Skip to main content

What Is an Information Silo?

If you are in the field of information management, you have undoubtedly heard the phrase “information silo” or perhaps “data silo”. I grew up in rural Michigan, so I understood the metaphor immediately when I first heard it. Wikipedia offers up a very concise definition of an information silo:

An information silo, or a group of such silos, is an insular management system in which one information system or subsystem is incapable of reciprocal operation with others that are, or should be, related. Thus information is not adequately shared but rather remains sequestered within each system or subsystem, figuratively trapped within a container like grain is trapped within a silo: there may be much of it, and it may be stacked quite high and freely available within those limits, but it has no effect outside those limits (Wikipedia).

Information silos aren’t simply an information system and architecture problem; they are also a knowledge management problem. In other words, what contributes to the making of information silos are the three pillars of the information landscape we have so commonly seen illustrated in presentations on the subject: people, process, and technology.

We Are All Silos

I’d love to tell you that “We Are All Silos” is an experimental rock band. However, I’m pointing a finger at all information workers out there: we are all both the architects of silos and the silos themselves. The comparison may sound strange, but hear me out.

When information workers build systems to create, collect, and retrieve information relevant to our work, we often have mostly our own work in mind. Too frequently, we don’t take in the scope of the entire organizational information landscape. Sometimes, we don’t think about it. Most times, it’s because the scope is too large. All we can do is focus on the task at hand. The result is a dedicated system built to do a particular type of work. Really, this makes sense. For as much as SharePoint tries, not all platforms can be all things to all people. Content management systems manage and store content. Product information systems manage and store product information. They are fit-for-purpose systems.

At the end of the day, we deliberately silo information in separate systems because they are best suited to manage that type of information. In so doing, however, we are separating the information and metadata from other systems. We may or may not recognize that what is valuable data for one group in one system is also valuable for another group working in another system. The result is working toward integration between systems delivering only the data and content necessary for particular processes or applications.

Communities of Practice

People also silo information. In knowledge management, communities of practice (CoPs) form around areas of interest and find among their membership subject matter experts and novices interested in the topic. These CoPs further silo information into segmented repositories of collected content pertinent to their discussions. They may even have navigational and metadata structures highly specific to their area.

I once worked in a role in which we moved away from small, specific taxonomies built for each CoP in favor of a more robust enterprise taxonomy. In many ways, this was the right way to go because the common needs of the CoPs was far greater than they thought. Everyone believed they were special and, hence, every concept was special. People argued for concepts in the taxonomy like “marketing meeting minutes” and “financial meeting minutes” as if there were something fundamentally different about “meeting minutes”. At the time, I countered that “meeting minutes” are just that. Adding another standalone concept such as “marketing” or “finance” got the content exactly to the same level of specificity without pre-coordinating every possible combination of a general content format type and who created it.

I was right in that there was no sense in managing hundreds of separate CoP taxonomies with their highly specific concepts. Where I was wrong was that in the creation of a single, faceted enterprise taxonomy, people were inundated by thousands of concepts completely irrelevant to their jobs. The missing component at the time was the technical ability to filter and deliver a subset of the greater enterprise taxonomy to every group. Users were overwhelmed by choice and struggled to find the limited number of concepts they needed among the thousands of options.

The people silo is even more granular. Knowledge management processes struggle with the tacit knowledge problem. If we are information silos in a group, what kind of information silo are we as an individual with a wealth of experience and knowledge relevant to our work? Individuals also contribute to process and technology silos by making information unavailable on their personal laptops and file server locations.

Bridge the Silos

I’ve heard people speak about bridging silos, but, since we are all farmers with our own silos, the metaphor doesn’t make sense. What good would a bridge do between two silos filled with grain…or information? Silos are purpose-built to store and separate grains. We should not try to bridge the silos, but find some other connection between them.

A metaphor I’ve used in the past is the rhizome, which is a decentralized collection of nodes connected by relationships. Rhizomes grow underground and then send up shoots above the ground. Maybe we should build tunnels between the silos? Well, metaphorically, we can.

The real-world rhizome sounds exactly like an electronic semantic graph. In a graph, there are nodes (subjects and objects) connected by predicates (a relationship expressed as a verb). The practical application of connecting data where it lives is often called a data fabric or data mesh.

From a taxonomy, ontology, and graph database perspective, we can connect all types of content and data using one or more centralized taxonomies. These can be bound by the rules of an ontology, across siloed systems into a knowledge graph. While the target may differ in each approach, the general idea is the same. Connect existing data and content where it lives through a superimposed layer of functionality.

We don’t all need to be content farmers filling up and maintaining our separate information silos. We can grow our content. This might be a good thing, but storing all that content for later use in isolated, disconnected silos doesn’t do much good for the organization. As individuals, we can share our knowledge for use in existing processes in systems which are connected and bound by a common semantic layer powered by taxonomies.