Data scientists spend days and days (if not weeks and weeks) cleaning and preparing data, building and training models, and eventually trying to come up with interesting insights or delivering valuable data products to improve overall company performance.
The thing is, what data teams are actually doing remains very obscure for a lot of business-oriented people out there including decision makers. Not only is it frustrating on a personal level because data scientists feel undervalued, but it can also (and somehow more importantly) prevent the company from actually using and taking into account data-generated insights. Which is kind of annoying. And costly.
There are probably several reasons for this misunderstanding. But most of the time it has something to do with business people not figuring out how to picture the whole data science process. They carefully listen to data scientists’ final recommendations but actually end up leaving the room feeling “in the dark”. Why? Because even though they (nicely) trust their data scientists, it’s always frustrating to accept advice without having full understanding and visibility on the “secret sauce” behind it all.
Basically, all the complex technologies and programming languages involved in the data science process make it hard for non-technical people to understand it. That’s when data visualization comes into play and allows less technical profiles to penetrate the black box and to really understand what’s really being produced in the company’s data department.
So you got it: data visualization is super-duper important. It’s basically the key deliverable allowing business people to understand data scientists’ work AND what their final recommendations are. In the end, it’s all about communication. But - believe it or not - there is something even greater than data visualization and it’s dash… wait for it… boards!
So what is a dashboard really? Dashboards are basically great places where you can pin whatever you want and organize the whole thing however you like (with sections, titles and so on). And by whatever you want what I really mean is maps, interactive graphs, forms, datasets previews, python notebooks, web apps, and so much more. How cool is that?
A Dataiku DSS basic dashboard structure for T-Shirts sales
The only thing you need to make sure of is that the information you choose to display is actually going to be useful. Either to your data scientists if they need to use the dashboard as a notebook; or to your “readers”, who might access the project dashboard and data generated insights without necessarily having to dive into the depths of datasets, recipes, and so on.
A graph visualization pinned on a Dataiku DSS dashboard representing how frequently two Marvel characters appear in comic books (if you want to know more about this project, check out Joachim’s free training about data visualization or Pierre's blogpost)
The main difference between “basic” data visualization and dashboards revolves around how frequently data is updated. While data visualizations are “just” generated from data, dashboards are regularly updated according to dataset modifications. Plus, dashboards generally tend to work with various shapes and sizes of datasets, which is a real asset.
If you’re interested in knowing more about the difference between dashboards and data visualization, you might like this article - and more generally this website - I came across during my investigation.
Dashboards not only allow business-oriented users to understand what is actually behind that huge amount of data by displaying graphs, tables and so on; but they also enable them to play with data and to easily extract what they need from it.
Simple use case: just by filling in a basic form from the dashboard with a customer id, marketing people can for instance be enabled to see whether or not this precise customer is likely to be a potential churner.
How we marketers feel when we see a dashboard
So to wrap it all up, dashboards are kind of the final step to a data science project made accessible to the company’s business oriented people.
Last but not least, dashboards are also a great way for technical people to work more efficiently together as a team. As they work on different parts of a project, data scientists can gather their insights on the project dashboard to share their results with their colleagues. That’s part of why we claim Dataiku DSS is a collaborative tool. Want to give it a go?
Try Dataiku DSS out now! The dashboard feature is limited to the Enterprise Edition but you can contact us if you’re interested in getting a free trial!
Please fill out the form below to receive the success story by email:
How can we come back to you ?