How to Cultivate Analytics Capabilities within an Organization? Design Options of Shared Service Centers for Analytics

Editor's Notes

• #2: Own experiences: BPS / CAO
• #6: By reviewing the major companies operating in Germany that currently have vacancies in the area of data scientists, a long list of 22 companies was created. After a first contact, only 15 companies turned out to run a shared service center approach, while the rest is looking for data scientists in a specific area only, and, therefore, is not relevant for our study. Based on availability and willingness for an interview, our final selection comprised nine globally leading companies from manufacturing, chemicals and IT services. All have established an ACC in Germany in recent years and have already completed at least a number of projects. Within this sample, we conducted 12 interviews, performed between May and August 2016. The interviewees have had different roles within the ACCs, such as executives, consultants, data scientists or have been an internal client of the ACC. A majority has already held at least two of those positions during their career. With three exceptions, the expert interviews happened face-to-face with the advantage of getting more information because of the private atmosphere (Mayring 2002). The interviews are analyzed based on inductive category building, a qualitative content analysis method, according to Mayring (2002).This method aims to build generalizable categories without referring to other theoretical concepts by using a systematic, rule-and-theory-based procedure. The major process steps are defining abstraction levels and formulating macro operators for reduction like skipping, generalizing, constructing, integrating, selecting, and bundling paraphrases. First, in order to structure the analysis, themes are determined. The themes correspond
• #7: All ACCs focus on supporting some or all of the objectives (cf. Figure 2) and adopt a series of corresponding functions to fulfill them (Table 1): The ACC should advocate analytics and identify potential for data-driven business models to lead the transformation towards a data-driven company. A data driven company is described as an organization that heavily relies on data to make decisions and take actions. To provide analytics expertise and act as a central contact, ACCs achieve economies of scale by centralizing the analytics skills across the organization and connecting ACC experts, business units and external service providers. The ACC identifies and creates uses cases that are supposed to optimize business process, products and services. It further evaluates the feasibility and value contribution of such use cases. Developing a data strategy and governance that lays out how to gain data access in terms of integration and migration is also one of the major responsibilities of an ACC. Some compa nies, therefore, decided to develop and maintain a data lake - a single data repository that tackles the challenges of data silos and monolithic systems (Woods 2011; Stein & Morrison 2014). The data governance is accompanied by the function of platform management with the main purpose of enterprise architecture integration. The ACC is further responsible for knowledge management and the definition of standards with the intention to benefit from best practices. Because the ACC has deep knowledge and capabilities in advanced analytics, it does not just support the end-user deployment; it also provides cross-organizational trainings to drive broad adoption of analytics within the organization. These aim to increase awareness of the services available from the ACC, to teach employees of BUs how to identify issues that data analytics can be useful in addressing, and to judge the amount of effort and resources required in different scenarios. The IT governance for analytics objective focuses on the appropriate management of licenses, vendors and external service providers as well as risk and change management that are required for analytics solutions. Objective Functions
• #9: While it is very common for business units to pay for services received by shared service centers (Schmidt, 1997), it does have certain shortcomings in the specific case of analytics competency centers. Allocating the cost of ACC services solely to the BU means that the number of use cases presented to the ACC is limited by the BU’s budget. Further, the BU considers the ACC a service provider rather than another unit of the organization on eye level. This leads to less collaboration and may hamper creativity in the creation of new use cases, particularly in cross-organizational projects. Lastly, from the perspective of the BU, paying for the service might be cheaper than developing its own analytics competencies by hiring data scientists, but it is still difficult to afford the service for smaller BUs with less resources. In comparison, pro-active cross-organizational projects are more likely to happen if the ACC is covering the cost. In this model, the ACC is acting independently and is willing to acquire and aggregate more data in an effort to get a bigger picture, benefitting the company as a whole instead of focusing solely on the issue of a single BU. Furthermore, there are no delays due to budget approval processes on the BU side, and the ACC can act with higher speed and agility. Acting on its own budget, the ACC experiences more freedom in selecting projects and making decisions. However, covering the whole project costs may lead to a massive amount of use cases presented by the BUs. In this situation, selecting the projects often does not follow a rigorous decision process, but instead selection is tainted by political power. These circumstances potentially lead to less pressure to deliver great results and a more lenient work attitude. To mitigate the pros and cons, organizations in which the ACC is solely responsible for the project costs have switched to a hybrid cost distribution model or plan to do so in the long run, effectively splitting the costs between BU and ACC. In a hybrid financing model, the ACC has a dedicated budget to finance certain defined parts of the project (e.g. use case creation), it has its own innovation budget for proactive, cross-organizational projects, and has a budget to cover short-term cash flow issues of the BU to start the PoC or implementation phase (e.g. start process of data integration while BU is still waiting for budget approval).
• #10: Plotting these characteristics in a simple visual representation allows for further visual analysis. Figure 5 captures all nine sampled ACCs along five characteristics out of thirteen1 that are relevant for the strategic direction of the ACC: Focus refers to the primary “customer” of the ACC. In two cases the ACC exclusively serves internal business units, in one further case the ACC also interacts directly with customers of the organization, thus offering analytics-as-a-service. In all other cases, the ACC serves internal stakeholders, but also focuses on product and service improvements that potentially impact the customer. Cost distribution defines the unit financing the biggest share of analytics projects as described in section 4.2. In hybrid-cost models, the lion’s share of funding still comes from either the ACC or the particular BU involved. Agility refers to the flexibility in a project; low agility means the project has a preset of inflexible restrictions in terms of cost, time and scope whereas a high agility means more flexible restrictions. Responsibility describes which unit is taking the lead in executing the projects: Either the business unit acts as the owner of the project or the responsibility is shared in a joint project. The last characteristic refers to culture: While some ACCs have a very hierarchical approach in structuring and managing their ACCs, others choose a very flat organization with high peerto- peer communication. Analyzing the plot of the key characteristics (Figure 5), two distinct types of ACCs can easily be identified: A number of ACCs screened exhibit characteristics on the left side of the visualization bar - the represent one type of ACC that we call efficiency-driven ACCs. Another type of ACC, in which the characteristics tend towards the right side of the visualization bar, we call innovation-driven ACCs. Based on these two types of ACCs, additional insights can be generated as the type of an ACC correlates with other information gathered. This section describes the two types of ACCs, and the additional correlations and their implications in more detail. Efficiency-Driven Innovation-Driven Internal Mostly Business Unit Low Business Unit Hierarchical coordination Internal & External Mostly ACC High Shared Peer-to-peer Focus