Technologies

The Emergence of the Analytics Architect

As analysis of big data matures, analytics architects bring a key set of skills for achieving business objectives

Big data analytics is gaining momentum as increasing numbers of companies realize they can leverage data and analytics to serve their client organizations and help improve their business performance. They recognize that deriving value from data is no longer a choice, but a business imperative in the emerging digital economy.1 And in many organizations, the analytics architect is turning out to be a key player in helping to extract insight from that data.

The evolution of using data to help deliver business insights and value has gone through a number of phases.2 The first phase focused on using data for reporting. The second phase began with the rise of big data. The commodity-driven MapReduce framework and its open source equivalent, Apache Hadoop, provided the blueprint for this phase. The third phase that is now emerging is characterized by significant innovations in technology, such as data discovery; interactive visualizations; and in-memory, predictive, and prescriptive analytics.

Businesses are leveraging these new technologies to implement bigger and more ambitious solutions than ever that analyze a wide variety of data, apply sophisticated models, and generate actionable insights. The insights can then be used to respond to events in real time. These new technologies, which are more complex than their previous, simpler, stand-alone counterparts, require additional planning, rigor, and coordination. That is, they require architecture—and an architect.

 

Architecture as a discipline for execution

The IT architecture discipline emerged and developed over the past few decades as a way to manage the relationship between business strategy and technology—a relationship that has been a challenge for many organizations to balance. Helping to reduce risk and ensure that IT projects achieve their business objectives is an established discipline for linking strategy, requirements, and constraints with a viable execution plan. Its benefits have been demonstrated on individual projects and on an enterprise-wide scale.3

Analytics architecture refers to the applications, infrastructures, tools, and leading practices that enable access to and analysis of information to optimize business decisions and performance. Architectural methods and techniques are designed to deliver a robust, cost-effective platform that supports current requirements and that can evolve to support future requirements without costly rework and disruptions.

The success of big data analytics programs calls for many considerations such as culture, governance, and sponsorship. In addition to the analytics architect, it requires an analytics ecosystem that involves collaboration between a number of roles, including the data scientist, the data steward, and the chief analytics officer (CAO).4

 

The analytics architect as data scientist

Of these data-centric professions, the analytics architect leverages the established architecture discipline to help ensure that business strategies align with the powerful capabilities of analytics to achieve business objectives consistently and cost-effectively. Because the analytics architect requires analytical skills and a data-driven mind-set, the role is somewhat similar to that of the data scientist.5 However, the analytics architect leverages knowledge of the organization’s information, application, and infrastructure environment as well as the current technology landscape to design a holistic and optimized analytics platform.

The role is central to the optimal analytics lifecycle, in which data from an operational system feeds the tools used for discovery, analysis, and modeling. Insights from this analysis feed back to the operational system to provide input that supports decision making and business processes (see figure).

 
The Emergence of the Analytics Architect – figure

The analytics architect’s central role in enabling the full analytics lifecycle

 
Key responsibilities of the analytics architect include operationalizing analytics, mapping requirements to implementation, selecting technology, and evaluating nonfunctional attributes such as security, usability, and stability.
 

Operationalizing analytics

IBM, in collaboration with the Massachusetts Institute of Technology (MIT) Sloan Management Review, conducted a survey of nearly 3,000 executives, managers, and analysts.6 Survey results found that embedding insights to drive business decisions and actions is one of the key incentives to realizing the business value of analytics. Performing discovery, analysis, and reporting on a snapshot of data in a one-off, stand-alone fashion is valuable.

However, organizations can amplify this value enormously by using the insights gained to inform decision making. Doing so involves providing feeds of live operational data in a constantly flowing loop. To operationalize analytics, analytics architects close this data-to-insight-to-action loop, which requires deep understanding of the applications and integration infrastructure environment.
 

Mapping requirements to implementation

Big data analytics architecture often needs to accommodate many and sometimes conflicting requirements and constraints. Requirements come from diverse stakeholders, such as line-of-business users, data scientists, analysts, and administrators. For example, data scientists need tools that enable free-format and ad hoc discovery, while business analysts need tools for structured discovery and tools that automate complex tasks such as data preparation.

Integrating a modern big data analytics platform with an existing data warehouse is also an important requirement. Constraints arise from the inevitable limitations with respect to budgets, schedules, skills, and current environments. Analytics architects analyze, shape, and prioritize these requirements as well as ensure they are implemented within current constraints. The implementation should satisfy current requirements and support future needs without significant rework.
 

Selecting technology

Analytics architects are responsible for selecting appropriate technologies from the many open source, commercial on-premises, and cloud-based offerings available. Integrating a new generation of tools within the existing environment is also crucial to help ensure access to accurate and current data. One example is IBM® SPSS® Modeler data mining and text analytics software, which allows business users and analysts to perform many tasks that were previously exclusive to data scientists with advanced statistical skills. Because technology in this domain is evolving rapidly, another important responsibility of analytics architects is ensuring that components can be replaced with well-suited alternatives that do not require any adjustment or downtime.
 

Evaluating nonfunctional attributes

When selecting technologies and building the analytics platform, analytics architects need to consider not only the functional requirements, but also the nonfunctional attributes of platform quality such as security, usability, and stability. Analytics architects plan, design, and monitor these key characteristics of the analytics platform to help ensure that it complies with enterprise standards and that it performs adequately as additional analytics solutions are implemented.

 

Cultivators of analytics architect skill sets

To help ensure a successful big data analytics implementation program, organizations should cultivate analytics architects from their existing talent pools of solution architects with analytics skills, or data scientists with engineering and technical skills. Conversely, practitioners in the architecture profession should develop analytics skills to fill the demand for this vital role.

Please share any thoughts or questions in the comments.

1 Big Data Imperatives: Enterprise Big Data Warehouse, BI Implementations, and Analytics, by Soumendra Mohanty, Madhu Jagadeesh, and Harsha Srivatsa, Apress, June 2013.
2 ““Preparing for Analytics 3.0,” by Thomas H. Davenport, Wall Street Journal, CIO Journal blog, February 2013.
3 A Systemic Perspective to Managing Complexity with Enterprise Architecture, by Pallab Saha, IGI Global, September 2013.
4Going Beyond Data Science Toward an Analytics Ecosystem: Part 1,” by Ahmed Fattah, IBM Data magazine, March 2014.
5Deriving Innovation from a Data-Driven Mind-set: Part 1,” by Ahmed Fattah, IBM Data magazine, January 2014.
6Analytics: The New Path to Value,” by Steve LaValle, Michael Hopkins, Eric Lesser, Rebecca Shockley, and Nina Kruschwitz, IBM Institute for Business Value, IBM Global Business Services, Business Analytics and Optimization, in collaboration with MIT Sloan Management Review, October 2010.

 

 
Previous post

Link into IBM Data Magazine

Next post

Delivering Minimum Viable Analytics

Ahmed Fattah

Ahmed Fattah is client technical adviser, financial services and public sectors at IBM Australia.

  • George Mouratidis

    Thanks for writing this article. I found it very interesting and it gives us an example of how the IT industry continues to evolve and mature. Here are my comments:

    1) I think closing the data-to-insight-to-action loop is key for organisations that wish to achieve an advanced level of maturity in information management as a way to place themselves in a stronger market position. I see this as supporting proactive evidence-based process improvement, whereby the actions taken are more clearly justified and traceable to the insight gained.

    2) Regarding mapping requirements to implementation, the prioritisation of the requirements as you describe, lends itself to an iterative/agile approach to implementation, which can help achieve faster return on investment, obtain buy-in from stakeholders, validate technology and architectural choices, and manage risk.

    3) The importance of data quality is probably worth a mention here. In this third phase of using data to help deliver business insights, it is even more significant to have valid and accurate data to rely upon, for improved discovery, visualisation and insight. Companies which have well understood and organised operational systems, where business information originates, will be able to obtain greater value from their analytics initiatives.

    4) Following from (3) I believe that the relational model (and good conceptual data modelling), will continue to play a key role here as foundation knowledge for achieving well-designed databases which can support diverse downstream uses of the information.

    Thank you :-)

    • Ahmed Fattah

      Hi George,

      Thank you very much for very valuable insights. Here are brief comments to your points.

      1) Agree and I like very much the “evidence-based process improvement” point.

      2) Agile is definitely a very integral to the analytics approach and feeds to the previous point.

      3) Fully agree. Many companies want to do advanced analytics realise that they don’t have the foundations of data quality.

      4) I need to understand this more. Hi George,

      Thank you very much for very valuable insights. Here are brief comments to your points.

      1) Agree and I like very much the “evidence-based process improvement” point.

      2) Agile is definitely a very integral to the analytics approach and feeds to the previous point.

      3) Fully agree. Many companies want to do advanced analytics realise that they don’t have the foundations of data quality.

      4) I need to understand this more especially in the context of ‘big data’. Although I agree that enterprise data with known value must be analysed, modelled and persisted in well-designed databases, in many ‘big data’ scenarios modelling is postponed and there are now emerging tools that interpret the data ‘automatically’.

      Thanks again for your feedback.