Data Science versus Business Intelligence
Data is everything and everywhere.
Imagine that you leave your house in the morning and you start wondering which route to take to work: That simple decision based on a quick estimation of time and distance is data related. Even from a straightforward commute, we can see that decisions, either driven or influenced by data, are made unconsciously in our day to day lives.
Over the past 4 years, I’ve taken part in data strategic initiatives for leading international firms across various industries. Nowadays, I look for opportunities in businesses, assessing how data can enable our clients to do better.
It is common knowledge that in the second half of 2020 companies are still not seizing the strategic potential of data and, according to MIT market research, less than 5% of companies use it well enough to gain a competitive edge. I have seen at first-hand how companies manage these projects and the common key challenge relates to stakeholders holding diverse points of view for some basic concepts. Everybody in today’s organisations has heard or talked about the terms Business Intelligence (BI) and Data Science, but not many have given proper thought to their meaning. In these circumstances, interactions during the commercial, or operational phase tend to cause frustration, misalignment of expectations and failure. A mistake that organisations clearly make is underinvesting in organisational culture and mindset change, which is fundamental if individuals are not to overlook the potential of adopting data-driven mechanisms at every layer in the organisation.
The root cause for this lack of success seems to be a misunderstanding of buzzwords.
After coming to this conclusion, I felt that it could be educational to go back to the fundamentals of BI and Data Science and shed some light on how they compare to each other.
What is commonly understood as Business Intelligence?
Business Intelligence (BI) is a generic term that dates back to 1989, whose original definition was along the lines of “mechanisms and the underlying technology to improve business decisions”.
Nowadays it is understood as the development of dashboards, digital reports or ad-hoc analytical visualisations. This could be basic KPIs digitally displayed or advanced analytical methods based on statistical models, either way with some governance and security around insights delivery.
Regardless of methods or technology, BI aims to provide bullet-proof facts for informed tactical or strategic decision making; the priority being to provide actionable information to the hands of management, allowing them to call the shots and act quickly on patterns or insights. Concisely put, BI aims to describe past events using data emanating from the business regardless of whether it comes from marketing, sales or operations.
Ok, so what is Data Science about then?
Firstly, it is considered a science because it aims to discover the unknown using methodical research and analytical techniques. It is about explaining and predicting events using a combination of mathematics, statistics, computer science and business knowledge within a domain-specific context.
The associated job title (Data Scientist) has been around since 2008 when it was introduced by D.J. Patil, and Jeff Hammerbacher at LinkedIn and Facebook respectively. It started to be a fancy trend when in 2015 The White House announced the first Chief Data Scientist and now demand for talent outpaces supply.
Recently, Harvard Business Review even called it one of “the sexiest jobs of the 21st century” and some articles claim that data scientists are the new investment bankers, but it requires a vast set of skills that are hard to combine. Some of these bright individuals hold PhDs in “exotic” fields like biomedicine or astronomy, but the majority have an academic background in computer science, maths or physics.
The main goal of a data scientist is not model building in Python but rather producing insights that allow tackling business challenges. Their role is about shaping large quantities of messy and disparate data sets to make their analysis possible, developing modelling and prediction mechanisms, to make conclusions about what happened, why it happened, and what is likely to happen next.
Data Science and BI: commonalities and differences
In common: Both practices provide fact-based insights for motivating, easing, and supporting business decisions, but these practices tend to focus on different temporalities. Also, both approaches require a visualisation layer, data management and governance.
Differences: In BI, it comes down to a validated formula or a known method of calculating a KPI. BI provides a reporting mechanism for showing updated values of previously known metrics, dealing with predictable “known unknowns”. BI mostly assists with descriptive analytics. On the contrary, in Data Science, the business comes with questions that have never been asked nor answered before, so Data Science deals with “unknown unknowns”. Data Science, as a field of automated statistics in the form of models, is able to go further than descriptive analytics, enabling future prediction and aiding in classifying and predicting outcomes.
Business Intelligence mostly assists with descriptive analytics. On the contrary, in Data Science, the business comes with questions that have never been asked nor answered before
In essence, BI is about interpreting and visualising data, whereas Data Science is about using statistics and other analytical tools to explain why something may have happened and to forecast what is likely to occur next.
Conclusion
Data Science is not a newer form of BI but both are critical milestones for any organisation that aspires to be data-driven.
Business Intelligence maturity fits well within a Data Science roadmap as a preliminary step to predictive analytics. If you really think about it, you must first understand and analyse past data and extract insights, to later build models that allow you to predict the future of your business.
However, the typical BI project framework is not applicable on Data Science projects because the latter demand specific operational requirements and the typical IT, Software as a Service, plug-and-play+configuration, mentality must go out the window.
In Bedrock, we strive to pave the way for the democratisation of Data Science and AI (in the form of Machine Learning) and our approach is built upon close collaboration between our data team and our clients’ business domain experts, counting on the appropriate support from management. This approach is what guarantees that our projects deliver tangible results.
Bonus: Many BI specialists’ teams have rebranded themselves as Data Scientists and this leads to confusion. Traditional Data Analytics differs from Data Science. Yes, some tasks overlap almost half of the time such as data wrangling, crunching, exploratory analysis and data visualisation. However, the difference resides in coding, modelling and in using algorithms which is why Data Scientists mostly use Python or R, developing models for correlation, causation, and counterfactuals. Being able to use SQL, PowerBI or Tableau is only a tiny fraction of what is required.
