The question of how well democracy works is often on my mind lately. Recently, I was particularly disillusioned as I watched the shenanigans south of the border. Trump supporters seem determined to vote for him even if the highest star power he can attract is Scott Baio, and the Democratic National Convention appears to have meddled in the campaigns of Saunders and Clinton.
Be it the potential of a Trump presidency, the mayoral run of Rob Ford, or Brexit, I am often left wondering whether everyone should have the power to make decisions. I know, I know, democracy is an incredible privilege – and it is only perched here in my secure world that I can even ponder such blasphemy. I say it a little tongue in cheek, but it does make you shake your head. However, this post concerns another democratic movement afoot that has nothing to do with building walls or rejecting Europe. It’s referred to as Data Democratization – and, about it, I am also somewhat sceptical.
Data democratization is the idea that data should not be concentrated in the hands of a few analysts but available to all business users who can use it to make decisions; in simple terms it means all of an organization’s data should be available to all. As long as I have been in the business of customer insights (over twenty years), there have been debates about giving the business user access to data. We have debated the pros and cons of Analytic Centres of Excellence and data warehouses – versus business-unit specific data marts and analysts. And we have tossed around the whether we hire Data Scientists or buy Automated Data Mining tools for business users.
With the advent of Big Data and its tools, the discussion has started again. Instead of a centralized structured Data Warehouse, the theory is that all data could be “dumped” in a free-form “Data Lake” available to anyone with a fishing rod or net. Vendors of data visualization tools are happy to tell you that anyone can create fantastic insights from the data if you just bolt their tool onto the Lake.
I am not here to debate I.T. or data management strategy; I’ll leave that to more technical folks. I love the idea of business users having the information they need to truly make data-driven decisions; however, I must admit, as an analyst, I am a bit concerned about this democratic movement.
Even the most seasoned analysts err when generating usable insights. Analysis is difficult to design effectively and insights are often elusive. Three areas concern me most of all – the potential for inconsistency across departments, incorrect interpretation, and the possible duplication of efforts. Please let me explain.
Inconsistency
If each department or project team defines metrics in their own way, it becomes difficult to draw comparisons and understand trends. Erroneous decisions can be made with respect to budget allocation or direction if you do not understand that you are not comparing apples to apples. For example, one team may report gross sales as their revenue number and another may count on the portion that is commission (net of payments to partners). When comparing the results, management may view one more favourably that the other without clearly understanding the differences that they are looking at.
Successful organizations will need to establish standard metrics and discuss how they should be applied at the business group level to minimize the chance of inconsistency. Scorecards and KPIs that are clearly defined and available to everyone will certainly make this easier.
Incorrect design and interpretation
Making data-driven decisions is critical in today’s world. And making a business decision based on data should mean truly understanding what the data is telling you. But, although they say “the data doesn’t lie”, it can certainly mislead!
Statisticians consider things such as statistical significance and sample size when interpreting the results – and when designing test and analyses. Can you really conclude campaign A did better than campaign B just by response rates? What if the sample for campaign A was very small and the high response rate was just due to random chance? Rolling out that campaign next time may result in considerable lost opportunity and disappointment when the results are not replicated.
Business users, untrained in analysing data, may misinterpret and make costly business mistakes; and requiring them all to understand analytic techniques is time-consuming and not necessarily optimal (you may lose some great minds who are not able/willing to become experts in data analysis.)
Successful organizations may want to consider analytic advisors to support business users in their quest to truly power their decisions.
Duplication of efforts
It has been our experience that in an organization when analytics is truly decentralized, multiple departments often try to answer the same question. Giving access to all users may mean considerably more duplication as everyone tries to come up with their own answers. And when the answers don’t match from team to team, the forensics required will perhaps be considerable as one tries to piece together how each different user came up with their number to explain why they don’t match.
Looking for ways to share knowledge will help companies avoid this duplication. In addition, organizations will need to consider which questions are strategic and should be answered for the organization, not simply for a specific department’s or project’s use.
I truly hope this democracy works; but, like in any democratic institution, it will work best if the right checks and balances are put in place, clear communication is used, and politicians rely on the expertise of trusted advisors with specific knowledge.
We’d love to hear from you…. are you democratizing access to your data? What lessons have you learned along the way? How are you looking at governance in this new world?
--
About the Author
Emma Warrillow of Data Insight Group Inc. has been helping her clients tell their story for over 16 years. If you would like to know what story your data is trying to tell you please contact info@datainsightgroup.ca. This blog was first published here.
