Data Science Change Management

Introduction

Change management makes or breaks large organizations. Organizations that have consolidated businesses and functions over the years are used to slow rates of change, and many as a result are out of step with the rapid decision making and change that’s required to survive in the competitive business environment of the post-internet era industry. In this post we argue that data-science-enabled-change-management can become the backbone of an organization’s strategic change management process, and we also discuss the flip side of the problem, which is to get managers to embrace data science at scale in their teams and processes, and to enable the implementation of these approaches. We first discuss the data science process that converts data to decisions, and the integration of insights for incidental purposes, and also the deliberate integration of such data driven insights and results into technical, functional and managerial workflows. Subsequently, we evaluate the change management frameworks that the established companies and institutions use, and concoct ideas around these change paradigms that data can enable. We conclude with a set of recommendations for organizations implementing large scale change, and the data instrumentation, analytics and data science that can improve the success rates of change initiatives.

The Evolving Data Science Process and its Industrial Impact

The data science process that enables the conversion of data to decisions has undergone a sea change in the last few decades. While statistical engineering approaches gained ground in industrial decision making in post war industrial era, operations research and industrial engineering management furthered the field in the multipolar decades after, and we foresee distributed data processing and the internet of things as enablers of the next revolution. With the advancements in statistical methods that computers enabled, we were able to develop large scale mathematical modeling approaches, both statistical and numerical, that contributed to the data science process. In recent years, data science has come to mean the development of mathematical and statistical models using massive data sets, and the deployment of novel algorithms. In the foreseeable future, the data science process will take on higher level abstractions of data analysis as being the human interaction level. Naturally, this has the potential to make business management and decision making more and more data driven. While there is a lot of hype about the process and the method, the essential benefit of data science to any organization remains informed decision making. For this reason, even though the data science process will continue to evolve over time, there is a need for managers and leaders in organizations to understand the challenges of the organization, know the relevance of data science methodologies and results, and to be able to convert statistically significant results into decisions in the practical and business domains.

Integrating Data Insightsinto Decisions

The integration of insights from data analysis into decision making has a long and turbid history in large and small organizations alike. The trend in major businesses these days is the tendency to move away from charismatic know-it-all leaders who are either functional experts or technical experts that have matured into people leaders, and away from democratic please-all leaders that are effective change agents but leave the details of the change to their functionally informed peers. The trend is towards novelty and change agents that can take difficult decisions, which may not find favor with their teams unless the leaders in question are armed with data and analysis enough to explain their decisions. No organization in the past has grown without its leaders being party to hard decisions, where decision making is complex and requires knowledge and situation-specific information. In the age of data, this problem can be rewritten as an information asymmetry equation, where there is not enough data or analysis that aids a specific decision.

Data Instrumentation and Change Management

To offset such hard decision making situations, data instrumentation is firstly key. Data instrumentation, which is the identification of key operational definitions for the business’ performance, the development of their data collection, storage and display plans, and the transformation of these measures into business-ready insights, is one way of enabling managers with dashboards.

Rather than just a steady stream of information which puts small and large decisions alike at the behest of the humans receiving them, there is a chance to automate the lower level decisions, when data instrumentation is done well. Sufficiently well-defined operational definitions also translate very well to higher level metrics, which can be managed by humans. Industry 4.0[1], which is the next wave of automated factories and businesses, requires high productivity from humans as well as machines, and decision support systems that are enabled by data are intended to play a key role in such situations. The transition to Industry 4.0, for a lot of companies, is a huge change management problem. The problem is more likely to be solved in the high technology manufacturing space, than in businesses with more traditional business models.

Data instrumentation can be integrated with established data driven models that provide multiple views of the organization. Examples of strategic management frameworks with which can inform the data instrumentation strategy of organizations are the balanced scorecard, which is widely used by company executives, and the Hoshin planning approach, which is widely used by large hierarchical organizations.

Traditional and New Change Management Models

Traditional change management models were driven by two kinds of organizational behaviors – commitment and compliance. The commitment model of change relied on the fealty of employees and their sense of belonging to the wider company’s journey. It relied on charismatic leadership, where the data and facts played second fiddle to the change that was to be initiated and deployed. The compliance model of change management relied on the development of systems, and the adoption of systems management paradigms. Here, the change management approaches relied on the prestige that surrounded global standards of performance, be it in the financial bottom line that’s core to the business, or what were once peripheral concerns, such as supplier and customer well-being, or environmental responsibilities. In the last several decades, corporate quality has become synonymous with change management in many organizations, and large change initiatives start as systems improvement exercises, driven by approaches such as Lean or Six Sigma. Some organizations follow more organic, people-centric approaches, such as Kotter’s 8 stage process for change management [2]. Other organizations have home-grown approaches. One essential need for all of these is the communication of some urgency to change, and to support this communication with data-backed reasoning.

Fig. 1: John Kotter’s 8 step change management process

Future Change Management Models

Organizations of the future will need to respond to rapidly acquired and rapidly changing consumer bases. They will need to respond to technological challenges as product obsolescence rates are influenced by consolidation of function and the increased generality of form and interaction in products. Companies will not only have to take decisions fast, but will have to automate the small stuff, and be very well informed, from their data and analytics practices, to take large decisions with confidence. The change management approaches of such future organizations are therefore more algorithmic and situational and less dependent on the past experience of leaders. A key element of this is instrumentation.

Learning and Reinforcement Plans for Data Science

Effective data science is at the core of the data science value proposition for business management and change management. One weak link in the data-driven change management is therefore the data science culture in organizations itself. Data science is a growing and rapidly changing body of knowledge, and companies that enable themselves to compete in their industries at large are undoubtedly those companies that leverage their teams’ data science capabilities through a culture of constant learning and the reinforcement of key ideas. Learning in data science happens on many levels, two of which are key – one of these is technical or functional learning, which is the learning that enables execution, while the other is systemic learning, which enables ideation and creativity. The need for the former is diminishing, because advanced frameworks and developments in AI are allowing automated analysis on a scale as never seen before, but the latter, systemic learning, taps into human ingenuity and creativity at its most fundamental level. This makes systemic and conceptual learning, and a reinforcement of the pattern of evolution of data science frameworks important for managers to understand.

Concluding Remarks

From a change management standpoint, organizations can benefit from diverse strategic management and change management methodologies that are synthesized with the data science constructs that have come to define the big data and machine learning revolutions. Executives can benefit from the learning that enables company executives to solve the big change management problems that the organization has to face, and such change initiatives often add the most value or are pivotal for the company’s success. Conceptual, higher level learning that will permeate the data science landscape enables executives to think through problems from a broad based perspective, while abstracting away the details to constantly improving and evolving data science frameworks. Change management frameworks themselves will evolve to face and embrace the new capabilities that new markets, new business models and new approaches demand. While not all of these are directly connected with the data science or machine learning revolutions, they will enable executives to face the key challenges, by leveraging data instrumentation and processing on a large scale.

Leave a Comment