Agile Analytics: 11 Ways To Get There
The accelerating pace of global business means that enterprises need more agile data-related systems and practices. Becoming more agile –- and succeeding at it -- isn't always easy given existing technology investments, constant technological evolution, and lingering cultural obstacles. No matter how agile your company is or isn't now, consider these important points.
![](https://eu-images.contentstack.com/v3/assets/blt69509c9116440be8/bltfd891477effc9ed7/64cb384adaed55322e7b2580/chess-1314359_1280.jpg?width=700&auto=webp&quality=80&disable=upscale)
Enterprises of all kinds are becoming more agile out of necessity. While the movement began with the software teams in many organizations, the rapid pace of today's business environment requires companies to become more nimble. Data-related initiatives, including analytics, are no exception.
"You're leaving value on the table if you don't have enough analytical agility in your organization," said Patrick Rice, founder and CEO of data science platform provider Lumidatum, in an interview. "A lot of people have standard reporting and dashboards in place, but we're going through the second wave of needing more agility in different places that go beyond standard reports for users."
Tool choices continue to proliferate, with the general trend being that they are easier to use. The idea is to put the power of data, analytics, and insights in the hands of more people in the organization faster without requiring PhDs to use the systems. Although, even in more technical contexts, technology layers or extensions are being added to existing systems such as Hadoop so that they too are easier to set up, deploy, and use. Of course, analytical agility isn't just about technology and solutions.
Create a culture where technology advances truly empower your business. Attend the Leadership Track at Interop Las Vegas, May 2-6. Register now!
"We're chipping away at the problem. A technology tool set is one thing, but if you haven't got the right mindset, the right people with the skills to use these tools, and you haven’t adapted the way you work, you're going to continue to work the old way," said Mark Marinelli, CTO of agile analytics solution provider Lavastorm, in an interview.
That's not to say organizations should attempt to reinvent themselves overnight. Becoming agile is a process that involves ideas, testing, iterating, and improvement. Rather than spending months or years building something that no longer maps to user requirements or the needs of the business, the point is to break large projects into smaller pieces, build on successes, learn from failures, and embrace an ethos of continuous improvement.
Still, many companies continue to chase the ideal of perfection at their own expense. They're attempting to build or deploy perfect systems or achieve perfectly clean data. The hope is to enable accurate data analyses and insights that can inform precise business decisions. Such a relentless focus on perfection can cause companies to spend unnecessary time and money on efforts that do not justify the incremental value they provide, however.
"I see companies fall down on this all the time. They want perfection and forget that good is good enough. They also forget that analytics are not perfect most of the time, so trying to create a perfect solution based on an imperfect signal, model, or outcome is actually compounding the imperfections. It's wasting time, money and effort," said John Lucker, global advanced analytics & modeling market leader at consulting firm Deloitte & Touche, in an interview.
Here are a few ways you can help your company achieve better ROI and greater analytical agility.
Fast access to data is important, assuming the right people are getting access to the right information at the right time. While speed is important, it's also necessary to keep agility in mind so it's easier to adapt to new technologies, new processes, new customer requirements, and new market conditions.
Culturally speaking, change is difficult because humans naturally resist it. As a result, companies need a sound and realistic approach to change management so that the shift from traditional business practices to agile business practices is a smoother one.
"I think the biggest thing is responding to changing demands within the business with some frequency. Traditionally, that has been a pretty difficult thing to do for BI shops in general. The question is how quickly they can move and address a new requirement. To me, that revolves around the changing aspects of the business," said Dave DuVarney, president and cofounder of BI solution provider Talavant, in an interview.
Companies that have adequate resources should consider establishing a center of excellence (COE). While the debate continues about where that organization should reside -- inside IT, outside IT, or somewhere else in the business -- hybrids are popular. In the hybrid model (which mirrors bimodal IT), some people in the group work in the centralized entity, while others are assigned to specific departments or business units.
The combination of centralization and decentralization allows the COE to have an enterprise-wide view of systems and approaches, and an understanding of what works, what doesn't, and why a certain thing relates to specific lines of business. The combination of broad and narrow views allows the COE to promote best practices throughout the organization. Without such visibility, duplicate efforts happen, including the purchase of redundant systems and solving the same problem again and again as if it were new.
"We see a center of excellence as a market of a pretty advanced analytical culture because they recognize a common tool set with some variability. Different departments can have different or supplemental tools, but there's a common approach and a way of harvesting the [best practices people are using] over time. Then, when people stand up a new project, there's a best way to stand it up," said Mark Marinelli, CTO of Lavastorm.
IT needs to be less of a gatekeeper and more an enabler to help their companies develop greater levels of analytical agility. Meanwhile, business units should avoid procuring data solutions without IT's involvement, because IT inevitably must fix any unanticipated problems, such as getting access to certain types of data or ensuring data quality and governance.
The centralized approach to IT that may have worked for traditional BI can impede analytical agility. As companies strive to embrace real-time businesses practices, there is no longer time to gather exhaustive requirements up front, enumerate them in a giant document, build the solution, test it, and then deliver it months, a year, or more later. Agile companies break down what were major endeavors into smaller pieces so they can be designed, tested, deployed, and improved at a much faster pace. In doing that, they are able to pivot the project as needed to better align with business objectives.
"You don't need to know the superset of all the data you're ever going to need for every question you're going to ask of the data before you can start. [You want to] deliver continuous value and also deliver continuous validation. As you're building these applications, you're questioning every assumption you made and validating it," said Mark Marinelli, CTO of Lavastorm
No single tool, model, algorithm, or problem-solving approach is appropriate in every circumstance. Even when a particular approach works, the dynamics of the underlying details may necessitate adaptation.
"Deep learning is great for things like image recognition and [for] some classification problems, but it doesn't necessarily solve some other types of prediction and forecasting problems," said Vince Jeffs, director of product marketing for customer decision management at enterprise application provider Pegasystems, in an interview. "The agility comes with your ability to get the better tool or model better suited to the job -- going through that test-learn-adapt cycle quickly."
Technologies, business dynamics, and processes will continue to change. Analytical strategies have to adapt. To stay competitive, organizations need people who are aware of emerging technologies and trends so they can anticipate how, whether, and to what degree those things may impact the business in the short term and over the long term.
"You have to have employees, especially data scientists and engineers, learning about new technologies and incorporating real-time analysis into your business," said Vadim Bichutskiy, director of data science at Innovizo, in an interview. "The old way of doing analytics is to have a central data warehouse with all your data and it might take days, months, or years to get insight from that data. The new way of thinking is to have real-time analytic systems that give you insight as soon as possible."
A common problem enterprises face is scalability. A lot of times the notion is dismissed, given the availability of massively scalable technologies such as Hadoop. However, Hadoop isn't a panacea for everything -- nothing is. Yet, organizations continue to search for silver bullets and in doing so disregard details that can hurt them competitively.
"Most enterprises have broken analytics stacks that don't scale. Their data is growing exponentially and the consumption is growing linearly -- and that gap is only growing," said Zubin Dowlaty, head of innovation at big data and analytics solution provider Mu Sigma, in an interview. "We're seeing a lot of clients that are unable to scale."
There are several keys to becoming more agile, many of which have to do with making things simpler -- breaking down large projects into smaller chunks, testing early and often to ensure that project is on track and on budget, replacing rigid systems with more flexible systems, and breaking down organizational and cultural barriers that make adapting to change more difficult.
"There's a rift between the analytics group and IT group, because they're just not agile enough," said Dowlaty. "If want to add weather data to sales data, I can call Weather.com, but I have to talk to my IT counterpart, and then the bureaucracy starts. You give me forms: What's my budget, who's my boss, am I politically connected? And all I want is weather data to correlate sales."
The business, data team, and IT all have to work collaboratively if an organization is going to improve its analytical agility. Deloitte & Touche recommends a six-step process to solution development that requires business-technology integration from Day One.
"The six steps are what frame out the well-constructed analytics solution. If you do that in an agile way, you keep in mind some of the basics like the 80/20 rule, which is important," said John Lucker, global advanced analytics and modeling market leader at consulting firm Deloitte & Touch.
The six steps are to start with a strategy, develop an analytics solution based on that strategy, embed the strategy and solution in the operational process, do whatever technology integration is necessary, manage organizational change, measure performance, and manage. Deloitte has clients imagine what they want to build and then assume that this imaginary solution has been built, including the system design and the data that needs to flow in and out of it. Then, a proxy solution is built with the understanding that it can be changed as necessary.
Organizations have potential access to more data today than they've ever had before, but they're not always aware of what they have inside their own companies, let alone what's available from third-party sources.
"The biggest impediment is not having data. You can get some data from the Internet, but you still need to know what's going on in your organization," said James Howard, a data scientist, analytics consultant, and adjunct professor of mathematics and statistics at the University of Maryland University College. "A lot of organizations aren't keeping their data, keeping summary data, or they're keeping the wrong data, which may have been helpful before, but now you need to know what a customer did on what day, what service they used, and the time they called you."
Organizations have potential access to more data today than they've ever had before, but they're not always aware of what they have inside their own companies, let alone what's available from third-party sources.
"The biggest impediment is not having data. You can get some data from the Internet, but you still need to know what's going on in your organization," said James Howard, a data scientist, analytics consultant, and adjunct professor of mathematics and statistics at the University of Maryland University College. "A lot of organizations aren't keeping their data, keeping summary data, or they're keeping the wrong data, which may have been helpful before, but now you need to know what a customer did on what day, what service they used, and the time they called you."
-
About the Author(s)
You May Also Like