Data analytics is an ever-evolving space, decades old but new at the same time. Data scientists, faced with an expanding universe of data and a steady stream of new tools, bring an artful combination of hard mathematics skills, creativity, and intuition to the search for solutions hiding in data.
The art of data science and analytics, unfortunately, isn’t limited only by the imaginations of its practitioners. Business considerations such as time, competing priorities, and, of course, money, have a hand in determining which ideas become working projects and which remain wishful thinking.
Data Informed asked several data practitioners what projects they would work on if time, money, and staff were unlimited. The answers reveal the passions and, in some cases, the frustrations of experts who see problems up close every day and know they have access to the tools that could help solve them, if only resources would allow.
Javier Aldrete, Senior Director of Product Management, Zilliant: Predictive analytics can go a long way in making sales reps more effective and efficient. A killer app for sales reps would not only provide answers to everyday decisions such as which customers and prospects to spend time with, what products to sell, what prices to quote, and the most efficient routes to follow, it also would recommend crowd-sourced selling tactics that are most likely to improve the rep’s effectiveness and specific actions to take to maximize their compensation.
Andy Cotgreave, Tableau Software’s Senior Data Analyst in the UK: If money were no object, I would build an analytical system that allowed us to throw out all static charts and interactive dashboards. By their nature, dashboards and static charts only answer the question or questions envisioned by the designers. But this doesn’t fit into our nature or our business. I see this trend beginning on blogs, but it’s not mainstream.
We ask questions each day. Each day, the overriding question will change. Yesterday’s dashboard answers yesterday’s question. It doesn’t enable a conversation.
I would create a platform that allows people to work together, simultaneously, on a canvas. They would bring in data from any source (the platform would, of course, intelligently shape and surface the data in the most appropriate way). People could drag and drop fields around and instantly see an appropriate visualization of the data. These could be dragged around a canvas, like Post-it notes. You could throw them away. Or add more data to them. Or share them with colleagues. Most charts would be temporary. It would answer maybe one or two questions, but inspire more. A colleague could take that chart and re-pivot it, or throw more data into it to develop the question and respond to the original chart.
This is a true data conversation, reflecting the way conversations work: we ask questions, we explain our reasoning, and then someone else replies, adding more light to the answer, or disagreeing, maybe. As the conversation progresses, people learn more. Consider the following example:
Andy says, “Look at this bar chart – it shows we’re struggling to sell some of our widgets.”
Scott adds new data to the chart and says, “Interesting, but look at our marketing spend: we’re not focusing on these widgets.”
Andy, as he re-pivots it to a time series, says, “Ah yes, I hadn’t realized that. But if we look over time, we used to focus on those widgets.”
Alyssa pulls in the competitor data and re-pivots the data into a map. “But look here. You can see our competitors our killing us in Europe. We need to focus marketing spend there in order to succeed.”
Scott, Andy, and Alyssa have learned something about the business by interacting directly with their data, collaboratively. They weren’t shackled by yesterday’s charts. They weren’t waiting for IT to complete a dashboard. They had all the data, miraculously formatted to fit their needs. And they had a blank canvas on which to bring things in, drag them around, or throw them out, in order to have a conversation.
Dan Graham, General Manager of Enterprise Systems, Teradata Corporation: The first application that comes to mind would be a massive supply chain monitor. The application would be a real-time application that would drape over everything we know about supply chain: logisitics, order entry, and all the analytics, and being able to look around and say, “I am going to track and trace every package, every piece, everything that’s being moved throughout the system. Whatever I am distributing, I want to track every one of those items at the most granular level, get it to its destination, correlate it with my suppliers’ on-time deliveries and with my shipping logistics: Are those particular service suppliers doing their job? What’s their average on-time delivery in this particular route?
And then, of course, at some point, a disruption occurs. So in real time, we discover weather has hit the Boston area and no trucks are moving. The application should have the intelligence to say, “How do I re-balance my commitments? Are there alternate routes? Are there roads where some shipments are moving through?” And I want to know what my contracts are saying, what penalties I have to pay. My buyer doesn’t care that I have a shipping problem or that there’s a snowstorm. They have penalty clauses. I am going to look at my penalty clauses and weigh my options, do some analytics, do some historical checks, and get these goods as optimized as possible to their destination, to the shelves and to the buyer that wants them.
The same kind of parallel application should be applied to healthcare, on a national level, for every patient, every doctor, every pharmacy, and every hospital. A real-time parallel application could provide diagnoses, compare potential drugs and treatments, and provide recommendations and probabilities of outcomes. This parallel application would also look across all of this and bring much more granular analysis, collecting more data and providing more analysis at the individual, city, and national level. You could also pull actuary tables and provide risk assessments to consumers. You could recommend diets, you could recommend specific tests. I have helped hospitals do a little of this, but it needs to be broader. It needs to be a much bigger picture. The possibility is within our grasp. The possibility is there.
Dr. Colleen McCue, Senior Director of Social Science and Quantitative Methods, DigitalGlobe: The new capabilities that come online that allow for the immersive experience, I think those are really, really exciting. And we can see as people begin to incorporate motion and different ways of visualizing, the analytics results become a much better fit cognitively for how we understand information and how we find those novel relationships and those insights.
The one thing that I would like to see in our community is more trained analysts. The technology is advancing rapidly, and we are seeing great access to truly huge and massive data like we have never seen before. The processing is even better, but the truly gifted analysts become a rate-limiting resource, and one of the things we have been talking about in the community relates back to analyst training.
For a long time, we had analysts who were trained in specific technologies or capabilities and it became almost like analyst technicians. And it was great to be really well versed in a specific source or method. But then they became a little bit unable to adapt rapidly as the capabilities came online. So now we are thinking more in terms of going back a little bit – and this is where I think the concept of data science is very well timed – in training analysts in analysis as a process. So (analysts will) understand from question formulation to acquiring and processing data to inserting their favorite analytic methodology, and then be able to validate the results and create output that is actually operationally relevant and actionable. Being able to create analysts who are open to new capabilities and sources that are being developed, including those that haven’t even been developed yet, that’s what I would like to see developed next, and I think that is where we really need to start doing some work as well.
Jack Norris, Chief Marketing Officer, MapR Technologies: The ideal analytics projects are those that integrate analytics with operations – processes that integrate production data flows with analytics to impact business as it happens. We have customers today that have deployed these types of analytic applications to increase revenue, reduce costs, and mitigate risk. Examples include a major retailer that integrates analytics with its online presence to personalize recommendations, leverage its nearby on-premise locations, and significantly increase revenues; a large financial institution leveraging new and greater data sources for real-time fraud detection; healthcare organizations redefining patient care; and many more. Perhaps the best examples of this trend are Web 2.0 companies that have built new businesses around these new integrated analytic approaches, companies such as Ancestry.com, Beats Music, and the Rubicon Project.
Subscribe to Data Informed for the latest information and news on big data and analytics for the enterprise.