<img height="1" width="1" style="display:none" src="https://www.facebook.com/tr?id=974250883405448&amp;ev=PageView&amp;noscript=1">
Data Science Process: 5 tips for effective data storytelling

Data Intelligence

Data Science Process: 5 tips for effective data storytelling

Jaime Perez Cuadrado | Jul 04, 2019

We are at the last step of our series of publications about Data Science Process. Finally, we have the results of our investigations, we have worked with a lot of data, we have loaded, processed, transformed and analyzed them, developed models using a set of variables to get the solution of the problem we were trying to solve.  It’s time to show our solution and to explain, especially to a business users audience, the reasoning that we have followed to reach the target and what’s in it for them.

This is one of the most difficult moments that you will face, you must set the “right tone” of your communication, you must be clear and concise otherwise all the effort put in the previous process steps will be wasted. I will try to give you some advice when presenting your work.

#1. Recall the original objective.

It’s a very common situation, when you’re immersed in all kinds of operational tasks and the time is going short it could be that you become so abstracted that when you explain the solution, you cannot link it with the original problem.

My first advice is to rewind a bit, reorder your ideas by putting the original problem at center and then going down to the details that made you follow one path or another, to select one solution or another. It looks like mind-maps? Well, they could be helpful to collect and sketch ideas in a “free flow” approach and then decide what to keep for your storytelling.

#2. Data can save your life, or kill you.

Should be clear now, the cornerstone where your house (solution) is based on are DATA, they play the main role in your story and that’s why they have to be unassailable. Evaluate the accuracy of the data and your ability to clearly explain them; contradictory data, unreliable sources, obscure transformations will mine the overall credibility of your story.

Do you have perfect data? It’s not enough, if explaining them is too complex and hard to understand for your audience (even if they are right), evaluate alternative options to express your concepts. More true than ever, in storytelling less is more.

#3. Talk the lingo

This is a classic pitfall; your Data Science Team is technically super-strong, during coffee breaks they talk in Python each other but they struggle to create “common ground” with business users. We already discussed in other posts how a Notebook, if correctly leveraged, can be beneficial in closing the gap between scientists and end users community but at this step it could be not enough, here we’re preparing the final presentation, the meeting with key stakeholders, it’s time for the silver bullet.

Best storytellers are the ones with functional background, they do not run the risk of becoming too technical (because they can’t) and they usually have (or quickly learn) the business vocabulary of the audience they have to present to.

Use a high level language close to the business, that is, do not talk about fields of a table that nobody knows or Python libraries your audience ignore. Talk about entities and business tasks, talk about business concepts, try to translate data into business terminology.

Omit unnecessary details; technicalities are always there, in your back-pocket, and you can satisfy your techy ego by showcasing them…just if and when somebody explicitly ask for them.

#4. An image is worth more than…

Think about your youth, the very first books were made of images only, then you moved to comics and only years later you faced text-only books; reason is simple, humans deal better with images than words.

Try to be as much visual as possible; in a process that is fully data-driven this shouldn’t be a complex task, as of today there are so many “visual tools” that you could use to support your storytelling, finding the right one (for you and your audience) it’s often a matter of practice. Most of the tools your team have worked with during the Data Science Process (e.g. Python, R-Studio, MatLab, Octave, …) have some presentation capabilities but limited to tasks they’re designed for, that means mainly manipulation, exploration and research tasks.

Your storyteller (remember, it’s quite probable he/she is not a techy guru) can be much more effective by using one of the many data visualization solutions available, more or less all of these have built-in functionalities to support story-telling. Tools like Tableau, Qlik, Power BI, SAS Visual Analytics, Oracle Data Visualization help the user to populate canvas by selecting reports, dashboards, high impact visualization frames and adding to them annotations, explanations, multimedia contributions and everything you could need to make your presentation a memorable one (obviously all of them can export contents to Office tools, first of all Powerpoint).

More than this, most of these tools can be extended by importing or linking external capabilities to address very specific needs; just to name a few, D3.js, Dygraphs, or JIT can be used to enhance graphic effects while CartoDB, Mapbox, Google Maps can be beneficial when you want to plot your data onto maps.

All this stuff tends to be cool at first sight so your storyteller could be tempted to “over-perform”; use them, don’t abuse them. Breath and repeat after me: “In communication, less is more”. To show only one graph at a time is a best practice: it doesn’t confuse the audience and supports the storyteller to clearly explain what is shown and what it means in the broader context of the problem being addressed.

#5. A good story has a beginning and an end.

You and your team have spent four to six months implementing this brilliant Machine Learning project, you followed the well defined steps of an agile methodology, you went through waves… essentially what you have to tell is the story of a journey.

The communication should consist of a brief, clear and simple description of the legs of the journey towards the solution.

Basically, you need to explain:

  •        The reasons for the journey - the business problems that led to the project.
  •        The starting point – as-is scenario, how things were managed before the project.
  •        The route – overview of main steps, explanation of the decisions you took in front of crossroads.
  •        The beauty of the arrival – the benefits you provide to the business users.

Ok, we’re now close to the close, just let me share the last, probably most relevant tip: for every story, for every audience there are many tools and techniques that can really help you in performing this critical task but remember that the storyteller itself is by far the most important element of the presentation.  


Interested in learning more?

Are you ready to start deploying a more scientific, data-driven strategy to your business operations? View our services and solutions for data intelligence.