Countdown to success with analytics

By Daymond Ling, Professor, School of Marketing, Seneca College

Is there an element of hype to analytics?

I’m purposely not going to mention Big Data if I can manage, because the very terminology simplifies the problem.

Analytics can lead to better business outcomes. This has been demonstrated time and time again, when the business can remain relentlessly focused on making itself more competitive. Sometimes, we get the impression that by gathering a lot of data and running it against algorithms, BOOM! Instant business results.

Unfortunately, it’s not that simple. Before a business can reap the undeniable benefits of analytics, there’s a lot of work to be done. And much of it has little to do with the technology. Much of it has to do with business culture and process.

I’ve prepared this countdown list to help guide businesses to a successful analytics regimen. I hope it helps prepare new organizations that are diving into the analytics waters, debunks the notion that analytics is easy, and demonstrates that the payoff in the end is worth the investment in time, resources, and focus on business outcomes.

5 Things Your People Need

Analytics starts with people. You can throw all the technology in the world at a business problem, but if the analytics staff doesn’t understand the business problem, you likely won’t get very far. For a successful analytics program, your people need the following five competencies.

Domain competency. This may be the most overlooked competency by your analytics staff—do they know what business they’re in? For instance, at a financial institution, you’re dealing with different strata of clients—low-income, middle-income, high-income, and extremely wealthy. It’s important to understand what parameters affect client service depending on where they fit in the client strata. Specialized knowledge is necessary regardless of the vertical: construction, legal, medical, retail, transportation, whatever business you’re in.

Process competency. This is related to domain knowledge; every vertical has different processes, though they may be fairly standardized across verticals. It’s important for analytics staff to understand them, because—as we’ll discuss later—gaining any business advantage from analytics depends on changing business processes to more efficiently and effectively serve the client. If they don’t “get” the process, it’s difficult, if not impossible, to change it in an advantageous way.

Data competency. With the two former competencies nailed down, analytics staff can move on to a grasp of what data is meaningful to the organization. We’re swimming in data. We’ve had structured data from spreadsheets and databases for years now. More recent sources are unstructured: data from social media sources; data from the Internet of Things (IoT), like sensor and telemetry data. What of these data sources are meaningful to the company? Sentiment data from social media feeds is much more important to a retail or service business than it is to a wholesaler or logistics company, whereas telemetry would be much more important to the latter.

Analytics competency. This seems a no-brainer, and it’s why data science graduates are in such high demand. It begins with an ability to see the relationships between data sets and extends to the ability to discover relationships that aren’t so traditional or obvious. It’s a competency that can be taught, but it’s most effective when experience can make it serve the competencies mentioned earlier.

Story-telling competency. Defining relationships among data sets, detailing how Input 1 affects Output 2, predicting how a process change can affect the bottom line—that’s material for a technical review. Analytics that don’t lead to action are meaningless. If you can’t tell a business audience—your C-suite—what it means to them, nothing will happen. Analytics staff must be able to tell a business story, in business language, for a business audience. Here’s the business problem; here’s the data; here’s what I’ve found. Keep the math and algorithms to yourself and your peers.

4 Building Blocks for Analytics

People. We’ve discussed this earlier. Business and process knowledge are critical to great analytics; being able to translate the math and algorithms into something meaningful to the business are necessary to achieve any results.

Analytics process. For your people to be able to tell that story, they have to have confidence the story they’re telling is accurate. Does the analytics and data collection process stand up to rigorous scientific scrutiny? Are the relationships between data sets an accurate reflection of the business?

Data. Almost as much as people, data is at the heart of analytics. Its integrity must be ensured by scrubbing. Its security—sometimes overlooked from an analytics perspective—must be preserved (though some data demands more security attention than others). Its collection, especially in an IoT environment, is challenging. Most importantly, though, its relevance must be evaluated before it is given any weight in an analytics environment.

Tools. The tools you need depend on the user and the application. While you need reports and visual dashboards, you also need machine learning algorithms that predict with high accuracy, and powerful optimization frameworks that make optimal decisions. Ideally, analytics tools should allow a semi-sophisticated user—a line-of-business user, not necessarily the IT department—to discover relationships among data sets in an intuitive manner, but also equip power users with state of the art techniques for heavy lifting.

3 Steps Analytics Management Must Look After

Governance. It’s a wide-ranging topic and deserving of a post on its own. It comes down to two things in the end. Governing the data infrastructure – what's stored, is it accurate, is it secure from abuse – is critical as data is the raw material for analytics. You also need to govern what problems to work on as demand will surely outstrip your ability to deliver, focus on what will move the business forward, and avoid curiosity questions that add no value.

Discovery. It's the job of the analytics team to assess the situation, do root cause deep-dive to identify core issues, and determine a course forward. They need to use methods appropriate to the situation, be it simple static rules, dynamic pattern recognition, Bayesian adaptive control, or real-time analytics.

Action. We’ll discuss this in more detail in the next section, but where the rubber meets the road, the rubber has to meet the road. The analytics team has to have the influence or authority to enforce changes to business processes. Without change, analytics is pointless.

2-Step Dance—Analytics and Action

We’ve been through the elements of usable analytics; I don’t think there’s any need to dwell longer on them. But what makes analytics useful is the action an organization takes based on the discoveries it’s made through analyzing its vast collection of data.

They say that “Culture eats everything”; analytics is no exception. Analytics will fail in organizations that don't have a learn-improve culture. Organizations must have the will to learn by setting aside pre-conceived notions, and have the will to change processes in light of the insights that analytics delivers, or it’s simply money thrown away.

All too often, analytics are employed simply to validate the opinions of someone in the C-suite. This contaminates the whole analytics chain—what data is used, what analyses are performed, what results are delivered. In these cases, analytics are used to prevent action from being taken. With your eyes closed and your ears plugged, analytics are meaningless.

1 Goal

Analytics is about delivering competitive advantage. Anything else, in an enterprise context, is academic. Keep your eyes fixed on the prize, and make sure your analytics team is working towards the continuous delivery of value to the business. Keep this focus, and the rest should follow.

Daymond is an advocate for the use of Advanced Analytics to create business value. Able to bridge between analytics and business, he has created and sustained analytics departments that drive for innovation, he is an educator and mentor for many analytics talents, and he is a frequent speaker at analytics conferences. Currently Daymond is Professor, School of Marketing at Seneca College. Previously, he worked at Canadian Imperial Bank of Commerce (CIBC) where he is Senior Director, Modelling & Analytics responsible for all facets of client analytics. Prior to CIBC, he worked at American Express Canada in Risk Management where he received the distinguished Chairman’s Award of Quality for his innovations. Daymond holds a Bachelor of Science degree in Honours Physics, and a Master of Science degree in Operations Research.