see above)
d) We need to take another look at our existing categories.
e) If we change the structure of the site, moving articles to the 
appropriate category is going to be a hard, and interesting. :)

Incidentally, one of the best ways to understand user mental models is to 
analyse the search queries on the site. What are people searching for? 
Nothing else is quite as valuable given our wide user base. And if we were 
CNET, we'd even track which link people clicked in the search results.

Are the search queries logged somewhere? Dan? Jeff? Rudy? Is there some way 
the IAs (Javier and I so far. Any other volunteers?) could take a look at 
this?

Regards,

Madhu
PS: Rudy, didn't I tell you some of my emails were this long? ;)
PPS: Chris, I don't think the O'Reilly IA book has anything much on this topic.



<<<   *   >>>
Madhu Menon
User Experience Consultant
e-mail: webguru at vsnl.net