It takes technology as well as people to make analytics viable. But what tools and technologies do the pros use? What advice do they give?
At IT World Canada‘s Analytics Unleashed, panelists Mike Armson, data associate, Sklar Wilton and Associates, Michael Morris, director, data and analytics, Global Furniture, and Eugene Y. Wen, vice president, global advanced analytics, Manulife, sat down with moderator Jim Love, chief information officer and chief content officer at IT World Canada, to chat about what they use, and to offer tips on how to choose what works for your application.
Love kicked things off by asking the panelists if there is a framework they should consider for data analytics, and what they should look for when choosing tools.
“It sounds kind of obvious, but really, what problem can the tool solve?” Armson said. “And you should have a variety of tools, and they should each serve a very specific purpose. It’s easy to get kind of distracted by some of the bells and whistles but stay close to what the tool can actually do for your business.”
Morris added that it’s also about looking for your pain point and asking how the tool can help solve the problem, and whether it will work for you. Also, he noted, it’s important to look at how it can help the organization in the long run, and ensure that “it’s not something that’s going to just sit there and everyone’s excited about, but nobody ever uses.”
Wen echoed these points, saying that companies should start with the problem, and then identify the tool that can solve it, its value, and its cost effectiveness. “The third thing, beyond the tool itself, is what happens after you get that tool in,” he said. “Are you going to get sufficient support from this vendor? And is this system fitting your own company’s architecture principles? … All these need to be balanced, when we decide what type of tool and what our selection going to be.”
And, said Morris, he looks for Canadian vendors, or those with a Canadian presence. “Sometimes, when you’re dealing with someone in Texas from an American organization, they really don’t understand how Canadians do business.”
Another key factor for him is the ability to pick only the components of a solution that you need, and expand from there, rather than having to take on a huge monolithic package. “We have a certain budget that we have to hit,” he said. “And sometimes a lot of these organizations come in thinking you’ve got C$100 million to spend, and we don’t, so that makes a big difference for us.”
Potential customers can fill in a lot of blanks during a trial, Armson observed, just by using the tool in the way they anticipate would be their daily use. Morris added that “sometimes we’ve actually sent a data set to them. I mean, we signed confidentially, and said, ‘look, we want you to do this, show us what you can do with it as well’, because they’re the experts. It takes time for the people you work with, no matter how great they are at doing their jobs, to get up to speed on whatever product you’re purchasing.”
In a large organization like Manulife, Wen said, the trial is important, but it’s also important to remember that the analytics team is not the final user. “We need to get our end users, which are business frontline decision makers, our frontline employees who serve our customers, and the people who make decisions related to different strategic choices, (to the point) where our analytics can support them either operationally or strategically. And those inputs are also equally important needs to be considered, in addition to the bells and whistles of the tool itself.”
He cited the analogy of a composer who sometimes writes music for other musicians, and sometimes writes for regular audiences. In an analytics team, he observed, “we do need to make sure we know what our core focus is going to be. If all the work and the tools and methodology and models that we develop are more trying to impress our peers, which is important; it means our methodology is advanced. And I do promote that in our organization among data scientists – but at the same time, most importantly, [we need to] make sure our work actually serves the business and generates value and the business users can use that in an intuitive way.”
Morris agreed, noting that you have to make sure you know who the real user will be. Most of his training time is spent on business users, since he uses a lot of self-serve analytics.
The product pricing model is also critical for him as per user pricing, and pricing depending on the function in use, can quickly get out of control. “Maybe in the beginning, you’re starting with five people, but then all of a sudden it becomes 85 people and there goes your budget.”
Purchase timing also matters, he added, citing the fact that sometimes if you wait until the end of a quarter, salespeople offer deals so they can make their quotas.
Winning over users is also important when adding a new tool. “The vendors are doing a selling job on you, but you actually have to sell it to the people. Part of your job is selling this product to the people that you want to help,” Morris said.
But how, asked Love, can companies get the best from tools they already own?
Most enterprises have multiple tools in house, Wen said, and many have been there for years. He recommended staying in touch with the vendors and learning about enhancements to these products. You’ve paid for them already, he pointed out, so know what you have, and take advantage of training and other assistance vendors offer. Both Morris and Armson agreed that maintaining a good relationship with the vendor is crucial.
“Also,” said Morris, “you have to find [internal] people that are champions for [the tool] and get excited about it who aren’t necessarily data scientists, data analysts, and ask them ‘well, what can you do?’ They pull it along, so that you explore more into what you can do with the product.”
During a question-and-answer session, the topic of data governance came up, with all three speakers discussing their approaches. Morris said he’d used an artificial intelligence (AI) product to help clean the data. “It helped our data people get us data sets that were cleaned, so we could get a win,” he said.
Armson, however, said his organization hasn’t gotten to the point of using AI yet; during the analytics software trial, they replicated all processes, including manual data cleaning.
At Manulife, Wen said, they take things a step further. In addition to using AI to clean the data, they examined how and where what he called “quality messiness” was generated. “Comprehensive data quality and data governance, structure, principles and practices – it’s very important to address these issues,” he explained.
Morris agreed. “If you don’t have good governance, which is the foundation, it doesn’t matter how many of these tools you invest in, without the foundation, all the other stuff, it looks great, but it’s not.”
