Awesome Plotly with code series (Part 2): Colouring bar charts

Don't create a rainbow coloured bar chart. But don't make your bar charts boring either.

Image generated with DALL-E

Welcome to the second post in my “Plotly with code” series! If you missed the first one, you can check it out in the link below, or browse through my “one post to rule them all” to follow along with the entire series or other topics I have previously written about.

Awesome Plotly with code series (Part 1): Alternatives to bar charts

A short summary on why am I writing this series

My go-to tool for creating visualisations is Plotly. It’s incredibly intuitive, from layering traces to adding interactivity. However, whilst Plotly excels at functionality, it doesn’t come with a “data journalism” template that offers polished charts right out of the box.

That’s where this series comes in — I’ll be sharing how to transform Plotly’s charts into sleek, professional-grade charts that meet data journalism standards.

What will we cover in this blog?

1. When should we only use 1 colour for all the bar charts?

2. If we wanted to highlight a bar, how should we do it?

3. An idea on how to tell the audience that there is a “data issue” with one of the bars.

4. A possible way to show a benchmark inside a bar chart.

Intro - A story is always less dull with colours

Most data visualisation libraries or tools have one thing in common: if you are plotting 1 series of data in a bar chart, then only 1 colour should be used. You actively have to work hard to colour the bars differently. My kid does it at pre-school and because he loves colouring shapes, but hey, he is 4. We are adults and need to properly convey a message.

However, because we do need to tell a story, sometimes this story requires highlighting something specific. This is where colours come into play. With a careful selection of colours, we can make a super powerful and engaging plot for our audience.

PS: As always, code and links to my GitHub repository will be provided along the way. Let’s get started!

Single coloured bars to avoid bias

As mentioned in the introduction, many data visualisation tools would, by default, colour a single series with one colour. Plotly is no different.

Theory: no contrast opens the door for the human brain to make it’s own conclusions.

• Contrast is a pre-attemptive visual attribute. The human brain is wired to immediately pick out differences in our environment. Applied to a bar chart, if there is no colour contrast, the human brain would look for other types of differences; for example, bigger or smaller bars. Therefore, in the absence of colour contrast, there’s no subconscious nudge to focus on one element over another.

• Colours might have meaning. When we use multiple colours without purpose, the brain might mistakenly assign meaning to those colours, leading to unintended interpretations. If your goal is to present the facts neutrally, then don’t emphasise one data point over another using colours.

The plotly.express example — 1 colour by default

Imagine that you are making a presentation about South American countries. You want to appear as impartial as possible and it might be very important for all the countries to be treated equally from a visual perspective. This is the data you have access to:

And with this data, you might build a simple default bar chart like the one below.

Not a bad plot, but it can be improved.

Where do I think this plot has issues?

All in all, the above default plotly.express is actually pretty decent, but some improvements could be done.

1. Redundant axis titles

2. Difficult to pin-point the exact GDP per capita amount for each country.

The plotly.go example(s) — an improved bar chart

Given how simple this one-coloured bar chart is, there are not many changes to make using plotly.graph_objects , but improvements are possible.

A better designed one-colour bar chart

Why do I think this plot is better?

1. Reduced clutter by not having axis titles.

2. Removing the x-axis and adding the specific values helps the user directly know the GDP per capita and not have to scroll their eyes bottom to top and viceversa.

3. Adding flag emojis could increase clutter, but flags are actually well known symbols which can help the reader immediately recognise a country without having to read the full country name text.

4. Always add the source of data where possible.

Summary of improvements

---

Highlight with 1 colour

Photo by Mitchell Luo on Unsplash

When you actually want your audience to care about a specific category, you need to do something about it. Your best option is to make use of colour contrast, as it is a pre-attentive visual attribute. 

Let’s say that we want to tell a story about how Uruguay is having a great GDP per capita success but how Venezuela is struggling. In order to highlight these countries using plotly.express there is already manual work to do defining colour categories. For that reason, I will directly jump on using plotly.graph_objects . Check both screenshots below.

Highlighting Uruguay

Highlighting Venezuela

What have we changed in this chart?

1. First of all, we have used a grey colour for all countries except for the ones we want to highlight. For Uruguay, we used the specific RGB blue flag colour. For Venezuela, we chose the specific red RGB colour as red can also convey something “bad”.

2. We have removed all specific data labels for the bars except for the ones we want to highlight.

3. We have re-included the x-axis. Given that we removed the data labels, we still want the audience to be able to compare across bars. But want we really want is for all eyes to be immediately drawn to the highlighted bar.

4. Compared to the simple title of “GDP per capita in South America” from the first chart, we are telling the audience exactly what we want to highlight through a title and subtitle.

Summary of improvements

Tips on how to create this plot

How to personalise the colour and text in a bar chart?

• Colour, text and text position selections are easily configurable within a go.Bar() object.

# Choose a colour for Uruguay and for the rest
marker_color_ = ['rgba(85, 181, 229, 1)' if c_ == 'Uruguay' else 'lightgrey' for c_ in df['Entity']]

# Depends on how your dataframe is formatted. In my case, I had to divide by 1_000, format to 1 floating point and add the "k" and the "$" symbols. For the rest of countries, just leave blank.
text_ = [f"<b>${val / 1000:,.1f}k</b>" if c_ == 'Uruguay' else '' for val, c_ in zip(df['GDP per capita'], df['Entity'])]

# Because only 1 value from text_ will be shown (the rest are blank values), we dont need a list comprehension to define the text position.
textposition = 'outside'

fig = go.Figure(
    data=[
        go.Bar(
            y=df['Entity_emoji'],
            x=df['GDP per capita'],
            marker_color=marker_color_,
            text=text_,
            textposition=textposition,
            orientation='h',
            showlegend=False,
        )
    ]
)

How to hide or show the x-axis?

• If you want to control showing the x-axis or not, it can be done configuring the xaxis parameter inside fig.update_layout() 

if remove_xaxis:
    xaxis_ = dict(visible=False)
else:
    xaxis_ = dict(
        title='',
        showline=True,
        linecolor='lightgrey',
        linewidth=1,
    )

fig.update_layout(
  ...,
  xaxis=xaxis_,
)

---

Telling the audience to be careful because of data quality.

As data storytellers, how many times have you encountered dodgy data points? And how many times where you in the fence of saying something about it or not in your documents or charts? I offer you a solution based on contrast.