Global, local and enterprise search

Written by , published October 30th, 2012

Categorised under: Enterprise 2.0, Intranets, Search tools


This is a sketch that I drew while sitting in one of the enterprise search presentations at the recent KMWorld conference in Washington, DC. It explores the question: what should enterprise search be searching?

The starting point is the global/local model for intranets, that we’ve described earlier. This distinguishes between two types of information:

  • Global information, that is common across the whole organisation, and used by all staff.
  • Local information, which is specific to a particular group of staff, according to geography, business unit and/or job role.

With this model as a foundation, it then becomes easy to outline the default scope of enterprise search:

By default, enterprise search should return results from all global information, plus the portion of local content that is relevant to me.

Easy to say, but much harder to do! Like all simple models, it’s deceptive how much thinking and work is required to deliver it in the real world.

This is not just a matter of installing a fancy search engine, and hoping that it’s smart enough to work all this out for itself. Instead, the these building blocks must be in place:

  • There must be a clear framework for the overall intranet/collaboration landscape, that defines the purpose and audience for the many different spaces.
  • The identity of staff must be known, along with an understanding of what information is relevant to them.
  • Enterprise search must tailor results by default, based on searching just that portion of local space that is relevant to the current users.

This perhaps turns things on its head. Instead of hoping that enterprise search will make sense of a chaotic landscape, it suggests that delivering an effective search relies on bringing structure to information first.

What are your thoughts on this?

Tags: , , , , ,


  1. As you mention James “Easy to say, but much harder to do!”. Our first and only attempt was not what you would call a success.
    The result was way too many results from internal drives and from across the Internet. This was purely the result of poor implementation planning
    and scoping. Next time the search configuration will be carefully planned, implemented, monitored and adjusted (repeat last two each week for ever).

  2. Nice article, James. What you describe above is a good model for thinking about search and I especially like how to relate it to good information structures.

    I have a couple of comments and would like to know what you think.

    1. People sometimes get confused between the source of the information and the users of information. A “local” source may produce information that is of interest “globally”. A global source may produce information for “local” use. That’s one reason I prefer to completely eliminate the terms “global” and “local” although it’s hard not to use them because they are so engrained in how we all think. I’ve found with my clients and in workshops that the terms “specific” and “common” make it easier to talk about who produces what, and what goes where.

    2. I like very much to push people into thinking of “corporate” as a “specific” source much like the other entities in the organization. I would love to see your diagram modified to show “corporate” as a specific source rather than positioned next to “global”.

    3. You make a very good point when you say “Instead of hoping that enterprise search will make sense of a chaotic landscape, it suggests that delivering an effective search relies on bringing structure to information first.”
    You’ll be very happy to know that “information architecture, taxonomies, tagging” is in position 2, just behind “mobile”, in terms of investments planned for 2013/2014 according to the 360 organizations that participated in this year’s Digital Workplace Survey. Mobile is highest at 40 % and “information architecture, taxonomies, tagging “is at 35%, tying with “implementation of a social network platform”.

    So, maybe things are finally looking up from a findability perspective.

  3. You’re right in saying that installing a fancy search engine isn’t enough: it’s usually the old GIGO problem that bites. If you could rely on decent metadata or a reliable security model, then it would be easy to deliver more personalised, local content in results: the trouble is people aren’t good at reliable metadata and security models only appear consistent up until you try them (with your first search of ‘bosses salary’). So here’s our approach:
    1. Don’t pick a fancy search engine. Open source engines such as Apache Solr are equally as good.
    2. Derive as much structure during the crawling/indexing process as you can – for example a file path can be useful metadata if it tells you what group a file was created by
    3. Use other free tools for automatic entity extraction – this can be indexed for faceting for example.
    4. Autoclassification & taxonomies are useful but are often just a sticking plaster for a badly implemented search. There are plenty of overpriced solutions – we knocked up an open source version in ten days
    Peter’s comment on how to run a search project are absolutely spot on – you also need a cross-disciplinary search team and to consider Agile development methodologies.

One Trackback

  1. By Intranet Lounge on November 1, 2012 at 6:28 am

    Global, local and enterprise search – James Robertson,…

    This article has been submitted to Intranet Lounge – Trackback from Intranet Lounge…