What exactly is Data Science?
Jean-Cyril Schütterlé – Anyone who’s read The Adventures of Sherlock Holmes will probably remember this pithy observation by Watson: “While the individual man is an insoluble puzzle, in the aggregate he becomes a mathematical certainty”. He was talking about statistics of course, but I figure what he said is just as relevant to Data Science. It’s a discipline that uncovers probabilities that can be multiplied to provide truths. That’s on condition, though, that it’s based on enough real observations within an infinite diversity.
Technically speaking, Data Science is where statistics and computer science meet. Its main purpose is data collection, correlation and visualization of something that happens or a process. The leap it’s taken is all down to two factors: the vast amounts of data the Internet is throwing out and the advances of computational power. It means age-old statistical processes – such as Bayes’ Law of probability theory – as well as brand new approach – Neural Networks – can be applied.
Sidetrade is one company that understood very early the data exploitation challenges its clients faced. It’s why it now has a dedicated Data Scientist team who work in close collaboration with its Product Managers to meet the needs of the companies that trust it.
A key issue for Data Science is the cleanliness and quality of the data it processes. This is why Sidetrade’s Data Scientist team started with formulation of an aggregation method to align the customer data stored on suppliers’ platform under different identifiers. With this goal now achieved, we have a unique 360° view of every customer company.
A further issue for Data Science is the identification of the relevant algorithms for analysis. In Sidetrade’s case, these relate to purchase and payment behavior. These then have to be validated for their reliability for automation and integration in business processes. Sidetrade Payment Intelligence (SPi) solution, a predictive payment behavior score, is our first outcome. We’ve got others under development.
The overriding aim is to put Data Science to work to quickly and efficiently apply findings through appropriate vehicles. These could be a simple graphic interface, or a conversational robot, or APIs that nurture an external third-party application.
Data Scientists, of course, have to work closely with IT development teams to guarantee the usability of any solution once it’s in production.
What are the advantages of the ‘data-driven’ model for cloud software users?
J-C.S. – Data Science lets us move from a very ‘cerebral’ approach of conceiving and configuring software in our data labs incorporating rules a priori to a more pragmatic data-driven method that leads to big user benefits.
Traditional software applications tend to lead to a conflict between the often simplistic operating rules that govern them and practical reality. The result is that the software has to be continually updated.
The ‘data-driven’ model, on the other hand, looks to the real world to capture as many data elements as it can to infer the software’s operating rules itself. Users can then leave the application to it to learn from their available data. Sidetrade’s users, for example, have access to client and lead engagement scenarios based on actual behavior data rather than the old predefined standardized criteria. We’re moving from an intuitive to an inductive framework. It’s pretty revolutionary stuff.
What are the take-aways in terms of customer experience?
J-C.S. – Today’s customers no longer expect systems that present their data in a static way or that don’t use intelligence to automate processes. The big challenge for Artificial Intelligence is to come up with the right recommendations for the best actions to take, the best targets to reach, and the best ways of doing all this. The combination of Data Science and Machine Learning means it can meet all these goals.
All that said, users need to adapt to these new technologies and the applications themselves need to be further developed. Overall control has to remain in the hands of the human operator, so that he can give context to machine-deduced recommendations and weigh up their relevance. No machine is infallible and human intervention is vital to identify possible exceptions to rules so the machine to learn from its mistakes and improve its judgment. There are plenty of ethical reasons too for not building ‘black boxes’ that take unjustifiable decisions.
Artificial intelligence is bringing a brave new world order to the process of collaboration and exchange. From here forward, the big change we’re about to see is that we’re going to be working less and less on machines, and more and more, with them.