One of the more frequent questions I get centers around the puzzle which presents itself when you have user generated content and need to categorize it. Do you start with a full fledged category structure or do you start lightweight? And when do you create new categories?

Take this example: A startup built a site which lets users upload instructions on how to “make things” (in the lingo of the maker movement). How should they go about structuring this content which optimizes engagement, highlights the breadth of content and provides enough structure to find stuff but not so much that you end up in a click-maze?

eBay provides an interesting case study here: The inventory on the eBay site is structured into categories and sub-categories. eBay has no control over what people will list on the site. You want to avoid categories which are so stuffed that you don’t find anything anymore and categories which are so empty that you only see a few listings in that particular category.

Here’s how eBay did it: Start out with a basic structure of categories you believe will make sense. Make this initial list fairly small and compact (10 is probably a good number). Add one category for “Other”. Now let your users go crazy. Watch what they upload into each category and watch for trends which emerge in the “Other” category. Once you find enough listings of a particular kind you break them into their own category (at eBay we had a rule that we need at least two to three full screen pages of listings to create a new category for them — as you want to avoid empty or near-empty categories).

That’s precisely how eBay came to become one of the largest used car sellers in the world. eBay didn’t anticipate that anyone would ever sell their car on the site. But people started listing a few vehicles in the “Other” and “Antiques” category. Once this trickle turned into enough of a river eBay created a “Cars” category. And when the river turned into a flood stream eBay created it’s own sub-site eBay Motors. The rest is history as they say.

Build What Matters.
Pascal ツ