The Illusion of Data Democratisation & Self-Service
Why are we still failing?
Background
This has been troubling me of late — the topic of data democratisation and data self-service (DD & DS). Over the past decade, many organisations have pushed hard to make (appropriate) data easily available to those who need it in the form of designed dashboards, structure tables to query through drag-&-drop, or even through distributed Excel pivot tables. However, time and time again, many organisations fail to see the “intelligence lift” afforded by having data at one’s fingertips — time-to-decisioning and quality-of-decisioning didn’t move nearly enough despite these data-enabling investments.
And so I dedicate my 61st article to unpacking this phenomenon of why DD & DS may be the wrong approach for many organisations.
(I write a weekly series of articles where I call out bad thinking and bad practices in data analytics / data science which you can find here.)
A Simple Framework
I would like to use the analogy of cooking to highlight why self-service or self-help don’t often work. If I want you to be able to feed yourself, providing you with access to the ingredients is insufficient to achieve that objective. Giving you a recipe or cookbook is also insufficient because you are not going to learn unless there is a decree that says no one else is ever going to cook for you. Even if you learn to cook for yourself, it’s likely going to be an unsatisfactory meal, and you will gravitate towards cooking some simple dishes and eat that most days. The effort to behave otherwise is simply too much. The million-dollar question here is — is it reasonable to assume that people want to cook for themselves? I would argue that it is NOT. (Most) people want to satiate their hunger; they are not interested in the cooking process. This is the same issue with DD & DS. No one really wants to play with data; they just need faster and better information to get their work done.
Let’s unpack the DD & DS further. I would like to ask the following questions:
- What is the definition of DD & DS? (Don’t assume that we all have the same understanding on this matter.)
- What conditions should precede for DD & DS to make sense as a business enabling strategy to pursue?
- What conditions should proceed for DD & DS to be successfully adopted so that speed and quality of decisions have improved.
Defining DD & DS
To say that data democratisation is to make appropriate data easily available is overly simplistic. Firstly, the consumer of “data” isn’t looking for data; they are searching for information to help either improve their decisioning or to confirm some hypothesis or hunch they may have about a phenomenon of interest. I’ve written here before about the difference between data and information, so I won’t go into it. But suffice to say, information is interpreted data; i.e. data with meaning. So the business user is looking for pre-processed, transformed, combined and translated data that aligns with how information is utilised in their decisioning framework. They need it pre-cooked; they won’t do the cooking.
We should therefore redefine data democratisation to information availability. It has to be information and not data. It has to be available in a format that the business user can quickly consume and digest. Furthermore, it has to be aligned to their decisioning framework. Efforts needs to be expanded to think through the best approach to meet these objectives. The solution can take many forms, and doesn’t have to be wedded to the traditional interactive dashboards or drag-&-drop database queries.
Preceding Condition
What conditions should precede for DD & DS to make sense? Your current organisation needs to meet 2 conditions:
- Have a fair degree of data fluency.
- Be already data-driven in critical areas of decisioning and actions.
If your organisation hasn’t really been exposed to sufficient complex (static) reports or Excel pivot tables or making their own simplified reporting, pushing for DD & DS is simply setting yourself up for failure. Without data fluency, the user will struggle to turn data into useful information, and DD & DS platform will be under-utilised.
Moreover, your organisation should already have a strong and trusted working relationship with your data analytics / data science function. This would ensure that they are sufficiently exposed to utilising data to drive faster and better decisions. This would ensure that the business users have the ability to form data-supported hypotheses. Without good hypotheses, the business users would be aimlessly playing around with the DD & DS data without landing on anything insightful.
Proceeding Condition
Even if your organisation is data fluent and data-driven, it does not mean that the DD & DS investments are fully utilised. What conditions should proceed for DD & DS to be successfully adopted? Assuming you have met the pre-conditions, then your subsequent DD & DS approach is about extracting even more value from your organisation’s data, and amplifying the power of data exploration and diagnostics by having more eyes and hands working on it.
The business users must feel that it is in their interest to more fully exploit the DD & DS platform. One way to go about it (and I’ve seen this in a client organisation) is to have your business leaders “make” their teams continuously generate worthy hypotheses that can then be re-channelled to the data analytics / data science function. The basis of those hypotheses should be supported by data that can be obtained via the DD & DS platform. Business users that do so will get their request prioritised by the data analytics / data science function.
It is important to note that an effective DD & DS platform will not be designed for solutioning. It’s designed for exploration and generating hypotheses. It may be able to do some simply hypotheses validation and diagnostics, but anything deeper will be fairly unstructured, and not easy to “templatise”, and hence best served by the data analytics / data science function.
Conclusion
Data democratisation and data self-help shouldn’t be your first move when you are starting out on your data analytics maturity roadmap. You are still better served by having a strong centralised data analytics / data science function at the start, and having them partner with the business to raise the bar on exploiting data for better decisioning. Only when the bar is sufficiently raised does it make sense to then push the next envelope by having the business have more direct access to their own useful set of pre-cooked data (i.e. information). They should then be incentivised / encouraged to generate even more hypotheses that will be acted upon by the data analytics / data science function, leading towards a productive virtuous cycle.
