Knowledge transfer barriers in the age of analytics
Analytic functions are becoming part of day to day operations for many organizations, such as predicting what quantity of a supply to order, or the likelihood of an insurance claim being fraud. But these functions are not implemented and left to run autonomously. Upkeep and changes will always be required. Changes requires knowledge of the function. As an organization evolves over time, the knowledge transfer about an analytic function will be critical, but potentially problematic.
To look at the challenges of knowledge transfer, the three categories of organizational knowledge from the article, The problems of embeddedness: knowledge transfer, coordination and reuse in information systems, will be used.
Knowledge-as-object – The tangible objects of an analytic function such as the data, queries, algorithms and outputs like reports. This category is the what of the analytic function. The primary knowledge transfer issue here is access – finding the knowledge objects and the means by which they are produced. The continuing democratization of analytic technology means the function can be highly decentralized – pieces in many places.
Knowledge-as-cognition – This is the sharing of knowledge, usually through social networks. These networks can be direct person to person, or through some other medium. The knowledge transfer barriers here include different interpretations of the same knowledge object. A continuous challenge in analytics is how the same data element can mean different things to different people. Knowledge coordination is another challenge. Over time, varying levels of exposure to the analytic function can create gaps or black boxes.
Knowledge-as-capability – The application of the analytic function to create value for an organization. Capability is generated from the experts’ work activities. With the application comes results and feedback which leads to fine tuning or further innovation of the analytics function. Knowledge transfer in this category is achieved by actually doing the work alongside an expert. It cannot be transferred like an object or shared through a medium like documentation. The hundreds of decisions, assumptions and nuances of an analytic function is learned here by doing.
The leading causes of barriers to knowledge transfer are the traditional ones of time and money. It takes time to:
- Create the required knowledge base for an analytic function.
- Implement and maintain processes to prevent fragmentation of knowledge.
- For people to learn and develop expertise.
- Experiment and innovate.
To do the above also costs. Additional capacity is needed beyond just executing the analytic function. To keep costs low, this additional capacity often does not exist. In turn, this creates risk for an organization of knowledge loss and the breakdown of the analytic function. This is particularly problematic if an organization is single threaded – relying on a sole expert. The sudden loss of that expert can set back an analytic function to the beginning, even one that is relatively mature.
Because analytics is enabled by technology, there is often a misconception that skills in the technology alone is sufficient to learn an existing function and traditional IT type documentation can provide knowledge. As anyone who has taken over a program written by another, even well documented, knows, there is a steep learning curve.
When developing and implementing an analytic function, consideration must be made of knowledge transfer, how it can be achieved, and what are the associated risks if it is not.