Squirrel Talk: Working with data as a one-man-army

David Hilzendegen became head of the data journalism department of the Heilbronner Stimme in April 2018. There, he is responsible for the acquisition and evaluation of data as well as programming more or less complex data visualizations himself. In this interview, he provides some tips on what to do if you are the only data journalist in your newsroom.

Should journalists learn how to code?

Short answer: YES!

Long answer: No, but I think it is important to know about the possibilities we have in terms of data procurement, data analysis, data visualization and storytelling. In my opinion, every modern journalist should at least have an idea of coding and should be able to identify JavaScript when he or she sees it.

What does the daily work routine look like for you in the Heilbronner Stimme data team?

You mean the one-person-data-team I am? There is one big advantage to being the only data journalist in the newsroom: I can work on any topic I want, whenever I want.

Most of the time, I have two or three smaller projects I am working on. Often I stop projects because it turns out that there is no interesting story in the data and certainly not the story I was hoping to find. Sometimes there is another dataset which is way more interesting to me (Yeah, I am really trying to get more organized. But there is so much data...).

So for most of my workday I am searching for data that covers ideas I got. After data processing the day is mostly over. Because that's the less romantic part of data journalism: Data wrangling takes a lot of time. A. Lot. Of. Time.

At the moment I am working on a machine-learning-project which is totally new to me. So right now i read lots of tutorials and blogs, try stuff myself, fail and start over. (Gosh, I hope my editor in chief doesn't read this....)

Which of your data projects are you most proud of?

I think our project on particulate matter was very good. My colleague Lisa Reiff and I built a measuring network for air quality in Heilbronn. We assembled measuring devices and distributed them to our readers. In cooperation with the local university, we created a platform where our readers could check the live data. Afterwards, we analyzed the data and isolated five learnings.

Our readers really liked the outcome. We had many readers and an average reading time way beyond our usual value. At times, it was at nearly five minutes, which is bombastic for my newsroom.

Analyzing commuter flows was a lot fun as well, especially the visualization in this cobweb-way. The project was the start of a series of articles about commuter and traffic in and around Heilbronn.

What didn't work out so well?

"Stadt, Land, Krank" didn't work out that well because I wanted too much. It was my first big scrollytelling attempt and I wanted it to look like a pudding.cool piece. I still like the project, and on my Mac and my iPhone it works like a charm. But it turned out that it is a little too heavy for the usual cellphone-user or older computers.

The quantitative as well as qualitative feedback was still good though. So I guess it is okay to just try stuff out and make mistakes. As I said before: Code, visualize, fail, learn, do better.

Which tools do you use to manipulate, edit and handle data?

Most of the time I use the iconic trio: The Python libraries Pandas, Numpy and Matplotlib. For interactive visualizations, I use Leaflet for maps and D3.js for the rest. For further designing purposes I use Sketch.

In the beginning, these tools are hard to learn, especially D3. But it is worth investing time. Because in the end, you can work, analyze, and style however you want without restrictions.

I try to avoid using commercial tools like Carto or Infogram. I think it is important to have your own look and feel that fits the topic and that can distinguish your work from others. Furthermore, there is no guarantee that commercial tools will still exist in three or six months. But your own code surely will.

Do you work with Excel? Or do you avoid it?

Personally, I avoid Excel. But I teach it to our trainees and in my courses. I think it is a good start to get in touch with analysis and visualization – as long as you don't publish Excel charts.

Working with data can be time consuming. How do you judge if it is worthwhile to do a data project?

I don't understand the question. Are there any good reasons at all to not start an interesting data project? Suppose I'm the only person in my newsroom who knows about data or how to code. What can I do?

Welcome to my world… My solution: Work work work. Scrape, code, visualize, fail, learn, do better. Then show it to some colleagues who are working on beats your story is about. Most will be impressed and you can discuss how you can improve and how they can cover your work.

Because the most important thing is: We are journalists and we should use data in a journalistic way. It is not our duty to visualize for the sake of visualizing.

The second important thing is: Data journalists aren't there to prettify articles. It's better to look for a real project you are working on... even if you have to do it after work sometimes.

To what extent does your newsroom support you in the development of data journalistic formats?

As I said before: It is great that I have the freedom to just do whatever I am interested in. So I guess giving me enough time to experiment is the main contribution of my newsroom.

What do you expect from your newsroom?

Sometimes I wish my colleagues would approach me more often on their own. But I think that's difficult with all the time problems that are typical for local newspapers.

Does data journalism have a future in small local newsrooms?

Do small newsrooms have a future without data journalism?


David Hilzendegen

David Hilzendegen believes that local journalism and data journalism are a perfect match. Bad enough that many houses still don't recognize this. So he continues his fight for more acceptance - in the editorchips and also inside the DDJ community.

Runs on:

How many pie charts have you built?
I built two Raspberry Pis. Isn't that enough?

Rate your CMS from 1 to 10.

How many times per week do you have to explain what "data journalism" is?
I hate you for this question!

How many screens do you have on your desk?
Not enough.

Swear words per day?
Not sure. But I guess it follows the Fibonacci sequence.

snow flake
© 2018 Journocode