Date science: The future of your business

By |  May 24, 2018
PHOTO: Istock.com/grandeduc/kasjato

Data science is becoming one of the most valuable resources for an aggregate operation. Photo: iStock.com/grandeduc/kasjato

A 2017 article in The Economist claims that “the world’s most valuable resource is no longer oil, but data.”

Back in 2012, Harvard Business Review published an article titled “Data Scientist: The Sexiest Job of the 21st Century.”

Statements like these refer to a field called data science, artificial intelligence, machine learning or data mining. Simply put, data science is the extraction of insights or knowledge from data with the use of statistics, computer algorithms, data visualization and other disciplines.

Great, but why should you care as a miner?

Early practitioners insist they possess a competitive advantage with it, that’s why. What does that mean to us in the mining industry? Is it just another management fad that will be gone in a few years, or can it help us improve our bottom lines?

In our industry, we are fighting for a nickel-per-ton advantage. Can we really afford to ignore this sweeping movement?

Why data science matters

Photo by Kevin Yanik

A wave of new production data is poised to make you a better operator. Photo by Kevin Yanik

First, though, let’s back up for just a moment. Computers have been around for decades. So why now? And where did data science come from?

The answers are quite simple. The amount of data that almost any business enterprise develops has exploded over the last 10 years, including customer data, production data and maintenance data – and the list goes on. This was made possible by the improved power and reduced cost in software and hardware.

Some equipment, for instance, incorporates sensors that can report readings of fuel usage, oil pressure, location and speed thousands of times per day over the internet. Are we taking advantage of all this new technology? Well, not exactly. New technologies also seem to add new wrinkles to the way we do business.

The world has changed forever. Business and technology are intrinsically tied forevermore. Twelve or 15 years ago, who would have expected a company’s social media presence to be so important? No one would have predicted the largest company in the world by market capitalization to be a computer/mobile phone company.

One reason for the boom in data is the development of free, open-source software that can find insights from mountains of data that were previously unfathomable. Anyone can play this game now if they have a laptop and the requisite data scientist abilities.

Kaggle.com, for example, which bills itself as the home of data science and machine learning, hosts machine learning competitions that are open to anyone. Recently, the internet real estate giant, Zillow, awarded $1.2 million to three teams that prepared the best algorithms for estimating the value of homes on their website.

Other recent competitions range from health care ($100,000 to improve cancer prediction) to grocery chains ($30,000 to predict sales). Lots of companies are forking out cash to use data science to improve their business.

Therefore, the manager of a mining company producing aggregate could well ask: What exactly could data science do for my company? After all, we’re not an internet business, our product sells because it is of good quality, consumers need it and we offer efficiency and good, old-fashioned customer service. What more can I do?

Well, here are some examples of how data science could be useful to aggregate producers:

  • Customer churn (turnover) and sentiment analysis. Which customers are more likely to leave you? Sales reports may contain hard-to-see indicators on how customers view your company. Is it getting better or worse?
  • Credit issues. Which credit requestors are not likely to pay their bills?
  • Sales trends and marketing strategies. How much will we sell next year? Is there a way to group customers for better marketing promotions or pricing strategies?
  • Maintenance. The age-old question: Which piece of equipment is getting ready to have a breakdown? (It’s more than just oil samples, now.)
  • Production and logistics. An accurate demand forecast can help to manage inventory. Are you shipping to your customers from the most efficient site? Are you taking advantage of shipping locations when transportation is a huge factor in delivered cost?
  • Red flag outliers and fraud detection. Are there anomalies in your general ledger that suggest fraud?
Photo by Kevin Yanik

A number of plant operators are in tune with metrics that report fuel usage, oil pressure and equipment location, but the depth of data available to aggregate producers goes much deeper. Photo by Kevin Yanik

Getting started

Convinced on the need for data science? Ready to go out and hire one of these experts to up your mining game?

Well, hold on. They aren’t that easy to find.

A data scientist needs to have a hard-to-find combination of database querying skills, fluent computer language skills in R or Python (the tools of choice for machine learning analysis), statistical knowledge, a good background in mathematics and quantitative analysis, an ability to make sense of large datasets, a curious and explorative mindset, visualization skills, business sense and experience, and the ability to communicate to others through storytelling.

A good data scientist needs to be able to work across functional groups and be able to lead teams of individuals with varying job responsibilities and backgrounds. If you don’t have one of these unicorns on your staff, you can take a few routes to land one:

  • Grow your own. If you have someone who has some parts of the above skillset, there is training available online and at universities.

Almost all technical schools are now scurrying to develop curriculums for data science and analysis degrees for both bachelor-level and advanced degrees. There are a variety of online courses available. There are even teaching organizations that offer immersive and intense data boot camps that attempt to teach some of these skills to applicants in a few weeks of in-class training.

This can take at least 18 to 24 months. It can take longer, but the advantage to this method is that your data scientist in training knows your business and its culture. You know they already are a good fit. The disadvantage of growing your own is the time lost before they are effective.

  • Hire one. The advantage of this is they are ready to start almost immediately. They are not easy to find, however, as they are currently in short supply and are expected to remain that way for a few years. Of course, bringing in a new employee is always hard and carries an element of risk.
  • Hire a data science consultant. While consultants may be up to speed immediately on the technical abilities for which you are hiring them, they may not be familiar with your company’s structure, culture or business needs. Hiring consultants carries its own share of possible problems, and you must have someone coordinating their work and squiring them through the ins and outs of your management style. They may be experts in data science, but neophytes as to what really drives the mining industry.
Photo but Megan Smalley

Data science can elevate your approach to equipment maintenance. Photo by Megan Smalley

Data science detriments

So what’s the catch? Are there any negatives to rubbing the data science lamp and hiring some genies?

Yes, it’s true that a large percentage of data science projects fail. The reasons for this vary, ranging from not enough data to not enough good data.

Although data scientists are skilled in adjusting for missing and erroneous data, too much of this can kill a project. The skills of the data scientist may not be adequate for the job. Also, there may not be a compelling business need for data science. Any half-hearted attempt to just create one to see what happens will usually be consigned to the managerial dustbin as time goes on.

Quite often, the insights of a project are correct and important, but the company is unable to act on them. Change is hard to manage in every organization. The recommendations that may evolve from an analysis may run counter to company culture, or the “we’ve always done it this way” syndrome.

As an example of how this can be detrimental to becoming a data-driven company, last year Microsoft sponsored a data science conference in which the keynote speaker was a world-famous economist and data science consultant. His firm was asked to help a very large retail fashion company improve its inventory control strategy.

They were successful, despite barely hidden opposition by the buyers who had previously been making those decisions. Their first year’s recommendations made the company significantly more profit, but when they met with the CEO to discuss the next year’s work, the CEO asked them to drop the project and work on something else. He gave the reason that while the data-driven work had been wildly successful, it hurt the feelings of their company buyers. Company culture ruled the day.

Absorb the opportunities ahead

Certainly, change is a muscle we don’t like to exercise in the mining industry. Let’s face it: Many companies make strategic decisions on the HiPPO method – that is, the highest paid person’s opinion.

There may be very good reasons for that, as that person usually has the most experience and has shown to have good judgment. But to be a truly knowledgeable company, the first question when faced with an opportunity should be: What data do we have? The first question that should not be asked: Boss, what do you think?

Data science is just a tool, not a substitute for management knowledge, experience, and judgment. Data science can’t tell you how to run your company, but it can be a valuable part of your decision-making capacity.

Sometimes, just the process of organizing data and visualizing its features may be enough to give the insights you seek. Data science analysis isn’t always needed to solve problems and should not be a knee-jerk reaction to fix things.

Balancing the possible benefits with the challenges and costs, you might be wondering if you want to enter this unknown territory. That is a decision only you and your company can make.

A recent article titled “Machines Learning Astronomy” in Sky & Telescope tells how even astrophysicists and astronomers have been slow to use the techniques of data science in their highly competitive field. They are now making up for lost time.

According to the article, “many astronomers predict that machine learning will take on an important role in the field, perhaps becoming as essential as the telescope … a set of tools that’s only beginning to be explored.”

Says astrophysicist Joshua Bloom, a convert to data science: “From my perspective, it’s a little like being a kid in a candy store – before all the kids wake up.”


Ron Hendrickson is a data scientist with Mining Insights.

Comments are closed