How To Build Compelling Stories From Your Data Sets

Every number has a story. As a data scientist, you have the incredible job of digging in and analyzing massive sets of numbers to find what that story is. The challenge can be that while you may have an artistic bent, you may not know how to turn that beautiful visualization into something more meaningful. Is it even possible?

Even the most mundane datasets can be compelling to an audience; it’s simply a matter of presentation. This post will aim to guide you through just how you can make a statistical analysis into a compelling narrative.

Visualization Is Your Friend

From the start, visualization is already helping you to make a compelling story—so you’re starting from a winning standpoint. In fact, one study shows that people who use visual aids in presentations are 43% more persuasive in their arguments.

Now, your job is to take that visualization and make it something that’s truly compelling. To do that, we’re going to focus less on the actual visualization, and more of what’s behind it: a well-crafted story.

Create a Narrative

Whatever the dataset you’re visualizing, there’s a story that comes out of it. This can be represented in something as simple as the change over time—what is important to realize is that it’s not just numbers. The visualization isn’t simply a representation of the numbers; it’s representing a point in a larger narrative. You just need to figure out exactly what that narrative is.

Narrative Structure 101: Every Story Needs Conflict

Based on this interview from The Atlantic, it becomes clear very quickly that a compelling story hinges on conflict. There needs to be some sort of tension in the story. While that might not play out in terms of “character development” or a plot arc, there is still a way to convey this tension—that something is wrong, or broken, or being fixed. There is significance to the data beyond it simply presenting something new.

The Different Types of Plots

According to Christopher Booker, there are seven basic plot types: overcoming the monster, rags to riches, the quest, voyage and return, comedy, tragedy, and rebirth. Most commonly, we see overcoming the monster—but we don’t get the full story. That’s the beauty of data visualization: you don’t have to tell the story, but you have to present some sort of tension that compels your audience to dive into your visualization.

In this video, surrounding U.S. gun death statistics, the monster is clearly gun violence. They do not present a solution, but rather simply show us the monster. But it’s not just the monster that makes this video compelling, they include several other narrative elements that draw the audience in.

Identifying The Narrative Elements

The five main elements of a narrative are the character, setting, conflict, plot, and theme. In the above example, it’s extremely easy to identify every single one: the characters are the victims of gun violence; the setting is the U.S.; the conflict is that they’re losing years they could have had; the plot is that every day, someone in the U.S. is losing their life to gun violence; and the theme is that gun violence in the U.S. is stealing lives.

They do not present a solution, that’s for the audience to conclude themselves, but rather than simply presenting the statistic that 9,595 people were killed because of gun violence, totally 413,342 stolen years, they used a beautiful visual presentation of each life up to the point of death and then the years that were stolen to make the numbers both tangible and significant.

Build On Your Story

The challenge for most data storytellers, however is that they’re not working with “compelling” data. You could be working with cell phone customer data in China, or consumer behavior based on eCommerce search queries. So how do you make that into something persuasive and beautiful?

Keep It Simple, Keep It Safe

The key is in simplicity and patience. Arguably the greatest teacher of non-fiction writing, William Zinsser, had a lot to say about simplicity that apply to data visualization, notably: “writing improves in direct ratio to the number of things we keep out of it that shouldn’t be there.

Here’s a great example: highway data, and what it’s costing us.

In this first chart, we see an easy to read, heatmapped map of the country, setting up the basics of our narrative. We’ve got a plot, a setting, and characters, and we’re even starting to see the beginnings of the conflict and theme: The roads in the U.S. are bad, and a lot of them need serious repairs.

In a basic conversation, highway data isn’t the most compelling thing in the world. And even then, it’s kind of a two-sentence conversation: “Yeah, the roads really suck, huh?” “Yeah, hopefully that damned government will fix them.”

Now enter the real driving point of this data story:

As it turns out, those roads aren’t just bad, they’re costly. Using the same heatmapping format, we now see what those bad roads are actually costing individual drivers. This information went from theoretical, and kind of boring, to a totally compelling story with a real conflict: every day that goes by where the roads aren’t getting fixed, it’s costing you dollars.

Start With a Kernel

Most often, you’re taking complex information and making a compelling presentation, so layer what you’re trying to say. The idea is that you have a kernel, and that kernel becomes a more complicated idea. You have to get people on board with a basic principle—in science, it’s a thesis statement.

From there you can develop the kernel, and begin to focus on “minor plot lines” and other information that in and of itself may not be compelling, but in the greater context adds value to the story. That kernel can work in two different ways.

Enhance: Start Big and Narrow In

Whether you’re using a series of visuals, a graph, a chart, or something completely new and different, you can layer the delivery of your information. The first method of layering is to put all the layers on at once, and then begin to highlight more specific, targeted areas of information predicated on the overall picture. We’ll call this the “enhance” method.

In this example from Jacob Vigdor over at Tableau, he presents an extremely full picture, and from there, allows the reader to explore different enhanced parts of the narrative that can lead them to a number of different, more specific conclusions based on the initial theme: immigration has boosted the housing wealth per homeowner in many different parts of the country.

He allows you, after seeing the full picture, to zoom in and find out how that plays out in specific parts of the country. Done in reverse, it would be much harder to identify the theme and conflict.

Snowball: Start Small and Build

The other option is to smart small and build out. By doing this, you may have a great effect on the delivery of the conflict, showing what may seemingly only be an isolated incident is actually affecting a more broader range.

This is a fantastic example, created by Ben Jones:

The gif here builds in three different stages. It starts by showing the zone in Europe which contains only “free countries.” Building out, it adds on a larger region where there are a few less- or totally not-free countries. Finally, showing the global map, continuing to lower the ratio of Free to Not-Free countries.

While these numbers might not stick out to the ordinary informed citizen as surprising, when put into a sequence that shows the contrast, and presents the reality in a straightforward visual manner, it shows just how startling the reality of the story can be.

Whatever data it is that you’re presenting, you have the ability to make it interesting. It’s a matter of discovering the conflict that’s within the numbers—taking the time in your analysis to decide not just what the conclusions are, but also the implications of the conflict for your audience.

How To Build Compelling Stories From Your Data Sets的更多相关文章

  1. 【Unity3D】生成工程报错解决—UnityEditor.HostView:OnGUI() Error building Player: Couldn't build player because of unsupported data on target platform.

    错误 错误1:An asset is marked as dont save, but is included in the build: unityEditor.HostView:OnGUI() 错 ...

  2. Working with large data sets in MySQL

    What does working with large data sets in mySQL teach you ? Of course you have to learn a lot about ...

  3. Result window is too large, from + size must be less than or equal to: [10000] but was [78440]. See the scroll api for a more efficient way to request large data sets

    {"error":{"root_cause":[{"type":"query_phase_execution_exception& ...

  4. Interviews3D: APlatform for Interactive Handing of Massive Data Sets 读后感

    横向比较: Inadequacy of current system design( 现代系统和一些软件的不足) 软件特点: Output sensitivity Out-of core data h ...

  5. machine learning data sets

    http://archive.ics.uci.edu/ml/datasets.html 例如 3 分类 鸢尾花 数据集: http://archive.ics.uci.edu/ml/datasets/ ...

  6. 最大信息系数(MIC)——Detecting Novel Associations in Large Data Sets

    本文介绍了一种发现两个随机变量之间依赖关系强度的度量MIC(最大信息系数,类似于相关系数的作用).MIC具有以下性质和优势: MIC度量具有普适性.其不仅可以发现变量间的线性函数关系,还能发现非线性函 ...

  7. My journey introducing the data build tool (dbt) in project’s analytical stacks

    转自:https://www.lantrns.co/my-journey-introducing-the-data-build-tool-dbt-in-projects-analytical-stac ...

  8. 【转】The most comprehensive Data Science learning plan for 2017

    I joined Analytics Vidhya as an intern last summer. I had no clue what was in store for me. I had be ...

  9. 深数据 - Deep Data

    暂无中文方面的信息,E文的也非常少,原文连接: A lot of great pieces have been written about the relatively recent surge in ...

随机推荐

  1. static块的本质

    在网上看到了下面的一段代码: public class Test { static { _i = 20; } public static int _i = 10; public static void ...

  2. HDU 5167 Fibonacci 筛法+乱搞

    题目链接: hdu: http://acm.hdu.edu.cn/showproblem.php?pid=5167 题意: 给你一个x,判断x能不能由斐波那契数列中的数相乘得到(一个数可以重复使用) ...

  3. HTML&CSS实体

    <!DOCTYPE html> <html> <head> <meta charset="utf-8" /> <title&g ...

  4. 1st 本周工作量及进度统计

    1. 项目:英文文章词频统计 项目类型:个人项目 项目完成情况:已完成 项目日期:2016.9.6 C(类别) C(内容) S(开始时间) ST(结束时间) I(耽误时间) △(实际时间) 分析 需求 ...

  5. (十二)Jmeter之Bean Shell的使用(一)

    一.什么是Bean Shell BeanShell是一种完全符合Java语法规范的脚本语言,并且又拥有自己的一些语法和方法; BeanShell是一种松散类型的脚本语言(这点和JS类似); BeanS ...

  6. jenkin重新注册用户

    http://www.cnblogs.com/xiao-fy/

  7. CentOS7 修改分辨率

    1. 修改文件: vi /boot/grub2/grub.cfg 2. 在linux16 开头的哪一行 增加 vga=0x341 修改为1024x768 3. 重启..

  8. sql 两列数据交换

    MSSQL的处理方法   update table1 set field_1 = field_2, field_2 = field_1 可是MySQL就不能这样写,不然一列会覆盖另一列记录   MyS ...

  9. docker+mesos+marathon

    前言 (Core) [root@docker-slave ~]# uname -r 3.10.0-229.4.2.el7.x86_64 [root@docker-slave ~]# uname -m ...

  10. ADB抓取日志和日志过滤

    建议抓取日志之前先清除缓存的log数据:adb logcat -c 网上有人介绍可以这样写:adb logcat -c && adb logcat 1.使用V.D.I.W.E.F.S优 ...