“The only thing that is constant is change” — Heraclitus
Today in the business world, we talk about change. The advent of AI is transforming our entire society. Tech companies heavily invest in R&D to stay on top of the game while older industries are trying to catch up on their digital transformation.
在当今的商业世界中，我们谈论变革。 人工智能的出现正在改变我们的整个社会。 科技公司在研发方面投入大量资金，以保持领先地位，而较早的行业则试图追赶其数字化转型。
There’s a widespread belief that companies which are successful in digital achieve their goals because they have the right internal structures in place. I am intrigued by how you can visualize these structures beyond simple org charts, and whether or not a tangible representation of them reveals insights into the way a company is managed.
As you would imagine, few corporations were willing to let me access their organizational chart under my own data art studio. My only option was to work as a collaborator in a large group to test this idea (among other things). As soon as I joined the Havas group, I got started on this pet project and was eager to discover how each business unit relates to another.
Spoiler: it turns out that the knowledge extracted from this analysis is extremely valuable and can drastically help to make informed data-driven decisions.
获取数据 (Getting the data)
Havas recently migrated to WorkDay, an ERP for finance/HR SaaS for large organizations. One of WorkDay’s key features is TalentSpace, which acts as as an internal LinkedIn-type system, through which people can expose their title, physical location, company name, ways of reaching them and their boss: an ideal dataset for us!
Havas最近迁移到WorkDay ，这是用于大型组织的财务/ HR SaaS的ERP。 TalentSpace是WorkDay的主要功能之一，它是内部LinkedIn类型的系统，人们可以通过该系统公开其头衔，地理位置，公司名称，联系方式以及与老板的联系：这是我们的理想数据集！
For those who fear leakage of private info, rest assured that as a regular employee, I could only access anonymized GDPR-compliant (no names) information with the consent of the HR department. Note that when I started working on the project not all the subsidies had migrated to the platform (hence the lower number of data points than might be expected).
As with all new data-driven projects, I started by agglomerating information. One point worth noticing is the gender balance of the organization, with more women in all four branches of the group, 56% to be precise.
Regarding the distribution of managerial functions, we can observe that the top-management is slightly balanced in favor of male executives with a turning at hierarchical level four (0 being the CEO Yannick Bolloré), which roughly corresponds to the executive committee of each individual digital agency that the group manages.
Of course, there is much more confidential information to be shown that I can’t share here, such as the distribution of job titles, the number of managers per agencies and at which level, the subsidies’ location of each agency and how they interact for instance.
When you start adding the job title in the mix, you can also anticipate the shortage/recruitment of talents you need to recruit to achieve your strategic goals.
This is just a bunch of examples but as you can imagine, it is extremely valuable and actionable info when your job is to manage and have a strategic vision of a company.
从数据可视化到数据艺术 (From data visualization to data art)
可视化层次结构(Visualizing the hierarchy)
So far we have talked about the value of collecting organizational info on a company. Let’s remember that we extracted a network connecting all employees to their direct superior. As you know, this simple one-to-one relationship does not necessarily reflect the real chain of command but it is already a good start to build a visualization.
到目前为止，我们已经讨论了收集公司组织信息的价值。 让我们记住，我们提取了一个将所有员工与其直接上级联系起来的网络。 如您所知，这种简单的一对一关系并不一定反映真实的命令链，但它已经是构建可视化的良好起点。
Here is an attempt at revealing the hierarchical levels of the group using a radial tree layout.
The dataset is too complex to be able to visualize all its dimensions in a 2D representation. As a general rule in a network visualization, proximity on the image does not necessarily imply proximity in the data. In our radial tree, we can clearly see the different hierarchical layers, but we miss a sense of quantity. Indeed, most nodes are drawn almost on top of each other, making it difficult to assess the volume of employees per branch.
数据集过于复杂，无法以2D表示形式可视化其所有维度。 作为网络可视化中的一般规则，图像上的接近度不一定意味着数据中的接近度。 在我们的放射状树中，我们可以清楚地看到不同的层次结构，但是却缺少数量感。 实际上，大多数节点几乎是彼此重叠的，因此很难评估每个分支机构的员工数量。
当dataviz不够用时 (When dataviz is not enough)
To solve this problem, we have basically two options: adding a companion data visualization (like the barchart on top) to indicate the volume per branch or per sub company or improve/modify our radial tree layout.
First, instead of having a circular shape we could put everything in line like a file system folder view. The problem with this one is that the list becomes very long and boring to look at: you probably need to scroll to see it correctly or unzoom it so far you won’t see anything.
Coming back to our radial chart, we also put more nodes on each line with no overlap and zoom out to even further away. The issue in this case, as with all angle-based visualizations, is that the human visual system has a hard time evaluating angles precisely and thus difficulty counting the number of employees. For the curious reader and avid dataviz designer, this observation led to the famous blog post: “Death to Pie charts”.
回到径向图，我们还在每条线上放置了更多的节点，没有重叠，并缩小到更远。 与所有基于角度的可视化一样，这种情况下的问题在于，人类视觉系统很难精确评估角度，因此难以计算员工人数。 对于好奇的读者和狂热的dataviz设计师而言，这一发现引出了著名的博客文章： “死亡到饼图” 。
This problem is also the perfect opportunity to get creative in the dataviz itself and gradually enter the realm of data art. Because it is already a compromise between different visual features, we could as well introduce aesthetics and emotions in the mix!
To iterate over a design, the dataviz practitioner must ask him/herself questions that criticize the current implementation. For instance:
- How would the audience judge its interpretability compared to the previous one? 与前一个相比，听众如何判断其可解释性？
- Is the color palette appropriate for the data at hand? 调色板是否适合手头的数据？
Is the data-ink ratio adequate?
- Does it look more emotionally engaging, if so why? 它看起来更具情感吸引力吗？如果是，为什么呢？
On top of that, other artistic concerns must be addressed such as:
- What is the concept behind this piece? 这件作品背后的概念是什么？
- Is there a sense of harmony, symmetry in it? 有和谐感，对称感吗？
- How does it compare to previous pieces, too much similar? Coherent with the series? 它与以前的作品相比有什么相似之处？ 与系列连贯？
- Is the shape supposed to be figurative or totally abstract? 形状应该是具象形还是完全抽象的？
- Do the colors resonate with the tone and idea behind the piece? 这些颜色是否与作品背后的色调和想法产生共鸣？
抢救数据艺术 (Data art to the rescue)
These questions led me to create this “fuzzy” radial tree.
While it is harder to count the number of layers, we have a better sense of the total number of employees and something which stands alone as a image for branding purposes. If you are interested in the technical details, it was created by using the previous radial tree and by applying a bit of a force-directed layout to it.
“Art is never finished, only abandoned” — Leonardo Da Vinci
“艺术永无止境，只有废弃” –达芬奇(Leonardo Da Vinci)
Once I “finished” this fuzzy radial data artwork, I put it as my wallpaper and moved on to other projects for a few months. I knew it was not in its final stage, yet I had no idea how to drastically improve it. Improving one aspect, such as visual compactness, would make a regression in another, like the interpretability for instance. In addition, I was so used to see a circular representation for these types of datasets that I had a hard time imagining something else.
一旦“完成”了这个模糊的径向数据图稿，我便将其作为墙纸，并转移到其他项目中了几个月。 我知道它还没有进入最后阶段，但我不知道如何进行彻底的改进。 改善一个方面(例如视觉紧凑性)将使另一方面(例如可解释性)退化。 另外，我很习惯看到这些类型的数据集的循环表示，以至于我很难想象其他东西。
I had to start fresh, a blank slate so to say, to be able to come up with something new.
打破死亡循环 (Breaking the circle of death)
If you are used to working with graph datasets, you know that most layout algorithms have a tendency to create circular-like shapes. It comes from the fact that they try to minimize the distance from the center of mass of the network without overlapping too much.
In layman’s terms, it means that most mathematical formula researchers use to draw networks on screen are based on the same physics model and that it often yields a circular shape.
Despite my love for these circular shapes, I wanted to create something radically different and kickstart creativity without doing a ring.
My first tries were leaning towards isometric shapes, but being a 2D artwork (to be printed) introducing a false sense of perspective didn’t really work for me here. I gradually came up to this triangular shape that I find compelling because it is so unusual in the network visualization world. It clearly shows each individual employee of the group, organized into each sub organization as we had before.
我的最初尝试是向等轴测形状倾斜，但是作为2D艺术品(要印刷)引入了错误的透视感，对我来说并没有真正的作用。 我逐渐想到了这个三角形，我发现它很引人注目，因为它在网络可视化世界中非常罕见。 它清楚地显示了该组的每个员工，就像我们以前一样组织成每个子组织。
Despite my enthusiasm for the previous triangular shape, I realized that it occluded the hierarchical nature of the network too much. I will keep this idea for a future piece where it really makes sense, here I clearly needed to go back to showing the links albeit more poetically…
After countless other iterations and refinements, I am happy to present the final version of the data artwork: “Together”. I hope that you find it visually engaging, not to say beautiful, and that it resonates within you as it did with me.
经过无数次其他迭代和完善之后，我很高兴介绍数据图稿的最终版本：“ Together”。 我希望您发现它在视觉上引人入胜，而不是说美丽，并且希望它在您内部引起共鸣。
最后的话 (Final words)
I hope that this article shed some light around the process behind the creation of a data artwork and the value behind this novel integrated approach: from data to visualization to art.
As we saw, the ability to analyze internal organisational chart data proves to be extremely valuable for the top management, shareholders as well as human resources.
More generally, the brand image is one of the most valuable assets of a company both for its clients and employees. Departing from classical marketing schemes, data art adds a unique perspective on branding, mixing emotions and cognition altogether, as we expose the essence of the company in one artwork.
Dr. Kirell Benzi is a data artist, science communicator and researcher. His primary interests revolve around creating visual experiences that inspire, educate and empower large audiences using state-of-the-art technology.
Kirell Benzi博士是一位数据艺术家，科学传播者和研究员。 他的主要兴趣在于使用最先进的技术创造视觉体验，以激发，教育并增强广大观众的能力。
In their whole, Kirell’s creations can be articulated and deciphered following an array of tones, shapes, dots, and lines that are staged according to the nature of the data. Through a hypnotic visual semantic, he works to show that algorithms have a soul… and that we can extract emotions from complexity using tools and methods which come straight from scientific research.
More data-driven projects: https://www.kirellbenzi.com