Bridging the gap between centralized and decentralized data analysis

Profile picture for user Brad Stillwell By Brad Stillwell July 17, 2019
Summary:
It's now possible to eliminate the traditional divide between centralized business intelligence and decentralized data analysis, writes Birst's Brad Stillwell

Hand reaching up to touch mesh network on blue background © metamorworks - shutterstock

Companies today often realize they need to become more data-driven to keep pace with market change, remain competitive and optimize profit opportunities. But, choosing the right Business Intelligence (BI) and analytics solution is a challenge. In addition to legacy solutions, the organization can choose between centralized and decentralized platforms, the latter often called desktop data discovery systems. Neither of these approaches, alone, provides for the scalability, trust and self-service that is necessary to make analytics pervasive throughout the organization. Organizations need one highly flexible solution, which can solve the needs of two types of users.

Diverse data needs

Traditionally, companies took a centralized approach, focused on creating a shared single source of truth across an organization. The analysis could be driven by the IT team, or a centralized team of data experts could be put in place as a shared service. Either way, the goal has been to publish and distribute that centralized knowledge to a broad number of people, many of which lack highly specialized data analysis skills.

The centralized approach focuses on providing trusted and governed enterprise data and making it accessible to a large base of users. The use cases tend to be about large, strategic issues and the software focuses on making analytics consumable, repeatable and scalable. These systems may take more time to set up, but they offer valuable economies of scale and consistency. Because this approach can generate a large volume of queries and reports, the process is highly efficient and trusted. There isn't a question of reliability that can accompany one-off reports.

By contrast, with a decentralized solution, the use cases are typically focused on the needs of specific departments or individuals, and they examine a contained data set. The issue at hand can be completely ad hoc, with no need to worry about scale or repeatability. This focus can make it as easy to go from raw data to discovery or insight, quickly and easily. Decentralized systems are all about minimizing the time it takes to turn data into visualizations that can drive quick business decisions.

For example, the C-suite may be concentrating on growth trajectories and examining predictive patterns, needing a highly reliable centralized approach. At the same time, an operational team can be focused on day-to-day opportunities to cut waste and optimize the workforce. There may also be times when analysts, business users, or local admins need to make well-informed decisions quickly to seize a competitive opportunity. In these cases, speed may be most important. It makes sense to empower users, so they can leverage the centralized models — but provide them with more freedom. User interfaces can help guide these users, who may not have expertise in data analysis.

One solution, not two

The dichotomy of needs can drive some organizations to deploy more than one tool. This can cause disruption, erode decision-making confidence, and stall progress while teams attempt to sort out the disparities between solutions.

Veterans in business analytics know that one version of organization-wide data is essential. It is very hard to formulate data insights that are beyond reproach when multiple, poorly-integrated solutions are put in place, each generating slightly altered mirrors of the underlying data points.

Adopting one highly flexible solution, which can fulfill multiple needs is a more effective approach. Recent innovation in the industry has made it possible to build a networked approach to data analysis, bridging the needs of the forward-looking visionaries and the real-time number crunchers.

This software technology uses the same data points, whether it is performing advanced data science projections for the IT team or finding rudimentary connections between spending trends and budget impact for the line of business managers. For both types of applications, the software uses the same tools for data preparation, visualization and the sharing of insights. The architecture is based on the principles of multi-tenant cloud computing and the virtualization of analytic services.

Networks that work

Now, new technology is taking the concept of self-service and combining it with ML and AI to improve ease-of-use and collaboration for every employee, not just advanced business analysts. Making the process easier frees up time so that business users — no matter their role in the company — can focus on higher-value tasks. It empowers business users to generate personalized insights themselves by simply selecting a KPI of interest. Experienced technical users also appreciate the time-saving interface, as it gives them more fine-grain control. User management, security, auditing, data orchestration, and data modeling and preparation, can all be accomplished through a single, modern interface that is highly intuitive and collaborative.

Knowledge is shared, and productivity is improved, without corrupting production-level data models. The architecture, based on the principles of multi-tenant cloud computing, uses a virtualization approach of services deployment, empowering multiple networked bridges that can share data, but also protect the data from being corrupted or losing its context.

The result is that organizations can seamlessly connect a common and reusable enterprise semantic layer with analyst-built, self-service data prep workflows to create a networked and unified shared version of the truth across the organization. While it may sound complex, the sophisticated processes are all behind a user interface that is easy to use. This means the front-line business user, needing to dive into factors influencing budget over-runs, can tap into the benefits of artificial intelligence (AI) and machine learning (ML)-driven analytics — often called 'augmented analytics' — without needing special training and programming skills.

Case study: Carlisle Fluid Technologies

Carlisle Fluid Technologies, a global manufacturer of equipment for the supply and application of paints, coatings and sprayed materials, is an example of an organization using this networked approach to integrate and serve centralized and decentralized analytic teams. Not only has the company saved money, but now, its internal teams can collaborate and collectively deliver trusted KPIs.

Before deploying these networked BI capabilities, the company tried to manage fourteen different ERP and reporting instances distributed globally, making one consistent definition of revenue a challenge to calculate, especially because of multiple currencies. The move to a networked BI system across regions has provided one version of revenue, plus given business units the freedom and agility to do their own analysis and better track information across finance, sales and the supply chain. The company also reports productivity gains from using a modern interface to onboard new developers and manage more complex workflows for data integration and loading, as well as granular user/role management and auditing.

Who will benefit the most?

Many types of organizations in multiple industries will find this ability helpful. Here are some examples:

  • Start-up ventures and small to mid-sized businesses (SMBs) that are starting to build their customer base and claim market share will find the advanced data insights valuable. As young sales teams, hungry for market growth are tempted to lower prices and over-extend offers, decentralized data insights will help them keep focused on account value, while centralized analysis will help the whole team learn the long-term impact of actions.
  • Organizations in highly competitive situations, where they must explore every opportunity to remove roadblocks, eliminate waste and enhance market opportunities to keep pace with the market will find the data insights helpful to fine-tuning pricing and market strategies - without sacrificing profits.
  • Enterprises exploring digital concepts, such as servitization, selling direct to consumers, ecommerce or bypassing traditional go-to-market channels will also benefit from the ability to examine profit influencers and determine drivers in account profitability. Teams throughout the organizations will be making shifts in operational processes and needing data to confirm positive results.
  • Companies planning or executing new structures, such as mergers, acquisitions, partnerships, co-ventures, or initial public offerings (IPOs) will benefit from the reporting that examines data in multiple views, from the perspectives of several key business units, as well as the overarching centralized view. The level of attention is often needed to get new ventures off the ground with confidence and sound investment of limited capital.

Getting started

While the IT team and the C-suite may be well-versed in advanced analytics, other users throughout the organization may need training and encouragement to adapt a data-centric mind-set, looking at their own personal Key Performance Indicators (KPIs), setting team goals and measuring achievements, and looking for ways to hone in on anomalies and spot early warning signs of issues before they escalate to budget-impacting ones. With the right tools, individuals and teams across the organization will start to find key business insights that enhance their decision making.

Some education, incentives and rewards may be needed to get this new, data-centric mindset to take hold. But, the result will be an organization that is in-tune to the data that reflects the core truth, past to future, end to end. Download a webcast on the topic here to learn more about smart business analytics.