Try this data framework for analytics advantage
Data has the power to help drive efficiencies, increase profitability, and bolster innovation.
Yet analytics skills remain outside the wheelhouse of most business leaders, who typically have little experience interpreting data from multiple sources or identifying what data sources are necessary to formulate the insights they are seeking.
Shoring up this experience gap among leadership can make all the difference in turning data-driven business aspirations into tangible and decisive actions.
With that goal in mind,
In his online course for executives titled Applied Business Analytics, Bertsimas, a professor of operations research, lays out an analytics framework that is designed to help businesspeople determine which analytics approach is best suited for their application and improve their ability to leverage big data for better business outcomes.
The framework is built upon data, models, decisions, and value, with an emphasis on decision-making. “The two protagonists in this process are data and decisions,” Bertsimas says in the introduction to the course. “Analytics leaders may understand the basics of the modeling, but it is their skillful handling of the data and the decisions that gives them an edge.”
The framework begins with data, which Bertsimas calls “the most important aspect” of analytics. “Finding the right data, cleaning it, and shaping it so that it works toward your need is a skill that takes experience and understanding,” he says.
Next, organizations use the data they have gathered to decide which models to use; the outcome of those models influences decisions around onboarding a data team, choosing the best tools, selecting the right variables, and outputting effective visuals — all of which should ultimately result in value to the organization.
As an example, here’s how the data analytics framework would look when used by an equity firm — let’s call it Acme Equity — planning to make a large investment in a growth firm it has identified. Elements would include:
Data. Acme Equity needs to collect relevant data on the earnings of the target firm, including the consistency, repeatability, and quality of the earnings.
Models. To understand how those earnings may grow, Acme needs to develop a predictive model that encapsulates all factors affecting earnings and considers past situations. There’s also the question of choosing the appropriate model, which in this scenario could be linear regression, a machine-learning tree-based model, or others.
Decisions. This stage is all about evaluating risk and making informed decisions based on the outcomes from various models, which should be able to predict the drivers that will lead to high or low revenue.
Value. By this stage, Acme Equity should be able to set clear expectations about the potential value of investing in its target company, including the level of return from a potential investment and the ability to maintain a relatively low risk profile.
Central to any business analytics initiative are algorithms, which Bertsimas defines as “a set of guidelines that describe how to perform a particular task.”
For business leaders still struggling to separate logistic regression from integer optimization (both covered in the class), it’s beneficial to think of algorithmic options as three buckets:
Determining which algorithm is the best fit requires an understanding of the type of data and how it’s organized, Bertsimas says.
Structured data is organized in rows and columns, where rows represent data points and columns represent a variable/feature of the data. Unstructured data is any data not organized in that way — for example, voice, video, text, or tweets.
Unstructured data eventually needs to be converted to structured data to ensure a complete and integrated data set — a task that is much on business leaders’ minds.
