Page 146 -

Knowledge Capture and Codifi cation 129

to purchase the fairly expensive automated software tools available. In all cases,
however, a hybrid approach is best. While automated systems can help provide a
good head start, especially in cases where there is a signifi cant volume of existing
legacy content, human intervention is almost always needed to correct and refi ne
the classifi cation — and, of course, to ensure consensus. A number of manual tax-
onomy techniques can be used to help groups work together to create the categories,
decide on the facets, and develop a thesaurus. The most popular techniques
used are card sorting (Nielsen 1994 2009 ) and affi nity diagramming ( Farnum 2002 ;
Gaffney 2000 ).
Card sorting is a very low-tech method of understanding users ’ mental models of
how knowledge should be organized. The best tools to use are sticky-note cards
preprinted with key concepts already known (typically derived from a survey of
documents and of intranet content). There should be some blank cards so users can
add terms. There are two general types of card sorting: open and closed. In open card
sorting, there are no preestablished groupings, whereas in closed card sorting, there
is already a preliminary taxonomy in place. Open card sorting is useful to better
understand participants ’ perceptions, while closed card sorting is useful to validate an
existing taxonomy (e.g., document classifi cation scheme or web navigation design).
The general steps involved are to distribute the cards to each participant and ask
them to group together those cards in a way that makes sense to them and to name
each grouping. The piles can be of different sizes and users can elect not to use some
of the cards (as long as they jot down why they were rejected). The user groups should
be representative, and they can be homogenous (if we are looking at a consensus) and
heterogeneous (in order to have a taxonomy that is broader in scope and to create a
thesaurus). Both types of groups are recommended if time permits. The recommended
number of participants is a minimum of six and the recommended time is a minimum
of thirty minutes to sort fi fty cards.
Users can stop when they feel they have exhausted all the possibilities. The facilita-
tor may ask them to try to aggregate into bigger groups if there are too many groups
(a good rule of thumb is Miller ’ s magic number of seven plus or minus two, which
appears to be the number of items our cognitive abilities are best able to handle). Once
everyone has fi nished, the facilitator enters everyone ’ s results onto a spreadsheet.
There will be some agreement right at the outset about groupings, while others
will differ. A statistical analysis called cluster analysis can be used to obtain a visual
representation of the results. For those groupings that were different, it may be due
to using different labels to denote the same concept, or additional subcategories may
be required. When the resulting preliminary taxonomy has been completed, the same