I’ve explained in other articles that Data Science is not a replacement for other data technologies, it’s a continuation of the data analytics process.
So if Data Science is used in analytics of large sets of data, how does it differ from Business Intelligence (BI), which is also used to analyze large sets of data?
Business Intelligence Is Exploratory
At its core, a Business Intelligence system has a large set of data that uses either a specialized query language or a graphical tool which allows the user to explore data. This is one of its primary purposes. The way a BI system stores its data is for this purpose.
Data Science, on the other hand, deals with answering a question. Most of the time that’s a single question – should we do X or Y, when should I replace this unit, Is this a mutated cell, is that a new star in a galaxy image. The start of the Data Science process is defining the one or few questions you want to have answered.
Business Intelligence Often Generates More Questions
Exploring data does not always lead to an answer. We might find, for instance, that we sell more umbrellas in London on a given day, but we probably won’t find out why that happens within a BI report. You might – but more analysis is needed most of the time.
That’s where BI and Data Science meet. The BI report shows the sales, and Data Science can help answer why those sales are happening. But if you stop there, you missed a big advantage that Data Science brings.
Data Science isn’t always just about looking for causal relationships in data. It also concerns itself with what you should do about it. You can take those causal relationships, and assign an action based on the result. For instance, the analysis can trigger more umbrella sale adverts on Wednesday, or bring more salespeople to work on that Thursday, or whatever the data indicates.
Business Intelligence Is More Likely To Have a Front-End
Which brings us to the next point – in Business Intelligence, you need a way to explore the data. Most of the time this exploration is done by business functions, who don’t want to spend a great deal of time and energy learning new query languages and writing programs. They’d like a graphical canvas that allows them to quickly compare, research, link and show data in trends, spot information, and aggregated information easily.
Since Data Science generates an answer, it’s more likely that answer will result in an action. The action might be an automated response to an event or some other follow-on process, or the data is recorded in another data destination. Sometimes there’s an interface, but often there isn’t.
Business Intelligence Is a Pre-Computed Set Of Data
To create a BI system, you need to point it at a set of transactional data, potentially ingesting it and transforming it into de-normalized shapes for processing into multiple pre-defined aggregates to be queried.
In Data Science, we also bring in the data, but we most often wait as long as we can to transform or join it – and most of the time we’re not using the data for the user to browse. We use that data to train a predictive, classifying or clustering model. That model – which we can think of as a kind of specialized formula – can now take in data it has not seen, and return that prediction, classification, or cluster to the caller.
So BI and Data Science are pieces of a whole – BI allows exploratory analysis, and Data Science takes the questions BI can generate and create prescriptive guidance for it. It’s all a sequence of providing insight and action from the organization’s data.