The promises of advanced analytics are inherently appealing to businesses: make real-time, data-informed decisions, optimize and automate processes, and monetize data in new ways. In the second installment of a three-part blog series, we’ll examine the role of the organization and technology in establishing a performant advanced analytics capability.
In the first blog in our series, we provided an overview of the success factors for establishing differentiated advanced analytics capabilities and explored the importance of team structure in greater detail. Today, we’ll unpack the importance of our second and third factors:
- Ensure analytics teams have strong leadership and alignment to business value
- Develop an analytic organization and culture through proper role identification, career development paths, and recruiting
- Establish modern tools that fit into existing technical infrastructure
- Have a well-defined development methodology to lead from idea to prototype to production deployment
- Operationalize models for proper business value
Success factor 2: Develop an analytic organization and culture through proper role identification, career development paths, and recruiting
As leaders gain momentum, they should shift to an organizational design that will evolve the team to the next level. This requires a focus on two key areas:
- Identify team structure
- Embed technology and process into the organization
1. Identify team structure
Depending on your situation, you may choose to tackle this by formalizing the fundamentals of job design, career pathing, and competencies. Or, if you are tackling this in a more rapid fashion this may be a less formal activity. If your analysts have the core competencies needed for your business objectives, the group may be able to make great strides initially as a small, agile team. But at some point, to attract and retain the right talent, it will be imperative to devise a compelling team structure with proper roles, titles, career paths, training opportunities, and inspirational leaders that candidates are proud to be a part of.
2. Embed technology and process into the organization
Another fundamental question is determining where the team will sit—in IT, in business, or in its own group? A case can be made for any, and we’ve seen all combinations. A team that sits in the business has direct responsibility to achieve business goals. However, a team that sits in IT has much more control over the toolsets and processes required to operationalize those achievements. Hybrid or external options attempt to address both but require empowering a leader with a high level of organizational control and a seat at the table for major decisions.
This can also be a great time to bring in outside help to rapidly establish the capability—complete with skilled resources, proper development methodology, and experience working with business teams—to deliver business rapid business value, and then slowly transition ownership back to the organization when ready.
Success factor 3: Establish modern tools that fit into existing technical infrastructure
The number of advanced analysis toolkits on the market has exploded over the last several years. Many organizations have a variety of legacy tools in their arsenal: SAS, SPSS, and others can be powerful but are often expensive, have limited offerings, and don’t fit into many modern developers’ skillsets. No matter the option you choose, it should satisfy two core criteria:
- Alignment to business needs
- Fit within current infrastructure and team capabilities
1. Alignment to business needs
Today there are countless commercial, free, open source, and pay-as-you-go services or platform options to navigate when establishing an advanced analytics foundation. Momentum has been shifting in these directions, with a lot of advances coming from popular public cloud providers.
For example, Machine-Learning-as-a-Service (MLaaS) platforms such as Microsoft Azure’s AI APIs support video and image classification, natural language processing, recommendations, and easily plug into existing infrastructures with little knowledge required of how they work.
New bundled tools like Databricks or H2O.ai enable automatic machine learning on very bespoke problems, allowing developers, who may not have significant mathematics background but know how to work with modern data platforms to engineer effective solutions.
There are also new initiatives in cloud processing that let you bring ML to your data, instead of the other way around—drastically cutting preprocessing time and data cleansing initiatives. Google’s BigQuery, for instance, allows user-defined functions to run on massive datasets, and can then run advanced analytics in the same platform. Apache offerings such as SparkML run on your pre-existing Hadoop platform with minimal reconfiguration and can easily become a part of your streaming data workflow. And Amazon Web Service’s (AWS) Sagemaker attempts to bundle these components into a single interface that’s familiar to existing AWS customers.
2. Fit within current infrastructure and team capabilities
Ultimately, the toolkit that you select should be based on your needs, existing infrastructure, and team capabilities. For example, not all organizations need to do machine learning in real-time on massive datasets, so simpler setups like using Jupyter notebooks may be appropriate for a small, growing team. Technology choices also need not be major commitments: with the advent of modern data platforms, agile architectural choices become possible, easily turning on and off certain features or capabilities within service providers you’ve already established. The most important thing is knowing it’s a journey, growing and adapting your toolkit over time so it becomes a normal part of doing business.
In the final installment of our series, we’ll apply learnings from leadership, organizational design, and technology to explore the role of development methodology. Most importantly, we’ll also discuss how to mobilize advanced analytics for value.
