Data, talent, and the big, black hole

Turing the buzzword into working reality

Article first published: Oct. 21, 2019. Image: black hole rendition (outer space with colored donut of light)

Big data has been a buzz word for a while now. Data analytics has, too. But as to the detail behind those words, there’s a bit of a black hole. Making sure you are hiring the right person for a job that is defined as a black hole is a bit daunting.

Let’s change that.

There are many ways to look at data today. We will start with the established descriptive analytics. The most familiar form of this arguably being the demographics used by marketing: age, region lived in, salary, gender, …. These are descriptive, but hardly explain the complexities of your buyers, suppliers, and customers. The data captured and used for these analytics happened in the past, and unless you are rewriting history, it doesn’t change. Measuring how much inventory is in a bin at 5 pm EST on Monday, Oct. 1 in year whatever-it-is? You get one descriptive number, or at least you should.

Our business models, marketing models, and other data activities needing descriptive data for defining customer segments or how an assembly line performed have used descriptive data and its associated analytics for a long time. Because we can point back to the past and say the decisions made using the descriptive data and models worked before, or at least the aspects we were measuring showed they worked, we assume making decisions using stagnant past data will also “work” for today’s competitive threats.

But we know things actually do change. In fact, the changes have been dramatic for a while. What happened before provides lessons and specific insights, but not a full picture of what is needed for your business future. Conditions, requirements, values, regulations, policy (ours and others, business and regulatory, markets and politics), competition, technology; these and more are all in flux. Descriptive has limited application for a world, or supply chains, in flux.

Ok then, we’ll use the predictive tools we already have. The simulations, what if case studies, and gamification that we’ve been using for a while. These allow us to adjust assumptions and perform needed what if scenarios to obtain high probability outcomes based on decisions made.

Predictive analytics need more, much more, data. We use descriptive data captured in our systems for inventory, sales, purchasing, operations, and so forth.  We add in sensor data when we can. Even HR data can go into the mix. Blend these into a statistics-based model and provide leaders and decision makers with options that can be evaluated and acted upon.

Yet in a way, these are also stagnant. We can change input parameters, starting conditions, or randomly program in risks, but the answers are still based on assumptions and biases, which are very difficult to eliminate. They provide a range of answers instead of a point estimate along with a probability the answer is correct. But nothing really happens that we didn’t program in. There’s no intelligence in the model.

Because we are able to capture so much more information on how people interact with our products and services and websites and on the phone, in real time, and ….we end up with a lot of data to play with.

There was a time when my models and simulations were considered to be big data at 8 million data pieces per run. Now, we are talking oh so much more. We’re up to yottabytes, that’s a whole yotta bytes (sorry, couldn’t help it). More specifically, it’s 1024 zettabytes, one of which is 1024 exabytes, one of which is 1024 petabytes, one of which is 1024 terabytes, which many of us have heard of but can’t even imagine. There’s a lot of data out there –it is not all good or useful, but the quantity is impressive.

So, what do we do with it?

We can perform prescriptive analytics. While relatively new for real world analytics use, the statistical methods behind prescriptive analytics have been around for a long time just waiting for us to be able to capture enough data to make them useful. We have arrived.

Prescriptive means that we can analyze out to the future using advanced statistics on what could happen as well as have the model attempt to tell us why something does happen and make recommendations for actions that have not happened yet. Prescriptive analytics uses a number of prescribed options to arrive at a solution about future decisions we don’t even recognize need to be made yet. Now we’re throwing into our models business rules, real-time data, machine learning, AI, and what seems like computational wizardry.

Businesses want and need to optimize important business aspects like schedules, inventory, production, and maintenance. To avoid wasted time and money, we want to figure out when something is needed before humans even recognize the need. Why bring in a helicopter for blade replacement or an aircraft engine for an overhaul before these things are required (wasted life cycle and increased costs). At the same time, we also don’t want to depend on schedules based on past data when we could use up to the minute insights from technology combinations like analytics and digital twins to help signal when helicopter blades or aircraft engines need to truly come in (perhaps before expected) to assure the craft does not go down.

Early recall is based on the huge amounts of data collected from weather conditions, sensor data, unanticipated stressors, and so on. The digital twin provides a visualization of the component (engine or blades) and data (affecting the engine or blades) under scrutiny somewhere remote from the actual component. AI and machine learning are integrated to better understand instantaneously the health of the component.

But prescriptive is not limited to parts or services. It is used based on automated lines, customer responses, real time reactions, outside influences, and data that previously we could only imagine getting in Sci-Fi scenarios. Those scenarios are now real.

But if many people, including those in descriptive and predictive analytics areas, need to enter the sci fi area of prescriptive analytics, how do we find them? Hire them? Keep them?

Do they exist?

Yes, but you must ask for them correctly.

Here is an example. I was presenting on integrating environmental performance into existing operations and supply chains using a modular methodology. Prior to the presentation, one of the attendees stated he would like to learn more about how prescriptive analytics worked. Before I could answer, a representative of a local university piped in and said he should take their python and a few other newer programming languages courses because he would understand advanced analytics after taking those courses.

Actually, that was an incorrect answer. Very incorrect if you are trying to hire talent.

(We’re here to help you avoid incorrect hiring requirements. Watch our FREE webinar)

Python and other newer languages are used to capture, organize, and present data. But the data itself is used with advanced statistical tools, like cluster analysis or SEP (structured equation modeling) or… Languages like python put the statistical models into usable form. If you want to analyze data, you look at what came out of the statistical models. The programing language is not what is important here. Mess this up and you get an excellent programmer who knows nothing of the statistical interpretations of the output.

It comes down to knowing your business, knowing the challenge you face, the real problem to solve, and what is actually needed to correctly evaluate analytic outputs and potential decisions to be made.

Big challenges, but ones that need to be conquered because your competition is/will be figuring this out as well. If your competitors are on top of this, or even just contemplating getting started, you need to be, too.

And because talent is tight, and talent that can help with today’s new technologies and their uses is even tighter, you need to start sooner rather than later.

It’s a matter of survival or entering the big, black hole, never to be heard from again.

 

Keywords and Concepts: Big data, descriptive analytics, predictive analytics, prescriptive analytics, talent, hiring, requirements, leadership, decision making, competitive advantage, supply chains, customers

Cynthia Kalina-Kaminsky is the president of Process & Strategy Solutions  The Recession-Proofing Your Business series helps owners, CEOs, and decision makers plot paths to growth, retain talent, and increase revenue by eliminating internal chaos through improved business performance customers pay for.

For companies ready to find the professional talent and executive leadership to drive your data initiatives and more, watch our FREE webinar