Demystifying Data Democracy
Big Data. Agile/Scrum. Machine Learning. DevOps. Smart Algorithms. Artificial Intelligence. Business IT sure loves its fuzzy buzz-words. Some of the latest rising stars are ‘data democratisation’ and its slightly more daunting sibling ‘AI democratisation’.
What all these buzz-words have in common is that, when facing an unsuspecting audience, inserting them in a sentence will make one look smart, agile and DevOps. This makes them an invaluable tool in any consultant’s toolbox. What they also have in common is that many (of said consultants) will remain awfully vague or evasive when pressed about what it actually means. Which is a pity, because underneath all the hype, there is substance to be found.
So for the first of those new kids on the block, allow me to paint a more tangible picture; what data democratisation could actually look like. I will even do better than that. In future posts, I will even outline strategies and architectures for achieving it. The other kid, AI democratisation, is for a large part about bringing accessible tools to data, so it is natural a next step.
To understand data democratisation, we need to go back to the root of the term: democratisation.
Democratisation could be defined as the process where the power is shifted from the small elite to the many. In a broader sense, it can be seen as a process of decentralisation.
Data democratisation is not the first non-political democratisation. Not too long ago, television (also known as: that odd, hardly interactive Netflix) was a scarce commodity. People would go to theatres or group up at early-adopting neighbours to watch it. But eventually most people got their own (black-and-white) television sets. And for the first time in history, people could experience, right in their living room, events taking place hundreds of miles away, as if they happened yesterday. Yes, yesterday, because it was still a long time ago. But once again: decentralisation.
Which brings us to data democratisation. What does it look like?
Data democracy should be like switching on the television and tuning into its channels. Multiple at the same time if you so desire and either in (big) batches or streaming.
Subscribe to a channel, get your data, do your analysis and cancel the subscription. Or leave it running for recurrent analyses. Subscribe to one or more streaming channels, add a twist of your own (like, gasp, machine learning) and publish it as a channel of its own. Or subscribe to some channel to receive its data in a simple spreadsheet every day. In short: consume and publish data as your needs require.
So where do people in a data democracy so conveniently get their data from? Where do they apply for those subscriptions? Well, that is simple: straight from the source, the data producing applications. You do not want or need intermediary stations. These must be maintained and could potentially distort data — the whisper game.
A data democracy lives by the principle that if you produce and therefore own the master data, you must also provide reasonable access to interested, properly authorised consumers.
This implies a profound change in the enterprise. Like with a real democracy, it not simply building a house of parliament. Nor subsequently stuffing people in there with the label ‘representatives of the people’. Nor telling them that they are the new rulers now. The (real) people would neither understand nor acknowledge it. And without buy-in, members of the old regime would try to marginalise it or tear it down as soon as they could.
Like a real democracy, establishing a data democracy requires breaking up and redefining existing structures.
In that light, the transition into a data democracy could be seen as a revolution. But not all revolutions are violent or abrupt. Silent, grass-roots revolutions might not be flashy, they might (appear to) take some time, but they make less victims and their effects tend to be longer lasting.
