The Art and Science of Data Storytelling

• Data Analytics

Let me start this post with an interesting conversation I had with a Head of Supply Chain Operations a couple of years ago.

During the conversation, this executive shared his ideal Monday morning where, he comes to the office @ 9:00 AM, opens his mailbox to find out just one email that talks about, what kind of exceptions happened in the last 24-48 hours, what were the contributing factors behind it, what the coming week looks like, and what are the 3 prescriptive recommendations for him. This entire set of Insights should be articulated concisely and crisply in the email body with no PDFs, PPTs, and XLS as attachments. To have this Monday morning, he was ready to invest anything!

For the past two decades, operational reports and interactive dashboards have been considered the ideal way to communicate and understand business intelligence from raw data. And even though they are still prevalent and offer value to businesses, there has been a shift in data, analytics models, and user expectations.

The future of business intelligence is dependent on the ability to tell stories with data. Data on its own does not convey a lot of meaning. But it is the context behind the numbers that help us understand the business implications of the data.

This article will talk about data storytelling, the constructs of a good data story, and the dawn of automated data storytelling using augmented analytics.

What is Data Storytelling?

To put it simply, data storytelling or data-driven storytelling is the ability to understand data and communicate the insights, through an amalgamation of visual and textual narratives. Where the insights are woven into a discernable story, offering valuable information and context.

Traditionally, organizations relied on standard reports and dashboards, packaged and circulated on a daily, weekly or monthly basis to narrate the business story to decision-makers. While in reality, it only provided the metrics on a dashboard, making it difficult to capture the “why” behind the data and to address the anomalies, often failing to inspire the audience into action.

On the other hand, a text-based reporting style enabled through informal mails, chats, long-form “big picture” narratives with visuals, make it easier for non-expert decision-makers at every level to derive the true value from the data.

What are the constructs of a good data story?

This art of revelation and delivering impact is at the foundation of good storytelling.

Let me talk about the top 6 constructs of a good data story and demonstrate both the art and science behind it.

1. Empathize with your Target Personas

Being relevant and humane is the key here. As an analyst, if we look at the semantics behind the term Persona, it has always been perceived as Individualistic; however, in the real world, it’s often a fusion of Business Function Specific and Individual-focused, and that brings confusion during the storytelling process.

For example, if you are setting up a Dashboard/Scorecard for a B2B Marketing team, are you going to stitch together your story keeping the department’s needs in mind or, will you focus exclusively on the CMO role and try to be strategic?

This differentiation is of utmost importance while designing the prototype of your dashboard/scorecard because, based on that, you would be shortlisting your measures, dimensions, data filters, default date/time granularity, and most importantly, the necessary column and row-level data security. In the case of a CMO Dashboard for a B2B marketing team, the focus would be on the following key measures:

• Marketing ROI
• Average Revenue per Account
• Customer LTV
• Channel Cost per Acquisition
• LTV to CPA Ratio
• Traffic to Lead Conversion %
• Lead to SQL Conversion %
• SQL Closed Won Conversion %

If your story is targeting the Individual (CMO) aspect of the Persona then, you need to be crystal clear about the Roles and Responsibilities of that Individual, their Department, their Weekly/ Monthly/ Quarterly KPIs, and most importantly, what is their definition of Actionability. Remember, by the end of the day, you want your Persona to trigger an optimization in collaboration with their team. If there is no trigger, then the story is just a GOOD-TO-KNOW factual observation.

2. Bring Context to your Story

Being contextual means enriching your story in the dashboard with relativity, combining the core KPIs with adjacent measures, designing narratives by smartly positioning your KPIs, respecting the time frame of the ongoing fiscal year, and factoring in the feedback shared during previous insights consumption.

Context is what lends power to a data story, transforming it from a collection of data points to a comprehendible narrative that drives business decisions.

For example, in the above CMO dashboard, if we have to bring the Context then, we need to make the following changes to our story:

• Compare the KPIs with the immediately preceding period as well as the same period last fiscal year.
• Associating the core KPIs like LTV with metrics like Cost of Servicing per Account, Gross Margin per Account, Churn Probability per Account, Share of Wallet per Account, and Monthly/Quarterly Target Attainment per Account.
• Now, when we publish the narratives against our charts in the dashboard, we need to be cautious about how we are arranging the metrics in our sentences because certain metrics are ratios while the others are independent. On top of that, specific metrics have cause-n-effect relationships among them, so; we need to smartly position our metrics in the narratives to make them easier and logical to consume.
• Time-frame is the most critical aspect to bringing the Context into the story because the CMO has to confront monthly and quarterly adjustments to the measures, fiscal ending, budget cuts during the recession, holiday season, demand, and supply shock months/quarters, etc. All these moments during the financial year need to be captured to maintain the relevancy of your overall story.
• Your target personas would always have feedback against your published story, and it’s imperative how you consider their feedback to optimize your tabular data model, underlying machine learning models, and the dashboard layout.

3. Respect the Hyper-Connected World of your Personas

In this hyper-connected world, your target personas are consuming the insights through multiple devices having a varied screen sizes. Some of these consumption mediums have both push-n-pull-effect, while others have only push-effect.

For example, Enterprise BI Platforms (Tableau/ PowerBI/ Qlik), Chatbots, Voicebots, WhatsApp, and Search are the mediums that facilitate interactivity with the Personas and build on the story collaboratively. However, Email Notifications, SMS Notifications, PPTs, DOCs, and PDFs have the push effect only because the objective is to push the insights to a broader audience with no opportunity for a real-time closed feedback loop. So, it’s essential to understand the traits of each one of these mediums before we publish the insights.

If you are pushing your story to the mediums which have both push-n-pull effects, you can’t be Verbose. You need to focus more on articulating the facts and providing quick diagnostic insights because, both your canvas and your attention span are short. In the case of mediums that have push-effect only, you have an opportunity to be verbose but, the focus should be more on anomalies, causal factors, and making smarter predictions.

Based on our experience working with large enterprise clients, we’d say try to be less tedious, and always have an executive summary associated with your dashboard.

4. Fusion of Facts, Anomalies, Causal Factors, and Time-Series Forecasts

For your executive dashboard to be actionable, there has to be a smarter fusion of Facts, Anomalies, Causal factors, and Time-Series Forecasts.

In the case of Anomalies, call out both the point and contextual anomalies in your dashboard. To expedite the Action TAT, classify your Anomalies as Expected vs. Un-Expected.

Try to articulate the causal factors behind your Anomalies in the form of narratives that should be associated with your trend-line or box-and-whisker plots.

In the case of Forecasts, always combine them with actuals for variance analysis. To make your forecasts more actionable, explain the variance in the form of narrative insights.

5. Smarter Data Filters

As soon as the executives start consuming the story, the very first action they take is to filter the default view of the dashboard storyline by applying different filters. Some of the fundamental filters that should be a part of every dashboard are Time Period, Market/Country, Channel, Audience Segments, Category, Brands and so on.

Here are some essential tips for deciding what filters to choose for your dashboard:

• Pick those dimensions as your filters, which have a lower cardinality. Based on our experience, the higher cardinality filters are rarely used, and can also negatively impact the load time of your dashboard.
• Time-Period granularity is the most critical filter, so while setting up this filter, you should adhere to your Persona needs. If your Persona has weekly and quarterly goals, then your Time-Period filter should have values like Current Week, Last Week, Last X Weeks, Current Quarter, Last Quarter, and Last X Quarters.
• Don’t add more than 7-8 filters per dashboard because more filters negatively impact the load time of your dashboard.

6. Robust UI/UX Capabilities

Last but not the least, your dashboard’s success in clearly constructing and communicating a data story, is largely dependent on the platform’s UI/UX.

The UI/UX capabilities should render your dashboard customizable and instinctive, with the ability to identify trends across time, with visualizations simplifying the business intelligence being generated. The interface needs to offer widgets that help users navigate, and the experience should be user-friendly. Moreover, you need to ensure that there is consistency in labels and colors being used.

If the UI/UX design is not up to the mark, there are a few risks that you are exposed to. Quite simply, if the UX is underwhelming users may avoid using the platform altogether. And if your BI dashboard is not being used simply because of poor design, your competitors will gain a competitive advantage based on your platform’s inefficiency.

The following are a few pointers to keep in mind while designing the UI/UX elements of your BI dashboard –

• Have a defined goal for the platform’s analytics capabilities
• Design for consistent UX, and uniform UI
• Prototype and text your platform for better product development
• Use simple and comprehendible terminology
• Keep the dashboard neat to help focus on the BI
• Display basic insights before a deep-dive into the data
• Display insights based on priority
• Ensure scalability, customization options, and collaboration

A consistent UI ensures quality UX. Superior UX leads to smart business decisions, which in turn supports a flourishing business.

What’s next: Automated Data Storytelling with Augmented Analytics

Data is being generated with unprecedented velocity and veracity, with variety adding to the complexities of analyzing the data. Enterprises cannot hope that their analysts and data scientists will be able to manually combine the data with a narrative, to deliver contextual insights to decision-makers in real-time.

Thankfully, the new breed of BI tools powered by AI and Natural Language Generation (NLG) capabilities, streamline the access to data across various levels of business personnel, and querying capabilities enable end-users to simply ask questions of the data and receive actionable insights in near real-time.

This is where data storytelling and augmented analytics intersect.

Narrative-building features of augmented analytics tools are programmed to simplify the data into a narrative form through automated data analysis. Providing contextual data stories to end-users for easier consumption.

Here’s an example from our proprietary tool, Course5 Discovery (more on it later), of insights served in narrative format forming a data story for the given persona.

What Problems do the Automated Data Stories Solve? 

Automation captures a greater number of insights than is even noticed by human analysts. Humans are unconsciously or sub-consciously subject to bias. What one analyst might find to be a valuable insight, may go unnoticed by another analyst looking at the same data set. Data storytelling enables enterprises to avoid human error and bias, making for a more holistic data story capturing all relevant insights.

Secondly, there are limitations to manually building data stories from data analytics, simply because of the human effort going into it and the varying levels of data literacy among end-users. End-users of self-service analytics platforms are not necessarily experienced analysts or data scientists, and they might face difficulties in understanding the data, interpreting the insights, and creating a data story that captures the correct business intelligence. Automation takes care of the pitfalls and delivers actionable BI every time.

Scaling the technology across the business to encompass every business function is a difficult task for manual analysis and storytelling, especially since the skillset is reserved for select persons. Covering different business functions and analyzing them individually is time-consuming, and effort-intensive, and it is almost impossible for a group of analysts to cater to an entire organization.

Automated data stories ensure accuracy in analysis, find meaning in metrics, and maintain clarity in communicating insights across the organization.

Let’s take a look at a real-life application of automation enabling access to actionable insights.

Get in touch to learn more about our Data Storytelling capabilities, and how we collaborate with our clients in generating relevant, contextual, actionable, and human-friendly insights through our AI-driven analytics platform Course5 Discovery.

Note: This article was recently updated to offer deeper insights into Data Storytelling, and how automation is becoming indispensable to Business Intelligence.