At a recent customer meeting, the top manager predicted that in the coming months, the most important IT projects will be the "small speed boats". He is the CIO of one of the five largest insurers in Germany, and he was talking about agile projects with agile teams. The prerequisite is appropriate software to allow such agility. And at this point, we are talking about the excessive use of open source in all industries.
Agility is a key goal. Mainframe and operating systems like z / OS are taken down or dismantled, replaced by rapidly-available open-source tools, just pulled down from the cloud. Companies are using them in particular for analytics.
Is this just hype? Or are we seeing the complete re-design of analytical approaches, in all sectors?
The truth is that rapid-access open-source technologies such as R are not really changing the analytical world. Instead, what is changing is the way analysts work with these quickly accessible software tools.
Will costs trigger new learning curves?
In 2015, one of the largest mobile telephony providers in Germany introduced Hadoop as the platform on which it stored consumption data for mobile devices. At the same time, the company also launched easy analysis of the data using supposedly-free open source analytics tools.
Why "supposedly"? Well, the IT department was buckling under the strain of managing high-frequency release renewals for each piece of open-source technology. This was certainly not without cost.
This sector is probably ahead of the curve. It is no coincidence that many banks are currently recruiting Chief Digital Officers from telecoms companies. Telecoms companies recognized the challenges of dealing with mass data very early. That was probably an important driver of digitalization in that sector, long before the financial sector saw the writing on the wall.
Analysis using open source tools has now opened up with solutions from established providers like SAS. But both established analytics users and latecomers to the party will recognize that initial savings in software acquisition can be lost over time through a massive increase in the need for care and a much more complex development process.
A new invention of mathematical software
The content reveals more about the old industrialized world of software platforms and the transition to small software ‘speedboats’ and short-term analytical successes. It shows that there may be very few differences between a traditional data mining tool, developed, grown and tested over 20 years, and a new open source data mining tool. We see point and click surfaces with nodes, which make dedicated functions possible via a mouse. In the Open Source Community, we see programming surfaces like the editor VI, which can be used to quickly develop and test languages like Python.
An editor, a language, mathematical functions: let's go! As a manager, I can generate results quickly, discuss them with my team - and then start again if necessary.
Speedboats only work when they can run smoothly
All these facets of new professional analytical processes and application scenarios have one thing in common. Operationalization plays a key role in all the newly-developed algorithms, whether self-learning machines or ordinary data mining..
So how do we use the dozens of speedboats, all occupied with different high-level data scientists and all going in different directions? The challenge for IT managers is to put the insights gained from this process into the reality of a company process, and make them productive.
And what about the challenge of delivering in a timely and error-free way? And so we reach full circle on agility and industrialization, or, as we call it, operationalization. I need professional model management for my mathematical functions, traceability of my analyses and, of course, to make my information gains freely available to other departments. Doing so means that they can also use the knowledge gained in the initial agile phase of the project productively for the benefit of the company.