Storytelling has been a buzzword in the data analytics/BI vertical for decades. However, very few products are able to really make a home run in this direction, either due to the lack of focus on the personas who would benefit from the storytelling or fail on delivering the experience it really requires. This article will dive into both aspects and discuss who the storytelling experience is really for and what it should do.
Who is it for?
It’s very important to identify the user persona before talking about what the experience should be. Data insights are used to inform decision-making, so clearly the decision-makers are the target audience for the storytelling.
Many products mistakenly assume the Data Analysts as the primary audience for storytelling. While they are indeed the main user for data analytics products, they are really the story maker instead of the consumer. In other words, storytelling is their deliverable instead of what helps them to deliver. With the Data Analyst as the user persona, the “storytelling” product evolves into a tool for Data Analysts to discover stories. The Data Analyst is still the one who needs to discover and tell the story. In that sense, Excel and PowerPoint can also be marketed as data storytelling tools, which is clearly not the case.
What should it do?
With decision-makers as the targeted persona, the solution to their problem is to inform them with useful insights given a data set. Therefore, it’s critical for the solution to generate good quality and useful stories. This process in the past was mostly driven by humans with a package that includes a Data Analyst, a “BI” tool, and PowerPoint. Not even to say it’s a very expensive package, the consistency in quality and efficiency is very hard to control. At the time where automation was a challenge, this might be the only way to go. But, at the age where automation technology is wildly available, human-discovery-based storytelling application becomes a solution that creates additional problems. While the argument for the only upside to this solution package is the flexibility with the human intelligence, if we exam closely, the workflow in each industry verticle is actually fairly mechanical. Not only it makes the problem fairly easily solvable by automation algorithms, with the algorithm comparing and finding all the possible interrelations between data points, it also produces better and wider ranges of stories consistently. Refocusing on our targeted user, an automated storytelling solution not only saves them money and effort but also democratizes the data from the Data Analysts to the end-users of the insights.
A storyteller has to produce stories that enhance the target persona’s Jobs To Be Done(JTBD). Usually, the stories are categorized into two different territories: opportunities and threats. The opportunity stories should present the evidence to help improve the north start metrics for the target persona. For sales, it can be the evidence showing their product’s successful track record for the pitch. For product management, it can be features with the most interest. For operation, it can be the workflow with the lowest efficiency. The threat stories, on the other hand, point out abnormal activities that can damage the north star metric. Usually, it reports a sudden change in the data set or a certain data point is entering a critical level. Many monitoring systems in different industry verticles are doing a good job telling threat stories. A good storytelling application should first identify the target persona’s JTBD and north star metrics and working backward for the storytelling.
People say a picture is worth a thousand words. A picture may be a nice aid for a thousand-word essay, but apart from a nicely-done chart, data storytelling also needs a point. Soley presenting a story with charts and graphs will only make the audience lost on the point it’s trying to make.
Here are two concepts to clarify: data reporting and data storytelling. These two keywords are being abused for online marketing. Different from the data storytelling mentioned above, data reporting doesn’t need to have any point. It’s simply a presentation with cold numbers that are up for interpretation. It would be fine for data reporting to have charts without a narrative.
Coming back to the mistake without narratives. It sounds like such an easy thing to avoid, but we can still find the market flooded with charts that have titles like “Sales vs Time Chart”. Therefore, a good narrative is critical for data storytelling. A good story narrative has to reveal comparison information and trigger actions. It can be the comparison between different cohorts or a benchmark. Comparison gives meaning to the data point. Actions on the other hand give the “so what” to the story. Without a suggested action, the story is still open-ended with many possible ways to execute on the meaning derived from the comparison. Action closes the loop as a data story so that a complete data story will sound like:
The data shows [data points] compare to [data points] so that we should do [action] to improve [JTBD].
Sounds familiar? Yes, the user story. All the stories are a close loop with goals, reasons, and actions.
The opposite of the consumption-based UX is the input-based experience where users have to fill out a “survey” of necessary parameters for the application to generate any content. This approach assumes users always know what they want, but the truth is usually the opposite. For decision-makers who are usually none data-savvy, the experience feels exactly like the command line in the terminal window to an average PC user. This input-first design creates a huge UX gap for the storyteller’s targeted user. Consumption-based UX on the other hand generates contents first and directly presents them to users without requiring any inputs. All users have to do is to “shop” the contents that are presented to them as if they are scrolling through a news app. Moving the interaction from pre-content to post-content can dramatically improve the user experience and usability of the software, which also results in wider adoption and better user retention. Data analytics applications in B2B space are usually in the none critical workflow of the operation, so adoption becomes critical for survival. Furthermore, a more fine-tuned default input based on a common selection logic study for the consumption-based experience can also largely reduce the noise from the stories produced. When a large number of stories are found, a recommendation/prioritization algorithm can also be very handy to improve the relevancy.
Keep user notified and hooked
As mentioned above, data storytelling is usually not in the critical workflow because of the nature that data insights are usually post-facts, so keeping users engaged makes a huge challenge to the product’s success. Similar to the news app, with the timeliness value attaches to most of the data story, it’s important to keep users notified of all data stories soon after discovery. To further enhance user retention, based on The Hook Model, we can add more investment features that allow users to further fine-tune the story production and recommendation preference so that we can: trigger the users by notification; let users take action to open the app; reward users with useful stories; allows user to invest to fine-tune their preference.
AI in storytelling
Data storytelling doesn’t require AI. Logic-based automation can already handle a sizable story discovery. However, with the aid of AI, Data storytelling can be dramatically enhanced. Here are, but not limited to, 4 aspects that can be helped by AI.
- AI can help with input parameters selection for the story discovery. Without AI, the input can only be a rigid configuration set by humans. However, in a more dynamic setting, humans don’t always know the best input for the stories they are trying to find.
- AI can help with advanced story production such as predictive and intercorrelation discovery.
- AI can help to recommend relevant stories when a large set of stories are found.
- AI can help to make storytelling conversational through an answer bot.
At the end
As data becomes increasingly important to fuel the right decision, a booming amount of data analytics products with various quality started to flood into the market. With marketers abuse the keywords to sell the products that don’t fit the claim, I hope this article can help the shoppers to better evaluate a product, and product makers to make better products to fit their claim.