a propos

Je suis architecte d'entreprise dans une grande entreprise publique de transport (devinez !) et depuis quelques années impliqué dans des sujets data "en grand". Vous trouverez dans ce blog quelques modestes réflexions liées à mon quotidien, les problématiques et réponses que j'ai pu trouver. N'hésitez pas à interagir !


What is datademocratization ? It is the opposite of a conservative approach. It's a liberal approach. It's about allowing everybody to use the data of your organization, to make the right choice, the right decision, to build better tools (with artificial intelligence for example). Everybody is to say  all the collaborators from the CEO to field people, all our partners, ouf subcontractors.
No more gategeepers or bottlenecks.

Datademocratization is not only a subject of data, we need to teach how to understand the meaning of the data, use the methods and the tools and mainly spread knowledge.
To undestrand the meaning you need a dictionnary with a three viewpoint approach : for technical experts, for bi expert, and for newbies. What is this data, where it comes from, what are the operational usages, what is is lifecycle from it birth to is death, the transformations made by all the systems, … You also have to document how to retrieve it (data owner, platform) and how to use it (including the risks, the quality issues, some algorithms, …)
Speaking bout tools, datademocratization have to be low tech. At first, you don't need a datalake or bigdata cloud infrastructure. You can start with an excel sheet ! You can use it to store, retrieve, visualize, document, … And then you can add gradualy some tools to improve your works (a lot are open source), then some cloud to be scalable, then some automatization tools (discovery, quality, security, …).
You can spread kwoledge inside and outside your organization. Inside your company, open all your BI, data, analytics to all in intern by default, except for risky data. I will explain later what i call risky data and what we can do to reduce the risks. Then teach collaborators on tools, on methods, on datas available, by organizing workshops dashboard in a day, IA Day, data day. In my #rockthedata team, we call them "Come as your data are" sessions. Colleagues comes with their use case and their data and we help them during a day to fix their issues. We also organize data tours, learning expeditions, …
Out of the company, you have to try to share data with partners in each use case you have. I've seen subcontractors comin to us with prototypes made with our open datas and  win contracts ! You can build easily an open data platform, organize open innovation days, hackathon, … The good idea is everywhere. But it only comes to you if you share who you are : your data.

It's an angelic approach isn't it ? Obviously there are some risks and you have to manage them before sharing  your data. For each data risk, use a systemic approach, the ebios framework for example. It's an easy understanding set of guides (and a freeware) dedicated to information system risk managers.
You can add all this informations to your governance documentation to be more efficient. DMBOK is an other usefull framework dedicated to data management. 
Note also that there is always a way to desensibilize sensitive data : mask or group data, use a symetric algorithm on keys, ... Keep your data still usefull for all but unrisky.

Wim Delvoye -Concrete mixer
With datademocratization, everyone can improve his own expertize with data, with facts, and everyone can contribute   effectively in decision-making. More, your organization is able to identify new data use case out of the box.

But be carefull, datademocratization changes the place of the manager in the organization  : is not any more the only owner of the information, now he has to help the team to evolve by itself. It's viral. At end,  sharing data will break all the silos in the company.
Data specialists can be afraid by this approach. They can be if they think that they can live alone. They have to improve their communication and sharing skills. A datalab nerd  is not valuable any more.

There's no cons, only risks to manage as explained below.

Democratization is an affordable lowtech game changer. It's about making our collaborators, our partners smarter. The conservative data policy is not an option. Except if you want to die alone with your data.
But deploy data is not enough. Making a decision without data is a big risk. It's a point. But making a decision with a bad uderstanding of data or with the wrong data is a big risk too.
Cross reference data and experience with the right methods and the right tools makes good decisions, makes the thing happen.
Let's democratize !