Let’s talk about being a data-driven organization. Becoming data-driven requires systemic change, where the data science is part of the feedback and decision loop of a business.
However, many fewer people than we’d hope know how to use and understand data. Even simple concepts such as the difference between mean and median, or causation and correlation, can be hard to convey.
So, maybe we should teach everyone about numbers? It worked for Google.
(photo credit: ansik on flickr)
Numeracy’s not enough
Even if it were possible, making everyone speak the language of numbers is not a recipe for success. Financial crashes tell us numeracy alone won’t save us. The wonderful thing about numbers is that we can use them to back up whatever plan we already decided to embark upon.
It’s still the case that to achieve a data-driven organization, one where data science drives real product and business benefit, the organizational culture needs to be more quantitatively oriented. But we cannot pursue a “you must go away and understand numbers, then we can talk” approach. Unless rigorously enforced from the top-down, that’s not going to fly—nor is it sufficient.
Something more collaborative is required, allowing communication across disciplinary boundaries. I’ve found that in software, the best way of having others understand what can be done is to have them work alongside you. You get domain expertise, and they get a feel for the medium.
Collaboration and exploration
A practical and feasible way to achieve a goal of organizational numeracy is to create social, collaborative and exploratory tools and processes, which people can use to get the fingertip feel for the data without having too learn much ahead of time. This is advantageous in two main ways:
- social and collaborative systems facilitate the two-way flow of domain and business expertise and feel for the data
- exploratory systems are able to constrain the tool to reduce the need for numeric expertise. We can create systems that guide exploration. It’s important not just to give bland reports, but to engage business users with the data by permitting experimentation and interactivity
Batteries included—and humans are too
Though useful, the phrase “data-driven” is an overstatement—it implies humans aren’t required. The ideal situation is when the data science enables and augments human capability, and there’s a collaboration between the business owners and the number crunchers.
The data-literate need not be frustrated that business doesn’t understand their language, but instead build directed tools to allow business to engage in their world. In turn, they will understand more about the business and be better able to meet its needs.