Executive Summary

As social tagging grows increasingly popular on the Web, organizations are curious to see how this trendy Web 2.0 approach can benefit the business world. Social tagging allows users to employ their own language to organize and retrieve content, and encourages social collaboration between peers by making those tags visible to others. Organizations are thus looking to social tagging as a potential solution for increased findability on intranets, news/blog monitoring and collaboration in workgroups. The enterprise context is different however, and many of the elements that make social tagging work on the Web make it a challenge behind the firewall.

These include a smaller and more defined corpus of content, a much smaller user population with less available time, and stricter information retrieval requirements. Other challenges include the quality of tags (misspellings, compound words, personal tags) and of tag search engines (no spell-check, no stemming), both issues which impact the performance of social tagging. These problems aside, social tagging is still a valuable tool for the enterprise, as long as these challenges are considered. A hybrid solution that employs social tagging and formal taxonomy as complementary information access approaches is emerging as a winning solution.

Introduction

You can tag just about anything these days: vacation photos, products, blog posts, friends on Facebook… But what about that travel approval form you can never find on the intranet? Or the latest company annual report? While the popularity and application of social tagging has been on a continuous climb on the Web since the launch of Del.ici.ous in 2003, the enterprise has been slower to adopt the trend. In the past few years, vendors – from niche companies like Connectbeam to large providers like IBM Lotus – have launched social tagging products aimed at the enterprise, hoping to capitalize on the growing curiosity around how this Web 2.0 approach can benefit the business world. But just what is social tagging in the enterprise: a cheap solution to information organization in difficult economic times, giving people the tools they love at home to use at work, or just a faddish attempt to incorporate “cool web 2.0 stuff” into the enterprise context?

 Social Tagging on the Web – Popularity Explained

Before we can determine whether social tagging works in the enterprise, we have to understand why it works so well on the web. Social tagging originally emerged as a solution for offering individual users control over findability. It allowed individuals to use their own language – tags with personal meaning – to organize and retrieve content important to them. No need to sift through unwieldy directory structures or guess at a good search term. When made visible to others, the value of tags expanded from the individual to the group: you could not only re-find your content but also explore content tagged similarly by others. So tagging moved from being a personal use tool to enabling serendipitous discovery of content, ideas, and peers.

The ability to support a multiplicity of views is an important aspect of this approach. Whereas taxonomies and thesauri force convergence around a preferred term (e.g. use Cinema, not Movies), folksonomies (“structures” that emerge from social tags) allow users to choose whatever term they find meaningful and there is no funnelling into a single point of view. Essentially, people tag the same content differently, and not only is this ok, but it’s the whole point: people who tag content as “cinema” probably think differently than those who tag using the term “movies”1. This allows users to get around the “problem” of semantics, that is, having to figure out what terms others have chosen.

Why Introduce Social tagging in the Enterprise?

So why does the enterprise want to bring social tagging behind the firewall? There are a couple of simple answers to this: it’s a quick, cheap and easy way to enable lightweight social collaboration and augment findability.  Social tagging software solutions are typically inexpensive niche products or modules you can add on to your existing software suite. They don’t represent a big investment, and are relatively simple to implement. Additionally, the technology is usually relatively easy to use (compared to an ERP or CMS, for example), so there is no need to spend heavily on training or special staff.

 

Fig 1: Sample tool view: Dogear from Lotus Connections

Fig 1: Sample tool view: Dogear from Lotus Connections

http://www-01.ibm.com/software/lotus/images/Dogear_screenshot.jpg

Organizations are interested in using social tagging technology both within workgroups and across the enterprise. Tagging can supplement information retrieval options in intranets and document management systems, allowing employees to use tags to enhance the findability of internal and external content without waiting for an information professional to categorize it. Many tools allow you to subscribe to what is called a “tag stream” and monitor content being tagged. This is an excellent way to provide trend monitoring, news/blog aggregation, and other external company-related information. Social tagging can also be used to help share documents, research, and more, both within formal workgroups and informal communities of practice. This increases not only collaboration, but also expertise location, as viewing a tagger’s profile can tell you a lot about their interests and expertise. Essentially, social tagging creates a richer set of options for users within the enterprise for locating content and colleagues.

The Difference Between the Web and the Enterprise

These all sound like great benefits, so why doesn’t every company implement social tagging? The problem with transposing approaches that work great on the web into a corporate setting is that we don’t often take into account the contextual differences that affect success. The first major difference between the Web and the enterprise is nature of the content. The Web is a seemingly infinite collection with no clear edges, no authority, and no structure. Outsourcing the organization of information on the Web to users makes sense in this context: people can be free to make up their own definitions and categories, and structure will emerge from the chaos simply through volume.  Corporate content is different: it is a more defined corpus of information that is meant to support specific tasks and users, entities are structured and there is authority to be respected. Finding information is mission-critical in this context, so employees have a higher need for precise and reliable access to information.

People are also different on the Web vs. the enterprise. One of the big success factors in many Web 2.0 approaches is population size. A Forrester study showed that 16-18% of users between 18-40 have tagged Web content. 16-18% is a lot when you consider the millions and millions of people who surf the Web, but not a lot in the context of a 30 person work team or a 500 employee company. Recent case studies published from MITRE and BUPA indicate that the level of participation in the enterprise tends to be more around 10% of users. People at work also have less time and motivation to participate in social software: they are focused on deliverables and deadlines and do not often have the spare time or incentive to focus on sharing and tagging information. They also have more concerns about privacy and security, given that their tags and tagging profile may be made visible to other employees.

Other issues that must be considered pertain to the quality of tags and tagging systems. Unfortunately, people are not especially good at tagging: they tag inconsistently over time and are usually more concerned about personal findability than the “greater good” (which is arguably the original point of social tagging). This translates itself into tags of dubious quality: misspelled tags (e.g. Sharepiont), inconsistent tags (e.g. dog vs. dogs), compound words (e.g. SocialTagging), personal use tags (e.g. toread), etc.2 For example, the top 75 tags from LibraryThing.com (book tagging) shows such issues.

Fig. 2: Top 75 tags from LibraryThing.com

Fig. 2: Top 75 tags from LibraryThing.com

This is compounded by the fact that tag search engines are not yet very sophisticated. Most do not have spell-check or stemming (expanding a search for “ski” to include “skis”, “skiing”, etc.), nor do they have the ability to search for synonyms (e.g. Ursa Major vs. Big Dipper). Some tagging sites, such as LibraryThing, are experimenting with an approach called “tag equivalency”, where a user can make two tags synonymous, but this is not widespread.

These issues lead to a problem of precision and recall in tag-based searching. Precision is the ability of a search engine to return items from the collection that are truly relevant to your search (exactness), while recall is the ability to return all the relevant results in a collection (completeness). In the world of the Web, precision and recall tend to be less critical, as you are dealing with collections with millions of items and less of a : does it really matter if you don’t find every single picture of cats on Flickr? It does matter however if you don’t find all the relevant cases in your law firm’s case history database.

The Content Continuum

This is not to say that social tagging is an irreparably flawed approach that should be avoided in the enterprise. Quite the contrary: as stated earlier, there are many situations and contexts in which social tagging is not only interesting but practical. What it boils down to is the nature of the content being tagged and whether its consumers can afford the drawbacks of social tagging. Some enterprise content is more mission-critical than others, such as policies and approved methods, and organizations should invest the time and effort required to ensure that this content is findable. Other content, such as blogs, external links and discussion postings can afford to be more serendipitous.

Fig. 3: The content continuum

Fig. 3: The content continuum

The Hybrid Approach

However, the best approach to enterprise findability is a savvy combination of both approaches. That is, don’t fire your libraries and throw out your corporate taxonomy just yet: keep them for tagging your high value content and ensuring that you have consistency of categorization and terminology. But do supplement your intranet search and other less formal sources with social tagging, show tags alongside official categories and result sets.  Finally, social tags can be a great source of terms to augment your corporate taxonomy, as they tend to represent the most current and natural user terminology. Some tools are also exploring taxonomy-directed tagging, where type-ahead tries to match user-generated tags to a controlled vocabulary (e.g. ZigTag).

Essentially, organizations should not be afraid to get involved in social tagging: it is a simple and inexpensive way to inject a little Web 2.0 into your employees’ lives and supplement more formal approaches to findability. Given that the technology is relatively mature, there are also some good case studies available to guide your effort3. To get you started, here are some basic do’s and don’ts around social tagging implementation and governance.

Don’t assume social tagging will replace other corporate metadata. View it as a way to enrich existing content organization and add life to your taxonomy.
Do investigate which tool is right for you: most enterprise software suites have add-ons, there are new specialized products, or you can custom build on based on open source tools.
Don’t skimp on marketing the use and benefits of social tagging to target users.
Do allow users to import their tags from online sources (e.g. Del.ici.ous) if possible, this will prepopulate the system and make the tool seem valuable faster.
Do evaluate whether your corporate culture can support open tagger profiles or would be better suited to anonymous tagging. Either can work, but an open system results in more social connections.
Do consider some minimal effort around tag clean-up and vetting to weed out inappropriate tags and basic misspellings/inconsistencies.

As long as you are aware of how enterprise tagging differs from its web counterpart, you will be ready to hit the ground running and make your tagging project a success.

1. See “Ontology is Overrated” by Clay Shirky for a more in depth discussion: http://www.shirky.com/writings/ontology_overrated.html
2. See “Folksonomies: Tidying up Tags” by Guy & Tonkin (2006) for more on tag quality: http://www.dlib.org/dlib/january06/guy/01guy.html
3.   MITRE case study: presented at Taxonomy Community of Practice Series, September 2007, http://www.earley.com/_September_2007.asp.
BUPA case study: http://www.socialtext.net/cases2/index.cgi?bupa_social_bookmarking