I have been involved in the setup of several data science practices in both industry and government. Here are a few key principles I use in establishing a data science practice.
Principle 1: Building a predictive enterprise is, first and foremost, about building a human infrastructure.
Many companies mistakenly believe that analytics is primarily about software and end up in a perpetual cycle of spending most of their analytics budget on the latest software that not many people in the company can utilise well or, worse, limits what they can do. This is the proverbial cart- before-the-horse scenario. A predictive enterprise places people at the top and build processes and technologies around people to enpower them.
Principle 2: Analytics is an iterative knowledge discovery process and needs to be managed as such.
Unlike most engineering endeavours that can be meticulously planned out, analytics is not a linear process. Most analytics projects are exploratory by nature and the outcomes are usually uncertain upfront. Some projects will lead to insights that have large impacts, but many others will end with little to no actionable findings. This is to be expected; after all, making mistakes is how we learn. The way to handle this is to have procedures in place to allow mistakes to be made as early and cheaply as possible. This implies many iterative short-duration (1-3 months) projects instead of a small number of large projects.
Principle 3: Analytics is not IT.
The third principle is a corollary of the first two. Analytics is primarily about the exploration, description, and modelling of data. Such activities need to be conducted within a human-centric and iterative environment that encourages creativity. This requirement is in contradiction with the IT view of the world, which is about pre-defined systems, known outcomes and large, top-down projects. Analytics cannot flourish within such an environment. The issues concerning IT in the context of analytics apply mostly in data access and the operational deployment of analytics results. In that sense, analytics is a major user of IT, but transcends all areas of the enterprise.
Principle 4: Discovery starts from asking the right questions.
We want to make sure as many people as possible are data and analytics literate, but it is futile to think that we can make a data scientist/engineer out of everyone. But every person is capable of asking the right questions inside their own business functions, and that can be as important as finding answers to those questions. The management framework of a data science practice must make it easy for people to ask questions and seek help to answer them with data analytics in a consultative manner.
In the next article, I will discuss the People dimension of setting up a data science practice.
2 thoughts on “Setting up a Data Science Practice: Fundamental Principles”