Developing a Strategy for Integrating Big Data Analytics into the Enterprise

by   |   January 7, 2013 1:15 pm   |   0 Comments

Editor’s note: This article is the fourth in a series examining issues related to evaluating and implementing big data analytics in business.

As with any innovative technology that promises business value, there is a rush to embrace big data analytics as a supplementary source, if not the main source of analytics for the enterprise. And as with the adoption of any new technology, it is important to consider the challenges and issues that become apparent when new technologies are rapidly brought into production.

Clearly, it would be unwise to commit to a new technology without assessing its potential value for improving existing processes or creating new opportunities. This value is manifested in terms of increased profitability, reduced organizational risk, or enhanced customer experience (among other benefits) in relation to the costs associated with the technology’s introduction and continued maintenance and operations. Essentially, testing and piloting technology is necessary to maintain an enterprise’s competitiveness and ensure the new technology is feasible for implementation. But in many organizations, the processes to expeditiously mainstream new techniques and tools often bypass existing program governance and corporate best practices designed to ensure new technologies work with existing systems.

The result is that pilot projects that are prematurely moved into “production” are really just point solutions relying on islands of data that don’t scale from the performance perspective nor fit into the enterprise from an architectural perspective.

Deciding What, How and When Big Data Technologies Are Right for You
The adoption of big data technology is reminiscent of other technology adoption cycles such as customer relationship management (CRM) systems or the desire to use XML for a broad spectrum of data-oriented activities. These are cycles in which potentially disruptive methods and algorithms insinuate themselves in ways that might not be completely aligned with the corporate strategy or the corporate culture because the organization is not prepared to make best use of the technology. Yet enterprises need to allow experimentation to test-drive new technologies in ways that conform to proper program management and due diligence.

Other articles in this series

Market and business drivers for big data analytics.

Read more»

Business problems suited to big data analytics.

Read more»

Achieving Organizational Alignment for big data analytics.

Read more»

For example, implementing a CRM system will not benefit the company until the users of the system are satisfied with the quality of the customer data and are properly trained to make best use of customer data to improve customer service and increase sales. The implementation of the technology must be coupled with a strategy to employ that technology for business benefit.

A strategic plan for big data adoption within the business intelligence environment will balance the need for agility in adopting innovative big data analytics methods and data management architectures in ways that are aligned with corporate vision and governance. That strategy will incorporate aspects of exploration of viability and feasibility of new techniques, selecting those techniques that best benefit the organization, as well as provide support for moving those techniques into the production environment. The strategy would incorporate these key points:

Standardize practices for soliciting business user input. Frequently the enthusiasm of the IT department for a new technology overwhelms the processes for establishing grounded business justifications for adopting it. Project plans focus on the delivery of the capability in ways that neglect solving specific business problems. In turn, as components of the technology are delivered and milestones are reached, there is a realization that the product does not address the end-user expectations. This realization triggers a redesign (in the best case) or abandonment (in the worst case).

Whether the big data activity is driven by the business users or by technologists, it is critical to engage business users early on to gauge their expectations and establish their success criteria. Clearly defining the performance expectation for big data analytics means linking business value to business utilization of the technology. Ask the business users: What do they want to achieve using the new techniques? How do those expectations support the company’s overall corporate vision? How is the execution aligned with the strategic business objectives? Directly interact with the business function leaders as partners. Enlist their active participation as part of the requirements gathering stage, and welcome their input and suggestions during the design, testing, and implementation.

Clarify Go/No-Go criteria. Allocating time, resources, and budget on testing out new technologies is a valid use of research and development spending. However, at some point a decision must be made to either embrace the technology that is being tested and move it into production, or to recognize that it may not meet the business’s needs.

Before embarking on any design and development activity, work with the business users and their specific corporate value metrics to provide at least five quantitative performance measures that will reflect the success of the technology. State a specific expected improvement associated with a dimension of value, assert a level of acceptable performance that must be achieved, and provide an explicit time frame within which the level of acceptable performance is to be reached.

For example, if the big data application is intended to monitor customer sentiment, an example performance measure might be “decrease customer call center issue resolution time by 15 percent within three weeks after an acute issue is identified.” Another example might be “increase cross-sell volume by 20 percent as a result of improved recommendations within 10 days after each analysis.” These discrete quantitative measures can be used to make that go/no-go decision as to whether to move forward or to pull the plug.

Failing to establish a structure around this process can lead to problems. In many cases, making this decision on incomplete metrics or irrelevant measures leads to committing to the methods even when it may not make sense, killing a project before it has been determined to add value, or worse, deferring the decision, effectively continuing to commit resources without having an actionable game plan for moving the technology into production.

Prepare the data environment for massive scalability. Big data volumes may threaten to overwhelm an organization’s existing infrastructure for data acquisition for big data analytics, especially if the technical architecture is organized around a traditional data warehouse information flow. Test-driving big data techniques can be done in a virtual sandbox environment, which can be iteratively configured and reconfigured to suit the needs of the technology. However, the expectation that any big data tools and technologies would be mainstreamed also implies a need for applying corresponding operations, maintenance, security, and business continuity standards that are relevant for any system.

Applying this standard means the enterprise needs to adjust its operations and systems maintenance model to presume the existence of massive data volumes. Considerations to address include using high speed networks; enabling high performance data integration tools such as data replication, change data capture, compression, and alternate data layouts to rapidly load data and access data; and enabling large-scale back-up systems.

Promote data reuse. Big data analytics holds the promise of creating value through the collection, integration, and analysis of many large, disparate datasets. Different analyses will employ a variety of data sources, implying the potential need to use the same datasets multiple times in different ways. Data reuse has specific ramifications to the environment and implies that the data management architecture must support capture and access to these datasets in a consistent and predictable manner.

To meet this need, the enterprise should to provide capabilities including increased data archiving, massive-scale data standardization to ensure level of trust in the data, and address other implications such as enforcing a level of precise data synchronizations across multiple sources or using data virtualization tools to smooth both semantics and latency in data access. Your organization’s information and business intelligence strategy must detail plans for large-scale data management accessibility and quality as part of the enterprise information architecture.

Institute proper levels of oversight and governance. A common challenge associated with adoption of any new technology is walking the fine line between speculative application development, assessing pilot projects as successful, and transitioning those successful pilots into the mainstream. This cannot be done without some oversight and governance. Governance procedures will direct the alignment of speculative development work with business requirements designed to achieve the enterprise’s goals.

Incorporating oversight of innovative activities within a well-defined strategic program plan prevents the evolution of “shadow IT” practices that bypass the standard approval processes. Consider the CRM example: many business people may try to deploy a CRM solution by signing up for cloud-based tools and populating those tools with data extracted from the corporate database. However, data that is migrated to that hosted system can no longer be monitored or controlled according to corporate data policies and guidelines.

A different approach is to fully incorporate innovation projects within the corporate governance framework to provide the senior management with full visibility into the potential value of new technologies. This allows management to track the execution of proofs of concept, monitor the progress of pilot projects, and help determine the degree to which the technology is suited to solving specific business problems. At the same time, the governance policies can include the specification of those go/no-go success criteria, as well as the methods for assessing those criteria prior to adopting the technology and incorporating it into the organizational technology architecture.

Provide a governed process for mainstreaming technology. Any speculative development using new technology is vetted using the concept of a “pilot project” or a “proof of concept.” But once the pilot is completed and the concept has been proven, then what? If the project demonstrates clear value based on the success criteria and a decision is made to adopt the technology, you must have a plan to move the system into a production design and development project. A technology adoption plan will specify details including: identifying the need for additional resources, assembling a development and testing environment, assessing the level of effort for design, development, and implementation, staffing requirements, and providing a program plan for running and managing the technology over time.

This process is essential for big data technologies. Big data has some particular needs, such as scalable computing platforms for developing high performance algorithms, acquiring the computing as well as the increased storage to accommodate the massive volumes of data. Don’t neglect upgrading the information architecture to support the analysis of many massive data sets, including semantic metadata, “meaning-based” and “context-based” analysis, high performance data integration, and scalable data quality management.

Reflecting back on our sample performance and success measures, presume that a pilot big data recommendation analytics engine did demonstrate the targeted improvement for cross-selling as well as the other specified success metrics. Meeting these objectives means that the proof of concept was successful, and an objective evaluation would lead to the recommendation that the application be devised for migration into production via the approved system development lifecycle processes and channels for design, development, testing, and implementation.

This approach to developing a strategy for integrating big data analytics is one in which the business users are directly engaged in providing input to articulate the value proposition, get alignment across business functions, and help prioritize activities. The approach means that any successful pilot project can be mainstreamed within an implementation plan that will guide the current development while reducing the complexity of enhancements and extensibility in future development. Doing so will reduce the complexity and pain that is often associated with “shoehorning” application code into the enterprise architecture.

Rather, taking these steps will confirm that the design and development are fully integrated within the organization’s project governance, information technology governance, and information governance frameworks.  This will ensure continued alignment with business objectives while observing the IT governance protocols for moving applications into production. This will also ensure that the results of the new applications can be fully integrated within the existing business intelligence, reporting, and big data analytics infrastructure and provide maximum value to a broad constituency across the organization.

David Loshin is the author of several books, including Practitioner’s Guide to Data Quality Improvement and the second edition of Business Intelligence—The Savvy Manager’s Guide. As president of Knowledge Integrity Inc., he consults with organizations in the areas of data governance, data quality, master data management and business intelligence. Email him at

Home page photo by JoeJohnson2 via Wikipedia.

Tags: , ,

Post a Comment

Your email is never published nor shared. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>