News

Bad
enterprise search (especially in combination with bad information
architecture) is the no. 1 frustration for intranet users in most organisations. Even
though there is quite a number of things you can do to make search
better (see for example IBF’s research report “Improving
Search”, available for IBF members only), in most organisations
there still is ample room for optimisation.

As
bad search is (in most cases) not a technical issue, it seems only
natural that social enhancements taken from the Web 2.0 space should
provide chances for improving search experience and results.

Let’s
take a quick look at some such options:

Utilising
user generated context

If
users can add context to content by tagging, setting bookmarks,
rating contents or voting on search results, all these additional
meta-data can be used in optimising the relevance ranking of a search
inquiry.

Pros:

• if
  people interact with content (e.g. bookmark it), this content
  normally is of importance to them (or of no importance, if for
  example a very low rating is given). Thus e.g. keywords associated
  with a content by social tagging typically express a higher
  relevance than keywords that have to be provided by the author of a
  content.

• user
  contributions relevant to a search term can also be used to provide
  ‘people results’ to a search thus not only delivering relevant
  content but also people potentially related to the topic searched
  for.

Cons:

• without
  a sufficient number of contributions across the whole spectrum of
  content available, there’s a high risk of falsification of the
  search results (e.g. content that is important for a topic is
  displayed far down in the search results just because it has no
  ratings or tags yet)

• relevance
  for a search term can be very different for different audiences,
  thus search results can become (falsely) biased to audiences that
  are more active in contributing that others

• user
  tags can provide ‘synonyms’ that might not be available in the
  content (and meta-data) itself. Thus more potentially relevant
  results for a topic can be found (which of course is a ‘pro’).
  Technically speaking the ‘recall rate’ of the search is improved.
  Higher recall (more results) automatically means lower precision
  (relevance of result ranking), though. For more, see: recall
  and precision

Search
Lists / Topic searches

If
user A (or team A) has researched a topic intensively and put
together a list of the most relevant sources for this topic, then
user B can be provided with an excellent starting point if looking for
information on the same topic by accessing this ‘search
list’.
Examples for this kind of search support can be found on
the web at Rollyo or Swicki.

Pros:

• searches
  are restricted (or focussed) to sources of potentially high
  relevance, thus eliminating large numbers of probable irrelevant
  search results

• utilising
  synergies by ‘reusing’ efforts already put into researching topics

• identifying
  experts for topics

Cons:

• the
  more dynamic information on a certain topic is, the bigger the risk
  that the sources included in the list don’t fully cover everything
  that might be of relevance to it (e.g. new blogs on the web come up
  by the minute thus making it impossible to provide comprehensive
  coverage in static lists).
  Thus potentially relevant results
  might be missing if the search lists don’t contain everything
  related to the topic.

• also
  see relevance for different audience above

Social
Best Bets

Manually
associating the ‘best’ result for a certain search term is a powerful
way of improving relevancy for the most sought after topics. This is
typically done by a central team that analyses the most frequently
used searches. With ‘social best bets’ everybody can contribute his
favourite search results in regard to a specific search inquiry.

For
example Google has integrated this concept as ‘Wiki
KeyMatch‘ into the Google Search Appliance.

Pros:

• while
  centrally maintained best bets are usually limited to no more than
  around the 100 most frequent or important search terms (due to the
  time and resources required for maintenance), social best bets can
  cover much greater number of search terms and a wider spectrum (e.g.
  for very specific themes).

• see
  also pros above

Cons:

• what’s
  important for user A doesn’t necessarily have to be of relevance to
  user B. If social best bets are displayed on top of the result list,
  users will get quite frustrated with this feature if it repeatedly
  delivers ‘bad’ suggestions.
  This issue can (partially) be
  addressed by enabling users to rate on the helpfulness  of best bets
  and adjusting its relevance accordingly.

• see
  also cons above

Web
2.0-style designed search results pages (SERPs)

Instead
of displaying a single search results list, you can think about
designing the SERP in a web 2.0 portal fashion. You might consider
displaying elements like:

• separate
  ‘boxes’ for different search sources (e.g. displaying search results
  from internal blogs in one box, those from regular intranet
  content in a separate box, results from wikis in yet another etc.)

• displaying
  alternate ways to navigate the content returned e.g. with tagclouds
  (e.g. like Quintura does),
  providing lists of what search results other user have clicked on
  for the same search or results that other users have voted as most
  relevant to them (e.g. like at Xibben).

• highlight
  picture and video results by displaying thumbnails (where
  appropriate)

• adding human touch by returning people with photos that are related to a
  search (e.g. that frequently use a tag that was search for by the
  user)

• …

Pros:

• information
  can be presented in a more encompassing and differentiated way thus
  potentially serving the needs and ways of working of more audiences

• provide
  a design that is more pleasurable to use and stimulates more
  interactiveness

Cons:

• if
  not well designed such pages can quickly become cluttered and
  confusing

• users
  are accustomed to google-style SERPs and might be reluctant to get
  used to anything that looks different

While
none of the options outlined is a magic pill to solving all problems
with internal search, it sure is a good idea to think about how they
could benefit your organisation and what existing issues and user
requirements they might address.

Also,
there are more ‘social’ ways for improving enterprise search than
just the ones sketched above. I’d love to hear from any other ideas
that you have already experimented with.