Mike Rucker, Ph.D.

Interview at the Motionsoft Technology Summit about Big Data

This quarter’s Business, Innovation and Entrepreneurship interview is the compilation of getting to discuss “big data” analytics with four exceptional thought leaders at the Motionsoft Technology Summit this year (2016). These four gentlemen in no particular order are: Jafar Adibi, Ph.D., the President, Co-founder, and CTO of re|unify; Jeffrey Cooper, the Senior Manager of Business Development at Samsung; Mark Newman, the President of Heads Up Analytics, and Keith Catanzano, a Partner at 2River Consulting Group. The answers below are summations of their respective answers, as such they are not represented as verbatim but edited for readability and context.


1) When a company is either building a data model (or working with a third party for this type of service), what considerations should an operator have regarding the crossroad of complexity and usability?  There are scenarios where too many disparate and incomplete data sets can make it difficult to find the signal from the noise; what are the trade-offs as the amount of available business intelligence information continues to increase? And what considerations should we take into account to maximize any investment in mining data?

[Jafar Adibi]: You need to figure out what problem you are trying to solve. Clients will come to me with data, rich sets of data, and say, “Jafar, now go figure out something to do. Find something interesting.” Generally, this is a waste of time. People believe finding correlations (any correlations) are going to help their business, but that is often not the case. When you identify your problem, we are better set up to solve it. There are different analytic methods for classification problems, association problems, and other questions that are not necessarily answered through correlative means. Getting to the right question will help you establish what data sets are important.

Then you need to figure out your budget. There will always be noise in your data, especially data from business intelligence. We can build a model to take the noise into consideration. However, using more data is obviously expensive, so that goes back to what are you trying to solve for. We can exclude data that will not answer your question, which saves you time and money. As such, you want to keep return on investment (ROI) in mind as you think about the question you are asking. Ask yourself, “If I answer this question, how much money with I gain/save?” The answer to the ROI question gives you a ballpark on what it might be worth regarding your investment in a data model.

2) It seems to me that a lot of ad hoc advice about using data for business intelligence is disseminated on broad-based assumptions derived from general population data. However, is this not one of the follies of “Big Data”? Companies are basing important decisions on arguably misleading benchmarks, rather than creating a narrative specific to their population (or at least a sample from their specific population); What are strategies to ensure we are making the best decision based on our company’s unique attributes?

[Mark Newman]: The most important thing is to trust your own expertise. You should intuitively know the customers you are trying to attract. You should have an idea of what strategies you are trying to pursue. You should already know what the important problems are you need to solve. What you don’t want to do is look to data to validate some preconceived answer to your problem. Instead, you want to devote your own educated guesses as to what to do — and then you want to use data to test those rigorously to keep yourself honest.

I think there are two ingredients to doing that. The first is to agree with your colleagues on the definitions of the terms that you are using in your data. If all the stakeholders do not agree on the definition of the numbers, then you all are not going to have an organized lexicon/narrative to work with. You have to agree on key metrics that you are going to use to allow for the monitoring of health and progress within your organization.

The second ingredient that you want to have is to follow an experimental approach that is constantly evolving. Your customers and prospects are going to react differently to your products and services over time. Reasons:

What works today does not work tomorrow. Instead of some one-and-done, super solution to what you are trying to accomplish — instead you want to have some kind of innovative, incremental approach in the beginning. If you follow that, then over time, the data is going to have a narrative that reflects who you are, and what you are trying to do, and what works best for you.

3) Until recently, most data aggregation efforts have told a fairly unsophisticated narrative, and inspired relatively unremarkable initiatives in an effort to capitalize on data mining. How can we improve our use of data? And, how can companies do better at making data more actionable?

[Keith Catanzano]: What is the question the company is trying to answer? It is important to not just say, “How do you make data actionable?” We are probably all guilty at some point of looking at a data model and saying, “Look at the results, they’re awesome!” I think intriguing insights can be challenging in terms of making data actionable. There is a ton of data out there. Once you find ways to bring yours together, there is a lot you can see using data by way of insights. At some point you need to do something with the insights. In order to do that, obviously, it’s important to know who your customers are [assuming trying to influence their behavior is your goal], but also why are they customers. However, in this use case the why is more important than the who. The “why” is ultimately what you are going to try to make actionable, because to take action you are going to need to pull some type of lever to influence consumer behavior. There are lots of ways to work with communication or outreach in an attempt to accomplish this, but the effort requires the company to take a deliberate approach regarding how data is used to take action.

It is also important to note that making data actionable is generally not a one-shot deal, and architecting a campaign that changes an entire group’s behavior in some way probably will take a series of events that includes multiple levers I mentioned. So to make data more actionable, an organization should sit down and say, “What is the level of energy I want to put into solving or addressing this problem?” And that’s probably both a financial decision and a brand decision. For instance, a brand manager might ask, “Is this the kind of consumer group that we want to continue to attract? Yes; OK, well … indicators show we may be struggling with this particular group, so let’s double down because from a brand perspective, that’s how we want to be seen.” An alternative scenario here is the data suggests (to the brand manager) that too much effort is being spent focusing on the wrong group. Without asking the right questions, the data just suggests that marketing is ineffective. To finish, a company really needs shared responsibility to make data truly actionable. Ultimately, as an organization you determine what resources you want to put against data analytics, but knowing what question(s) you wanted answered first is important to making data actionable.

4) How will health club and health club member data evolve over the next several years — what will prove to be important signals for our industry in addition to financial, transactional and activity data?

[Jeffery Cooper]: So besides activity data from wearables, there will be a lot of contextual data the health clubs can now potentially get. With corporate wellness taking off you are going to see deep integration with insurance companies and insurance data. I believe, along those lines, health clubs will also be integrated more with the medical industry. As prevention becomes more associated with a basic level of fitness, I believe you will see medical data become relevant.

In that regard, I think prevention of chronic diseases is eventually going to drive a lot of people toward health clubs from the medical side of things. Right now, in most cases, doctors cannot write a prescription for a health club, but that could change as more complex sensors begin to validate the efficacy of fitness interventions.

Genomics data is another revolutionary area. You already have things like 23andMe, but there is a company Helix, which has been recently funded. Their idea is to sequence your genes, and license this data back through health care providers and fitness applications. With genomic data, consumers can make better choices (and health clubs can cater to them better). With this data, people can ask:

As science becomes more advanced, these companies will snapshot your genome once, and then as the science learns more and more about the genome — health clubs can take preemptive, proactive actions from that data to keep their members healthier longer, keep them out of the hospital and improve their overall quality of life.

5) Why does “Big Data” often fall short on delivering on its value promise?

[Mark Newman]: Personally, I feel that part of the problem is the way output data get reported. I feel that in data science to deliver a static report, it is potentially a sign that we have not done our job properly. The reason for that is because when we deliver a page of numbers, there is often no context to the end-user. When you are able to create/refine a business question, you generally make the presumptive problem simpler than it first appeared. Before you set off looking to get value from data, your organization should come up with your desired thresholds and metrics. Then instead of looking at static reports that, at best, will give you trailing indicators — build a dashboard that gives you real-time intelligence based on the most important metrics for your business. This dashboard should be something that your employees can always go to — not just some report that gets delivered on your desk — but something that is readily available on an ongoing basis. You also need to evaluate and monitor the efficacy of this dashboard on an ongoing basis. For instance, if you have a forecasting dashboard and there is a forecast your company is trying to meet, is the dashboard valuable and helping you meet your forecast?

I believe both dashboards that monitor things that drive your business forward, as well as insights that are actionable, are at least two things that give you some evaluation of whether “Big Data” is helpful and valuable within the context of your own particular situation. The other thing is that you really want to be doing analyses all the time. You want your data strategy to evolve past sending out graphs and numbers — to actually be working to build a story of what’s going on in your organization — and back up your story with reliable and meaningful communication so every stakeholder is seeing the same thing and you can all agree that your chosen data model(s) is providing value and is meaningful within the context of your particular business.

Exit mobile version