NZZ scrolly telling the German economic crisis

I stumbled over this NZZ data story of November 2025, because I was looking through the NZZ's data visualisation department. Out of curiousity I sometimes look through the works of DDJ (data driven journalism) teams even when I didn't subscribed to it. Sometimes there are free articles like in this case.

Keep in mind I'm not a journalist. I practise data visualisation so I might not be aware of some aspects. If you are journalist and you have an opinion please let me know.

What is it about

The German economy is crumbling according to the article at hand and they can proof just how bad it is with data.

The decline began gradually—and long before the pandemic. The following charts show how industrial production in Germany has evolved since 1995—across nine sectors, ranging from the automotive industry to the arms industry.

As a format they chose to present the largest portion of the story as a scrolly telling format. Meaning you get data and facts as you scroll through the page. By this pattern we'r told a story. This is a very popular web presentation pattern (as of March 2026, when I write this).

The story at scroll

This is a longer section since there are different chart layouts due to the scrolly telling format. Things are just chaning while you scroll, so it's only a fixed chart for a certain scroll state. First we are introduced to the data and as we scroll we go through the following steps. You can skip this section and look for yourself, it's very long and tedious like the original.

  1. We are shown five line/area charts
    • three in a red color (labeled industries)
    • two in dark red color ( industries)
    • below we find four hidden lines
  2. The upper lines are hidden now and the lower ones are introduced as:
    • two green lines (labeled as industries)
    • two more for industries (meaning sucessors) in blue
  3. The bottom lines are hidden again and the key industries take over the space
    • This time they get more space since the two traditional are hidden too
    • Axis appear y are percentages, x is time/year spanning from 1995 to 2025
    • Also we got two dashed vertical lines appearing in 2008 (financial crisis) and 2020 (corona pandemic)
    • At first glace all these chart behave the same
    • Line charts are growing until 2008 and stay almost static until before 2020
    • We also get a fixed bottom div with an question mark icon labeled "Whats the percentages about?", we get to this one later.
  4. The first chart is highlighted and reveals it's the car industry
    • employs 720k people at 24% industry revenue share
  5. The car industry chart is seperated with a vertical line that highlights the section since 2018.
    • It tells electic cars hits the classic supply chain and 120k jobs are gone.
  6. The car industry chart moves forward to 2023 (next shock)
    • China produces subsidized cars, shock stays to today and energy prices and CO₂ regualtion hit this industry the text box says.
  7. Next mechanical engineering
    • Another text box informs us 930k people employsed with 11% revenue share
  8. Next highlight section, again in 2018
    • Industry slows down, demand goes back, china strong and low energy and labor costs, as a result 70k jobs are gone, the text box informs us
  9. Moving on to the third key industry we now know is the chemical industry
    • 320k jobs at 6.7% industry share
  10. You guessed it we hightlight another section but no! It's 2022.
    • CO₂ pricing, energy transformation and nuclear phase-out make energy expensive
    • Even worse environment and climate regulations also target this industry, we are told by a lovely text box
    • Another text box, just to shake it a bit up plants are closed in Frankfurt and Ludwigshafen. China in contrast keeps growing.
  11. We surpassed the key- and move on the the traditional industries. First: clothing and textiles
    • More text boxing 64k employees, ...
    • Textiles once been a heavy weight for the German industry, ...
    • This is the first line/area chart we see that is degrowing heavily

At this point I'm already asleep. At this point I'm just writing for achivements sake. Let's scroll on:

  1. Next text box, ... breweries
    • Not as rapidly degrowing but noticable. 23k employees in the beer industry, ...
    • Oh another text box informs us that demographic change makes older drink less younger differently
    • Exports static, while netherlands companies dominate interationally
    • Interesting I always had the impression young people drink way less compared to older, but anyways.
  2. the former charts are hidden now and we are shown the latter 4 charts. But the labels have changed
    • blue is now "against the trend" are post crisis better than before
    • green "just risen" industries are rising through tax money, unknown if they succeed asia is strong the text box tells us
  3. Scrolling on we see the batterie industry
    • Another variant this chart (an the other green one) is not sharing the space with other expanded charts thus it's larger.
    • The first rising chart this lightly growing until 2020
  4. Highlight section at 2020, our loyal companion the horizontally centered text box is speaking:
    • Rapid growth due to e-car boom, drop in 2023 due to governmental funding is running out.
    • Past 2023 the line is continuing to grow on a volatile coarse
  5. Next green chart: Weapons industry 15k employees, 0.6 global revenue share
    • Since cold war manufacturing of weapons and ammunition was drifts along states the text box
    • The chart is the most volatile shape and inceases lately
  6. Highlight section at 2022 Ukrain war, weapon industry grows at steep angle
    • Tank industry is not included (often confidential, thus not included)

We almost got it, now hang in there.

  1. Shared view next we get the last two (blue) lines (logically at half height)
    • First highlighted chip industry 77k employees, 1.7% total revenue share
    • Car and instry chips produced by Bosch and Infineon. TSMC also coming to Germany
    • Thanks to a billions (1000k, German and English billion differ) in funding program
    • The blue lines both growing at a stable rate
  2. Medical tech industry - 110k employees, 1.3% industry revenue share - increasingly older people make this sector grow reliably over the shown time

About the data

Sorry If this bothered you a little, but I had to finish this an for me it felt exactly the same way it reads. At one point I even felt like in an IKEA store walking the indended way when you don't know where the shortcut is. You get the point of beeing guided. But I want to analyse and learn from their work so first I want to see what we got without the scrolling.

Therefore we need to find out what data was used, the authors added a note at the bottom of their data story, which says the used destatis data (data from federal statistics office). If you go to the destatis database and type in 'production index', you get the statistics they used. More precisely it's the statistics code 42153-0003. You can find the way I processed the data in the repos datawork folder.

The authors additionally said: The figures are adjusted for price, calendar and seasonal effects. So we know we need to get the correct X13 series. Make a few adjustments so we get a date type in the data base by concating year and month. This comes handy now because we need to reduce the data to the articles time span. Which I believe is around march or july. Doesn't really matter in this case since a few 1-4 months won't make much of a difference in a 35 year scope.

Without the scrolly

After a little while of searching we find can find the actual codes used in the visualisations, so we can filter for these and reduce the data further. We create for colors, english labels and the maximas. We need the maximas because of a small thing the authors did to the y axis. Remember the percentage scale? This is why. When we hover the question mark icon on the bottom of the original article we'r informed that the 100% tick on y axis corresponds to maximum value of each industry sector in the given time span. Further it reads the percentages are not compareable amongst each other. In short the authors mapped max(car_industry) to 100%. We can replicate that by using a linear scaling.

The basic plot outside a scrolly telling scenario would look like so:

Looking at the scales

We got three different types of scales: (1) y which shows the performance of this sector in %, (2) a classic time scale on x and (3) a colour palette of four colours. The x axis ranges from 1995 to 2025, which are 30 years of data. We can also note that the only tick we get is 2010, so we can suspect that we don't need much more than that context (at least for the time scale).

The y axis ranges from 0 to 100, which obviously correspond to the percentages, mapped to the maximum value of each sector. The authors also state that the values are only comparable inside one graphic. We'll later explore why (also why you should).

The color scale shows us four different colours for the different lines and another four light versions for areas. Colours are some common pattern to guide readers perception. Red and dark red are preserved for key and traditional industries, while. Blue is preserved for structural winners that go against the (suggested) general decline. Green sectors are recently growing surprise industries while orange is used for the key industries and red is used for two examples that are declining in a long term trend.

Tone

My impression is that if you read through the text boxes the basic gist is everything bad and if something is good it's required because of war or tax payers are subsidising the trend against general decline. The authors mention regulations or political decision as negative factors: environmental and climate regulations, energy transformation, CO₂ pricing and energy prices in general.

But I'm not experienced enough to finally rate a tone here. That's journalists work. I'm a visualisation nerd with an impression.

Percentages on Y

The authors have chosen to normalise the production index values to a scale of 0 to 100%. This causes every line chart to start at 0 and end at 100. Which for the setup they went for is a good idea. People are not used to the production index and if every chart starts and ends at the same scale it's more intuitive to users.

On the other hand they lose context, because the index values are normalised individually, not globally. That means and they also mention it: you can't compare the sectors anymore. The production index as stated by the federal statics office is based on the volume of the production in each sector reported by these sectors on a monthly basis. When you normalise the way the authors did you loose out on the possibility to show users the volume in comparison. Here is an example that illustrates the influence of the normalisation in this case:

I included a few examples so you can directly see the influence of rescaling for this y axis by looking at the grey area between the line. It illusatrates the relative distance between the lines. Most impressive for me was the comparison between the mechanical engineering and the weapons and munition industry. If we have spiking data and we are rescaling like this, we get an effect where the spike is mapped to 100% and the rest is perceived less.

After remapping the spike is assigned 100% (largest industry size), while the rest of the time series is decreased relative to the spike. Choose the example regarding mechanical engineering vs weapon industry and see what happens when you toggle the rescale button. It would look as if the German weapons sector is just recently catching up to the mechanical engineering sector. But in reality the industry was always (since 1995) been as large as mechanical engineering in terms of production volume. Since around 2022 the sector is even exceeding other sectors not only mechanical engineering. You can see it in the next chart.

If you look at the unnormalised chart one thing really sticks out the weapons industry is raising above all other sectors since 2023, this is something you don't see in the normalised version. Another thing you cannot see is since 2007 the weapons industry is closing up to levels comparable with the chemical or mechanical engineering industry.

The unnormalised verion of this last chart would be my resdign proposal it gives users real interaction and additionally I'd include another dataset with non-manufacturing sectors, such that readers are able to inspect service sectors as well. Care sector should for example also grow in parallel with the med tech sector I'd suspect.

Since the article does frequently mentions that jobs have been lost over time in certain sectors, I'd like to know where they work now. I'd guess many german employees that lost their job in the manufacturing sectors might moved to service industries.

Conclusion

To sum up what I learned from this DDJ story, is where I get data on the german economy, roughly how the production index works (there is far more to it as it's used to calculate parts of the GDP). But also that scrolly tellings get very boring very fast. Sadly this navigation or I'd better call it presentation pattern is taking away most of the readers freedom and cages you into a narrow path. I'm more a fan of less fancy but more freedom for readers. But that's just me. Many DDJ stories currently are scrolly telling you to sleep.

I also learned that there is a trade-off between simplifications for users (not having to think about what a production index is) and the linearity of numbers and diagrams. The more you simplify the less exploration you posibilities your readers have - at least in this case. Giving the possility to compare the sectors over time would have gotten more user interaction (maybe I'm just naive). It didn't even hurt the story they told. But there might been other reasons I'm not aware of.

I do want to acknowledge that this story part was fully reproducible so that's really nice and helps me to learn from their work. Although I think it's not intended that random guys on the internet can learn from their content, it's more about the credibility as journalists. Please note that there is more but I didn't want to check on the rest of the story, where the authors show a chart to argue that government is growing while industry is shrinking, that's a part of journalism I don't want to get into. There may be a part of relation but that would require me to dig far deeper into the matter - again journalists work.