Why I changed my mind on dual-axis charts

This week, we at Datawrapper launched two new chart types: waterfall charts and dual-axis charts. “Dual-axis charts?” I see you pondering, raising your eyebrows. “Wait, what? Are you sure? Dual-axis charts? Those awful things? Didn’t you yourself write a lengthy article on why they are bad?”

Oh yes, dear reader, I did. In May 2018, I wrote "Why not to use two axes, and what to use instead." And then in May 2022, the Financial Times' John Burn-Murdoch designed this beauty:

A readable dual-axis chart! Well-designed! With important (disturbing, really) data! And without an obvious “this should’ve been an x chart” alternative! It changed my mind on dual-axis charts, as I wrote on Hacker News two years later.

Toph Tucker helped with that, too. Like him, I’ve worked in mainstream media for quite a while — so when I think about charts, I think about how best to communicate data to the general public out there. Toph saw the other side. He spent some time working at a financial analytics company, and then wrote about the experience in "Doing enterprise financial data visualization after data journalism." Let me quote him, because Toph is great and what he says is great, too:

Dual axes are sometimes fine. I agree with basically all of this wonderful Lisa Charlotte Rost post on why not to use dual axes — and yet! Some smart, experienced, well-informed, well-intentioned people with skin in the game sometimes really, really want to use them! So I am forced to conclude that they know something we don’t. Like: maybe they are not being misled because they are not looking for what we think they’re looking for. Sometimes they aren’t looking for a statistic or trend or thesis; they are navigating a space. When this happened to this instrument, what happened to that? This isn’t a correlation, it’s: when there was a single big move in one, what did the other do? It’s easier for the eye to leap from one line to another when they’re close. Toph Tucker, 2019

We saw the same thing at Datawrapper: people want to use them, especially people working in financial services. That's why, starting today, you can create dual-axis charts in Datawrapper as part of our new business plan.

And so in today's Weekly Chart, I present to you a dual-axis chart — and two alternatives that may or may not be better.

The dual-axis chart

First, the dual-axis chart. This one combines stacked columns and a line and shows data on vastly different scales: 0-5.1 million bikes sold, and 0-7.4 billion euros in revenue made from selling them.

As we can see, the total number of bikes sold in Germany was the same a few years before and after the pandemic peak: 3.8 million in both 2017 and 2025. Sales value, on the other hand, more than doubled over the same period. The reason: e-bike sales. In 2025, more than every second bike sold was an e-bike; in 2017, it wasn't even every fifth.

What we also see is that both sales value and e-bikes sold peaked in 2022, while the total number of bikes sold hit its high in 2020. Since then, sales value has fallen faster than the number of e-bikes sold. (The chart doesn't tell you why, but it's because e-bikes have become cheaper.)

Alternative 1: Small multiples

Let's try to tell the same story with a small multiples chart. It's one of the alternatives I suggested eight years ago:

This looks nice! And we have more room for annotations here than in the dual-axis chart, which I like. Because the metrics sit next to each other rather than on top of one another, we can tell a story with panels, like in a comic strip. It's also easier to see the decline in regular bikes sold, because it's no longer part of stacked columns.

What works less well:

• To show that the total number of bikes sold hasn't changed much, we need to add a panel (because Datawrapper doesn’t let us put a stacked column chart next to a line chart – but even if it did, we wouldn't have space for all the annotations anymore).
• It’s a bit harder to compare how the metrics did at certain dates, like the insight that the most bikes were sold in 2020, but the most money was made in 2022.
• It's really hard to learn anything about the share of e-bikes sold because of the different axis scales.

Alternative 2: Indexed chart

So how about the indexed chart, another alternative I suggested back then?

This looks… hard to understand.

• As with the small multiples chart, we need to add an extra metric (the total number of bikes sold). So now we're showing four lines, and all of them fairly important. That's never an ideal situation in a line chart.
• We can't learn anything about shares (how many bikes were e-bikes in 2025?) or absolute values (how many e-bikes were sold in 2025?).

But hey, now the data series are on top of each other again instead of next to each other, so once (if!) we understand the chart, we can easily see that combined bike sales peaked two years before sales value and e-bikes sold.

Here’s why Toph appreciates that dual-axis charts show non-normalized values instead of the normalized ones in an indexed chart:

Non-normalized lines preserve references to absolute levels, which are essential landmarks, like “when AAPL was at $100.” And in a world of nonlinearities, a 1% move when rates are low is not the same thing as a 1% move when rates are high.

Toph Tucker, 2019

So what's best?

I think we have two winners for our data! I like the dual-axis chart because you can compare best when each metric peaked, and you can see the shares in the stacked columns.

But I'm also a fan of the small multiple lines, for their comic-like storytelling and because it shows so clearly how regular bike sales fell.

The indexed chart can stay home, though. Too many lines.

Agree? Disagree? Let me know at lisa@datawrapper.de or on social media. If you want to learn more about dual-axis charts, I put together a whole new guide on when to use them and how to make them less misleading. And we explain what makes our Datawrapper take on them special in this announcement. We'll see you next week!