I came across an good article yesterday from one of The Guardian‘s data science team that explored how many (media) organisations invest into analytics tools mostly because everyone else does, and then have a tendency to star at the pretty pictures and the flashing numbers without really having the first clue what they should do next. Author Stijn Debrouwere then gave an overview of the ideas behind data science – essentially, a scientific method of hypotheses and testing. It’s a good read, and uncovers an issue of “Huh?” that I’ve seen in many organisations not just related to areas of web and content analytics.
But the article also got me thinking about an idea that’s been bouncing around my head for a few weeks – alongside data science, the necessity for, for want of a better term, I’ll call Data Jazz.
Back in my university days, studying social science (and therein lies a debate about whether the likes of sociology are sciences or arts) I tended to take a fairly qualitative approach to my research. I was interested in being able to explore information, data and ideas in a way that enabled me to identify hypotheses that could then be tested. I was (and am still) more intrigued and motivated in the processes that generate ideas than those to prove or disprove them (I don’t think that the latter is unimportant, just I get more from the former).
In that idea generation stage, the use of data is far less structured. It’s something I’ve been finding in the past couple of years with my work on http://stampsocialCEO.com/ – that the act of investigation and research has thrown up a number of questions, but in turn has given me greater insight into a number of areas, both the ones I was expecting (the social networking habits of the FTSE100 leadership) and some maybe less so (how much turnover there is in those roles; how obscure about 15% of the FTSE100 companies are…)
This wasn’t using data to test specific hypotheses, but much more the improvisation of paths through the data (both qualitative and quantitative) that has expanded my knowledge and thinking in the area. If I’d simply set a brief for an intern to go off and investigate, providing me with a report at the end, it would have taken much less of my effort, but I would be way less the wise about the domain.
So what do we need for Data Jazz? If you know anything about the genre, you’ll know quite how broad it can be. I personally think that the best Data Jazz probably works with some constraints – the twelve bar blues or modality of people like Gil Evans-era Miles Davis. Free Data Jazz, a la Ornette Coleman, is a little too random even for me. Trad Jazz by numbers is how many organisations approach it.
Data Jazz probably also works best in small groups, with those involved able to riff off of each other – jumping on idea snippets from one player to explore and develop further. The virtuoso solo performer can produce some great material, but they are rare (with the socialCEO project, although mostly my own work, I’ve been taking great efforts to get ideas from others to keep me going).
Data Jazz requires some pretty good knowledge of your instruments – being able to understand how to query data, construct and deconstruct data, appreciate both numbers and more qualitative information that brings breadth to understanding.
So there we go – a new term for the massive lexicon of analytics – Data Jazz.