Data Science for the C-Suite

What Do Executives Do?

Make decisions. Ideally, they make decisions driven by information. That information is usually distilled from various data. As far as “actionable insights” go, the decision as to which actions to take falls into the hands of company executives.  They set the course of the company’s direction by making important business decisions on a regular basis. It is thus imperative to the company’s long-term success that these executives can see the results of the actions they take. 

Does this information lie in various processed data sets? Maybe. But only if the data processing is focused on the right questions. Readable and informative models go a long way in helping decisions be made for the sake of the company’s future. Risky business moves with the potential for high future payoff cannot be made in absence of this information. It is only with substantiated business records and evidence that the potential of these high-level decisions can be assessed. For all this to work, the analytic end report must be directly useful to the business.

Superpowers for Executives: Data-Driven Decision Making 

Being able to interpret massive sets of data generated by companies allows for appropriate conclusions to be drawn about what it all means. With a strong analysis of business data the firm now has information on its side in the market, allowing a company to make informed decisions that allow them to better optimize operations. “Data-driven decision making” enables us to better guide the course of the company because it allows us to understand the implications of operation. 

So two simple questions:

1)  If this true, why doesn’t everyone be deeply data-driven?

2)  And why do so many do it badly?

One answer to both is that data is rarely enough. Transforming that data into actionable information is what truly matters. Examples abound of high-level data analysis in optimization. It minimizes unnecessary energy consumption and automates efficient traffic flow. It brings great predictive power as well, such as in the prediction of machine breakdowns or for stock market forecasting. But only if the analysis turns the data into information.

How fast can you really make a datainformation-driven decision?

The timeline for all this varies, of course, depending on the size and ambitiousness of the project taken on by the analysts. The first challenge is getting the analysts and business executives on the same page. The best way to do all of this is better question-asking. Getting both managers and analysts in sync is already critical. After all, miscommunication of expectations can be an expensive failure if the data model is useless to the company. 

Data Analysis for Dummies

A quick look at modern data analysis might be instructive as to the potential hurdles. How long does it take to establish with full clarity what the goals are? What are the questions we NEED to ask versus the easy to ask or, worse, what we have always asked? Only then, the data is passed into the hands of the analysts.  They organize it, run critical evaluations, create a good model, and, in the case of “smart” technology like deep learning, apply an appropriate algorithm so the model learns and develops itself. It is in the best interest of the analytics company to work both thoroughly and efficiently in its process. While it should not take too long in generating a model to present, the analytics company also doesn’t want to overlook potentially important information. Data Scientists are expected to perform their tasks in days or weeks rather than months, since important business decisions are rarely allowed to sit for that long. 

Finally, it is on the shoulders of the business to integrate the results of the analysis report into their business model. For a company, the ultimate goal of all this effort is to optimize business practices and gain a competitive advantage.  This translates into additional profit over less-informed companies. Optimizing might cut costs or improve efficiency, making a company more viable in its space. 

Data Mining versus Data Analytics

Too many organizations look at the data they have and try to tease out what predictable relationships can be found in that data. That runs the risk of finding spurious relationships that are ephemeral. It is far better to start with identifying the right question, only then do you ask what information will answer that question and only after that, ask what data do we need and want? It’s all about testing a hypothesis that’s based on experience, not convenience. Finally, we have arrived in an era where analytics should worry less about explaining the past and more about predicting the future accurately.

Implications for Cutting-Edge Analytics Efforts?

Are the cobblers’ kids going barefoot? Using the logic of the old proverb, it’s also in the best interest of a data analytics firm to analyze its own efforts and maybe streamline its own processes. They might create software that optimizes functions pertaining to report generation, the finding of appropriate models and methods, or the tailoring of the end report so that it is useful to the business. 

Moreover, whether the analytics operation is part of a larger organization or even as one-person shop, a big part of any streamlining effort is identifying where a data scientist’s time should be spent. We think of data scientist tasks as gathering data, shaping it, coding ways to manipulate and analyze it. But if the most important function is asking the right questions, shouldn’t that be the biggest use of precious data scientist hours? Similarly, interpreting data findings to create actionable information is also massively critical.

So what if we could begin to automate what we currently think data scientists are doing? 

Is the potential from automating Data Science bigger than we realize? 

In theory, we could teach a computer what this process should look like using intelligent learning methods, lowering and eventually eliminating dependency on human operators. The amount of time saved by this process will counterbalance the amount of time spent developing intelligent algorithms. This alone should be vastly more profitable. 

If a program can learn quickly and generate appropriate results completely on its own, then it makes sense that streamlining dramatically improves processing time. This would be a massive additional source of profit for analytics companies as they can take on much more work with the extra time.

Most important, the focus of human minds will be on asking the right questions and turning those answers into useful information.  It’s exciting to imagine the ways in which fully automated data science will transform analytics. 

Imagine… being able to ask much, much better questions.

Imagine… being able to translate data findings to action quickly.

Imagine… doing all this affordably.