Background
Knowledge management (KM) and knowledge transfer (KT) are the processes by which an organisation formally captures and records the way it operates, defines key assets such as data and models, and establishes internal standards. Typically, this information is held at an atomic level within departments and even by individuals. Ensuring it is not dependent on specific practitioners or process owners allows an organisation to sequence the DNA of its business, ensuring it can be rebuilt, licensed, sold or developed in a sustainable and replicable manner.
Data and analytics professionals will probably be most familiar with the field of KM through data dictionaries and meta-data. These are a form of knowledge management since they centralise critical information so that it an be accessed independently by multiple users and generations of users. As organisations become ever more data-driven and pioneer advanced analytics in the form of artificial intelligence and machine learning, ensuring such fundamental building blocks of understanding are in place will become more important.
Organisations already recognise the value which resides within human intelligence to some extent – anywhere that operates a travel policy under which senior executives are not allowed to take the same flight is reflecting the value of their knowledge and abilities. As part of the automation of business processes, however, that dependence needs to be reduced.
A good example of how dependence on human skills can lead to increased risk and exposure to failures can be found in legacy systems. In a study carried out by Forrester Research for Compuware, 57% of enterprises said they currently run over half of their business-critical applications on mainframes and expect this to rise to 64% by 2019. Yet at the same time, these same enterprises report having lost 23% of their skilled mainframe workers in the last five years, with 63% unable to fill those vacancies. As a consequence, the ability to maintain and develop those essential applications in that environment is constrained because of a lack of human knowledge.
This CARBON™ guidance considers frameworks for the implementation and measurement of knowledge management. It focuses on both soft and hard methodologies that can help to ensure business-critical knowledge stays within the organisation and not just within the minds of its workers.
Measuring what you know
All organisations have a vision. But to what extent do they consider how much they need to learn and improve in order to achieve that vision? It might not appear the most obvious area on which to focus, until you consider the extent to which intellectual property and intangible assets are at the heart of a company’s value. If the vision is underpinned by the use of data and analytics, then the organisation needs to have a proper internal “memory” for how these were developed, how they operate and what knowledge has been built in but which is not immediately visible or interpretable. For the individuals and for the organisation itself, there needs to be a process of continuous learning which is supported by knowledge management and knowledge transfer.
Management gurus David Kaplan and Robert Norton recognised the importance of this approach when they developed the balanced scorecard between 1989 and 1991, making learning one of the four quadrants, alongside financial, customer and internal measures (see Figure 1).
Figure 1 – Four quadrants of the Balance Scorecard (Source: APO)
In doing so, however, they also recognised that the majority of focus for an organisation will typically be on financial and customer perspectives, suggesting weightings for these of 60% and 20% each. Learning was meant to have a 5 – 10% weighting. In reality, the continuous disruption of internal processes through digitisation and datafication has meant this dimension has taken virtually all of the remaining attention, leaving knowledge management neglected or even ignored in the majority of firms. Among the membership of DataIQ Leaders, only one organisation – a global mobile telco – had a formal approach to knowledge transfer and an individual dedicated to this process. This whitepaper draws on the expertise and experience of that practitioner in many places.
Within the DataIQ Leaders CARBON™ assessment, there is similarly a limited degree of focus, yet a fundamental one. As part of the Organisation battery of questions, respondents are asked, “To what extent do other business functions understand the capability of the overall data function and how it can benefit the wider business?” This reflects a view of the importance of internal communications and disseminating knowledge about what the D&A function does and is able to deliver.
Typically, organisations score at Level 3 (Defined), with some lines of business having an understanding, but no enterprise-wide knowledge sharing. Improving knowledge transfer involves promoting a better understanding everywhere in the enterprise of the capabilities and their benefits, although this does not always involve formal business partnering at Level 4 (Managed). Where there is a strong degree of understanding, promotion and partnering, an organisation will have reached Level 5 (Optimised).
Introducing KM and KT processes
Commercial organisations are typically behind the curve for academic and government bodies when it comes to knowledge management and knowledge transfer, except for those with a very strong and specific focus on R&D. If investment is being made into developments that may result in new patents, strong KM will be in place, for example.
Where funding is being made by research bodies, this practice is built-in from the base – Innovate UK has the Knowledge Transfer Network to ensure innovators and funders are able to connect and learn effectively, for example. Cambridge University has the Institute for Manufacturing which formalises the spin-out of academic research into new businesses or new process for companies.
Dr Tim Minshall at the Centre for Technology Management in the IfM, offers a good perspective on this practice:
“Three key factors seem to underpin successful KT. First, it’s not a ‘zero cost’ activity – it takes effort and time to make it work. Second, it is a ‘contact sport’ – it works best when people meet to exchange ideas, sometimes serendipitiously, and spot new opportunities. Third, it needs practical, timely and active support at an institutional level – within companies and universities – encouraging a culture of open access and open innovation.”
As a checklist for how to build-out knowledge management from the D&A function across the enterprise, the following serves as a useful starting point:
- Document processes from a “know-nothing” baseline – To become the first officer on a commercial airplane involves at least 1,500 hours of flying time during training. Yet even when a licence holder, pilots follow a written checklist before every flight. They even walk around the aircraft to check its condition visually for themselves, despite the teams of engineers who will have checked the airframe. Consider this the baseline for knowledge transfer in and out of an expert function. In documenting a process within data and analytics, assume nothing is already known and needs to be included so that even a non-expert could follow it. Just as a non-pilot could run through the same checklist by following the book, even though it would take them much longer because of a lack of familiarity with the controls and their location, your knowledge documents should be capable of step-by-step execution.
- Develop “human back-ups” – Business resilience and disaster recovery mandate mirror sites and data centre redundancy to ensure critical processes are continuously available. Yet some of the most fundamental knowledge about data – such as definitions, sources, flows, etc – and analytics – such as the assumptions built into models or which techniques were applied – live in the memory of a single individual. Ensuring that there is always one other practitioner in possession of this basic knowledge is a foundation of effective KM.
- Maintain training for white and black-out operations – Ongoing training and validation of that training is at the heart of continuous learning within an organisation. The focus is usually on core technical skills sets as well as essential process, such as data governance. But it is also worth considering a training exercise for downside scenarios, such as the absence of a key worker when a project is on deadline which could lead to an operational black-out. This is likely to reveal the true depth of knowledge and formalised understanding.
- Introduce knowledge management tools – Finding the data required for an analytics process is usually the most time-consuming part of the job. But it can be just as frustrating for a new hire to try to understand how a key process operates, especially following a sudden departure or as a result of a merger or acquisition. Internal communications systems can begin to capture some of this knowledge by at least bringing informal conversations into a discoverable channel. Next generation tools based around artificial intelligence are starting to emerge which can automatically read from operational systems and map out the underlying business processes (including where exceptions are taking place). Data science should put itself at the forefront by building or deploying this type of technology against its own operating processes.
- Enable the internal, contract the external – Internal communications both within the D&A function and between it and lines of business are vital for ongoing, dynamic knowledge transfer. Formal presentations, meet-ups and networking events, intranets and physical communications should all form a part of this. Bringing knowledge in from external contractors is just as important, but often more difficult or overlooked. Historically, third parties have argued that the work they do is their intellectual property. Yet in most commercial contexts where the use of the client’s data is involved and models are deployed by the client’s systems, a contractual obligation should be included to document and transfer ownership of this knowledge from the start. The risk from relying on an “exo-memory” is even higher than failing to capture what internal employees know since there is less of a legal basis on which to force any transfer unless stated in the contract.
Knowledge transfer in a global mobile telco
The European-headquartered mobile network has 265,000 employees and 265 million customers. It operates across Europe, South Africa and South America under different brands. Each country has a high degree of autonomy with separate technology stacks.
The business has a research and development culture that understands the role of knowledge sharing and, uniquely among DataIQ Leaders members, has a specific function and individual dedicated to this practice. This is not about picking up predictive models from one territory and dropping them into other countries because the cultures and conditions are all different. It is about working at a process level to identify what is repeatable and will help each country to operate more efficiently. That might mean something that reduces the number of tools required to complete a task from five to two, for example.
Typically, analysts within each territory are developing innovative solutions, but are unaware of what has already been developed elsewhere. To address this, knowledge management has focused on the customer and data, rather than technology. To date, it has been through four stages in this evolution, beginning with retail systems, then CRM, moving onto OTT services and is now building out an API-based platform for enterprise-level services. These are being built in the leading six countries first to ensure the biggest upfront gains.
Knowledge transfer is pursued within a four-stage process:
1 – Process development: identifying transportable processes that can be offered to each territory;
2 – Technology development: introducing new processes into existing technology stacks;
3 – Standardising technology: gradually reducing the number of different tools to create more consistency;
4 – Migrating to a single data model: creating a global standard adopted by every territory.
Stage 4 has seen this model developed and applied to 2,800 common items of data which all countries are required to capture, manage and make available to the global group. This has allowed local adoption to be part of a global progression seamlessly.
In order to gain buy-in, countries are not placed under any obligation to adopt new solutions and can continue to work with their existing processes or technologies, or buy from external vendors if they prefer. There is a natural human resistance to experimentation and innovation, with local countries not looking ahead to the environment in which they will be operating in five, ten or even 20 years’ time. The task of knowledge management is to put that roadmap in place and help each country along the journey.
Creating visibility for these standardised solutions and innovations is a critical part of the role, not least because there is a decay curve for new processes – adoption needs to happen within a window of opportunity that is typically six months long. The single data model has helped by removing a common obstacle to adoption.
One unexpected obstacle arose around physical interaction between the knowledge manager and target territories. Building relationships, active listening and stakeholder management are all more effective with face-to-face meetings. However, following a change in the C-suite, a ban was placed on international travel with the CEO’s approval needed for any costs over €100. This forced international teams to adopt video-conferencing, but even for a network with dedicated, high-bandwidth broadband, the absence of a booking system meant there was a constant struggle to schedule online meetings. Eventually, pushback from local countries led to the travel ban being reversed.
Now, data and analytics teams work on new projects with the idea of transportability at their heart. That means keeping in mind different technology stacks and local cultures. That means ensuring teams have a mixed skills base which includes data engineers as well as data scientists so end-to-end processes can be built, moved and dropped.
Valuing knowledge
Making a commitment to knowledge management and introducing knowledge transfer processes requires investment in time, resources and money. Given the historically low level of focus on this aspect of business performance, KM has struggled to gain the support it requires.
However, there are significant signs of change as a direct consequence of the rise of data and analytics. In enterprises that have fully adopted these, it is now recognised that there is a need to formalise, control, develop and exploit valuable assets being developed by the D&A practice. This is creating three areas around which a KM practice for D&A can win support:
- Intellectual property – new models, algorithms, robotic process automation and potentially patents based in data all create value in a business. For investors and shareholders, knowing this is being given the proper attention is increasingly important.
- Best practice – before D&A is centralised, it often operates in pockets around the organisation. Identifying where best practice is already in place, quantifying and codifying it, then transposing it into other functions reduces the cycle time for performance improvements.
- “Knowhow” – people are an important asset to any business and need careful management. However, the understanding they have about business-critical processes, from data definitions and data models to the rules within algorithms can not be allowed to reside solely in their memories. Instead, documentation and communication of this vital knowledge needs to form part of their working practices.
​For a detailed understanding of knowledge management and the processes involved, DataIQ Leaders recommends “Critical Knowledge Transfer: Tools for Managing Your Company’s Deep Smarts” by Dorothy Leonard-Barton, Walter C. Swap and Gavin Barton, available to read online at Google Books here.
