PAW Visualisations, Planning Analytics Workspace (PAW), Point Chart

The Point Chart, PAW and Beyond – Into The Light

The point chart is a mysterious creature. It lurks in the shadows of academia and statistics. Beyond this realm, point chart sightings are rare, their purpose often misunderstood. Observe it, however, and you’ll see that the point chart is intuitive and effective – it will bring your disparate data points together and communicate them clearly and accurately. Nestled in the darkest corners of Planning Analytics Workspace, it’s time to bring it into the light.

This article will show you:

The research and rationale behind the point chart
Point chart misunderstandings, and how they don’t hold up
Examples of the point chart in action
How to determine if the point chart is right for your audience and your data
Methods, tips and tricks to make your point chart shine in Planning Analytics Workspace (PAW)

History of The Point Chart

The point chart begins as the Cleveland dot plot. It was conceived in 1984 by Cleveland and McGill after years of research into how humans perceive charts. This knowledge has evolved since – Kieran Healy summarises the history and results of this well – but the point chart continues to hold up.

The point chart encodes quantities in the most fail-safe way possible – through their position on a common scale (see here again).

A point chart showing the revenue for five different product lines for 2012, 2013 and 2014. Each year has a vertical line, and each product line has a point on a each line. Higher revenue products lines are higher up on the line. Click to visit the IBM PAW documentation on the point chart. — The point chart from the IBM PAW documentation.

This example from the IBM Planning Analytics Workspace documentation demonstrates this: each year has its own number line, but they all follow the same revenue scale on the left. This makes it easy to compare revenue across years and product lines. Despite its deep and obscure academic roots, the point chart effectively communicates its data. Let’s take a closer look at it in PAW.

IBM Documentation

A point chart is good for showing “trends over time”, according to IBM. The documentation insists that “the x-axis should show time”, and if the x-axis shows anything else (it lists countries as an example), we should use a bar chart instead.

This is a misunderstanding.

While this is a suitable role for the point chart, the IBM documentation risks banishing it into the dark. We can’t confine the point chart to this narrow role. It can do so much more.

“A Line Chart Without The Lines”

The IBM documentation says that “a point visualization is like a line chart without the connecting lines”.

It’s true, but why do the charts work so differently with or without them?

A diagram of eight Gestalt principles.
Two rectangular groups of six circles represent the proximity principle - the close shapes are visually grouped together.

Outlined and filled circles mixed together represent the similarity principle - the viewer groups the outlined and filled shapes together, even when they're not next to each other.

A group of circles with a filled box behind two circles and an outlined box behind three other circles represents the enclosure principle. The boxes group the circles inside together.

A grid of twelve circles with a line through four of them represents the connection principle. The line groups the four points together.

For the continuity principle, two paths of circles cross. They are filled before they meet, and outlined after they meet. Because they're arranged in continuous paths, the viewer groups the points on the same path together.

For the closure principle, circles form the outline of a box. Because they're arranged in a closed shape, the viewer perceives the shape.

For the figure-ground principle, a path of dots leaps out from the page left-to-right, with a wavy grey area behind it. The user perceives the figure (circles) and ground (grey area) separately.

For the common fate principle, a line of circles has arrows pointing out of it, with the direction alternating. The viewer perceives points going in the same direction as part of the same group. — Above: a selection of Gestalt principles. Click for original image.

Gestalt principles are different methods of grouping visual objects together.

Line charts unify a series of points over time with the connection principle. This principle is stronger than proximity and similarity principles, so a viewer will group different points along a line together, even if other points are nearer or more similar.

Once you get rid of the lines from your line chart, you lose that strong sense of connection. Other groupings on your chart can then shine through.

The Point Chart In Action

This example from Pew Research Center shows the point chart at its brightest. The items here are countries, one specific case where the PAW documentation discourages a point chart. Countries don’t have an inherent order to them, so a line chart would be inappropriate. A point chart, however, can effectively communicate the data with careful application of Gestalt principles.

A point chart from Pew Research showing attitudes on social media's effect on democracy by age group.

The title says "Young adults see social media's impact on democracy more positively than older adults in most countries". The byline says "% who say social media has been more of a good thing for democracy in their country, among those ages..."

The chart has a horizontal line for a selection of 12 countries. There are three points on each line: a green point for 50+, a yellow point for 30-49 and a blue point for 18-29. The higher the value, the further to the right each point is. A thicker grey band connects the lowest point to the highest point along each line. Each point is labelled with its value. To the right of each line, the "youngest-oldest diff" is written out.

The caption reads
"Note: Only statistically significant differences shown.
Source: Spring 2022 Global Attitudes Survey. Q28.
'Social Media Seen as Mostly Good for Democracy Across Many Nations, But U.S. is a Major Outlier
PEW RESEARCH CENTER"

Click this image to see the original article.

A Caveat: Aim

Before we begin, let’s establish that Pew’s aims are different to yours (probably). Charts have two main purposes:

Explanatory: the creator has already drawn a conclusion from the data, and the chart is there to back it up.
Exploratory: the creator presents the data as-is, and the audience draws their own conclusions.

This Pew chart is explanatory. You can see the conclusion they’ve drawn in the title: there’s a generational gap in the data for each of these countries. The chart exists to back up this claim.

Most dashboard charts are exploratory: the aim is to give an overall view of the data for the audience to draw conclusions and make decisions from.

Keep this in mind as you analyse and create charts.

Back To The Chart

There are two main Gestalt principles at play here: connectivity and similarity. The connectivity principle groups the points together by country: both the grid lines and the grey lines between points do this. The similarity principle brings each age group together with colour.

As mentioned before, the connectivity principle is stronger than the similarity principle. If this chart were a line chart, then the country and age grouping would both use the connectivity principle: they would have equal emphasis. As a point chart, the country group comes first, and the age group comes second. This guides the viewer to Pew’s point: there’s an age gap in each country.

A snippet from the Pew point chart above. It shows three horizontal lines, labelled UK, Australia and U.S. Each line has three points on it, one for each age group, and the points are labelled with their percentage values. The further along each point is, the higher the value. A thicker band connects the lowest and highest point on each line.

One line (Australia) is highlighted to demonstrate how the chart groups together points for each country. — The lines join together points from the same country: the connectivity Gestalt principle in action.

Pew also takes advantage of sorting to highlight its point.

A line chart typically has time on one axis. Time is always fixed in its order: July will always follow June, and there’s a high risk of misleading your viewer if you re-arrange your months, years or days.

Instead of time, Pew has countries. Since countries have no inherent order to them, Pew can sort the rows as it likes. Sorting allows viewers to make quick comparisons. It’s common to rank by raw values, but Pew wants to highlight the generational gap. So, Pew has sorted by the gap between the 18-29 group and the 50+ group. This way, Pew makes the gap more apparent, and the viewer can make faster and more nuanced comparisons.

Overall, Pew’s point chart showcases the simplicity, versatility and clarity of the point chart. It brings a jumble of countries together, and it shows what what each age group thinks of social media’s impact on democracy with grouping and sorting. While this grouping and sorting would fail in a line chart, they shine in a point chart. The chart both demonstrates Pew’s point and allows the viewer to make their own comparisons. Despite the PAW documentation, a point chart is highly effective in a situation where a line chart would fail.

Why You Should Use The Point Chart

So you’ve seen the point chart in the light – you know what it can do. But is it the right tool for the job?First, consider your aim and audience.

Here’s the Pew chart’s approximate aim and audience:

Aim: highlight the generational gap in attitudes across countries about social media’s effects on democracy
Audience: English-speaking, has internet access, baseline numeracy and literacy level. Otherwise, general

Yours might look like this:

Aim: show actual, budget and forecast data by branch to facilitate business decisions
Audience: your company’s finance department

Once you have a defined answer for both, your chart will be easier to create and more useful for your audience. If you want to consider your audience in further detail, then read here.

Now that you’ve sorted that out, here are the situations where a point chart is especially suitable:

You have two dimensions to visualise – one categorical and one numeric: “categorical” means your dimension’s data fits into neat categories, not a continuum. These groups can have an inherent order to them, such as months or years, or they can be unordered, like countries or branches. The groups can be disparate, and they can even overlap: see another example from Pew. The point chart brings different groups together and displays them well.

A template point chart shows what type of dimensions a point chart can encode in PAW.

The title says "The point chart displays one categorical dimension and one numeric one". The byline says "Optional additional categorical dimension on colour". The word "colour" is in the three colours of the points below.

The chart below says "numeric dimension" on the horizontal axis and "categorical dimension" on the vertical axis.

The chart has bold horizontal lines, with three points, one green, one blue and one purple, along each. Each point is at a different length along each line.

The chart has lighter vertical gridlines to help align the points with the numeric dimension.

Once you have your groups, each group gets a line to put points on. Then the numeric dimension determines where on the lines your points will go. If your main dimensions are both numeric, consider a scatter chart instead.

You want a chart with low visual weight: each quantity is represented by a small point, no matter how big the quantity is. As a result, a point chart has a lot of negative space. It’s easy on the eye, and it won’t visually compete with other charts on a dashboard.

A point chart and a bar chart, side by side, show the reasons to choose one chart type over the other.

Both charts have the title "Units Sold By Branch by Machine Type, 2023" and a legend for machine type below it: "Cardio (20000)", "Cable Machines (30000)" and "Single Stations (40000)" each have their own colours. Both charts have the same axis labels: location on the vertical axis, "Units Sold" on the horizontal axis.

The point chart has one horizontal line per location and three points on each line, colour-coded by machine type.

The bar chart has three horizontal bars per location, colour-coded by machine type. — Above: a point and a bar chart, side by side. Which is easier on the eye? Which one makes it easier to compare by location? By machine type? Which one makes the totals more apparent?

Your parts are more important than the whole: Point charts make it easier to compare parts to each other than to their combined whole. As demonstrated, the point chart is great at showing individual quantities. Try adding up the values across each category, though – it takes a lot of concentration.

With a different chart type, viewers will get the gist of these sums subconsciously. Bar (and column) charts represent amounts with areas of colour, a pre-attentively processed feature which is quick to understand. Some types of bar charts also take advantage of length for quick part-to-whole comparisons.

The following chart types emphasise the whole more, in order from most to least:

Pie charts: effective at communicating ratios, if ineffective at communicating actual quantities (see 2.1 and 2.2, as well as the example below). Hotly debated, but that’s beyond the scope of this article.
Stacked bar charts, especially single-bar versions: the parts form one contiguous shape, so the proportions are clear. The viewer can use length and area to determine ratios.
Bar charts: unlike stacked bar or waterfall charts, each bar starts from zero. Viewers need to consciously add lengths, but they can still compare areas at a glance.

Four charts in a 2x2 grid: a point chart, and three alternatives which show aggregates and ratios better. Each chart shows the number of units sold by machine type by region.

The top-left chart is a pie chart. It has four pie charts in a row. Each is labelled with the region it represents. Each pie chart shows the proportion of units sold for each machine type in each region. The watermark "Bad" spans the chart.
This chart effectively shows the ratios of units sold by machine type within each region, but it represents the ratios between locations misleadingly, and it fails to show absolute units sold entirely.

The top-right chart is a stacked bar chart. Each bar has three coloured sections stacked next to one another, one per machine type. The longer the bar, the more units sold for that machine type in that region. The stacked nature allows the viewer to compare totals by region, and the viewer can also compare individual machine types. Because of misalignment, however, comparisons within regions and for machine types on the right side of the stack are harder.

The bottom-left chart is a bar chart. Each region has three horizontal bars stacked on top of each other top to bottom. The left end of each bar is aligned with the horizontal axis, and the right end of each bar shows how many units of each machine type were sold for each region. The alignment makes comparisons by machine type and within region easier, but region-wise totals are less obvious than for the stacked bar chart.

The bottom-right chart is a point chart. Each region has a thin horizontal line through it, and each line has three points on it, colour-coded by machine type. The further to the right the point is, the higher the units sold for that machine type in that region. This allows for individual comparisons across machine types and regions, but the lack of proportional areas makes aggregate comparisons harder. — The point chart and a few alternatives which emphasise the aggregate data more.
Eagle-eyed readers may notice some smaller pie chart wedges actually correspond to larger numbers.

The above list is not exhaustive, of course. This flowchart may also help you in such a situation.

You want to emphasise all your points equally: in a point chart, each point is the same size, whether it represents a small quantity or a large one. Other chart types, such as bar charts and pie charts, associate quantity with area, so bigger numbers are represented as bigger areas, and the viewer notices them more than smaller numbers.

You need to truncate your numeric axis: you’ve thought about it, and you’ve concluded you absolutely must not include zero in your numeric axis. If this is the case, a point chart is more effective than a bar chart. When you truncate the numeric axis of a bar chart, your bars change size, and you falsely inflate the differences between your numbers. With a point chart, your points are still the same size, so the differences are less misleading. This is the principle of proportional ink in action, vital for honest data visualisations. I encourage everyone who ever looks at charts to read more about it here.

You’re short on space: you can fit multiple points on one line. Other charts, such as bar charts, will take up more space if you add in dimensions. So, a point chart will be more compact and easier to interpret in a smaller space than other charts.

Your audience has basic numeracy skills: as said earlier, the point chart was designed to be easy to interpret. Many people encounter number lines as children, and most people in a business environment have the numeracy to read a point chart properly. If you expect especially low levels of numeracy in your audience, a frequency-based icon chart may be more effective – see here for a review. However, a point chart is appropriate for a general audience.

Now that you’ve decided the point chart is right for you, let’s put it into practice.

How To Create a Point Chart In Planning Analytics Workspace

There are two ways to make a point chart in PAW:

Go to the “Visualizations” tab, then drag in a point chart.
Create an exploration with your categorical dimension in the rows and your numeric dimension in columns. Then, convert to a point chart at the top of the page.

Fields

The point chart in PAW has multiple fields to encode information.

Item Axis

Your categorical dimension goes here. As mentioned, this is a dimension that can be split up into groups. If your groups are ordered (e.g. months), then keep them in order. If they’re unordered (e.g. countries), then you can sort them:

Recall your aim: what quantities does your audience want to know about most? Determine what value you’re going to sort by.
Create a new exploration. Put your discrete dimension as rows and your numeric dimension as columns. If you want to assign a dimension to point colour, then put these dimensions under the numeric dimension.
Sort as desired.
Convert to a point chart.

I strongly recommend item axis gridlines: they hold your data together. Ensure the points are still the primary focus of your chart, but otherwise, make the gridlines as dark and bold as you dare.

You can omit the item axis line, but keep your labels so your audience knows what they’re looking at.

Value Axis

Your numeric dimension goes here – any dimension with a low end and high end works.

Place your value axis on the left side of your chart, or the top if your chart is horizontal (see Tips and Tricks below). Your audience will read the axis before they read the chart, so the data will be quicker to understand. If your audience speaks a language that isn’t written left-to-right and top-to-bottom, then modify accordingly.

There is merit to directly labelling your data – the viewer doesn’t need to look back and forth repeatedly to determine the value of each point. However, it can make your chart look cluttered. If you don’t label your points with their values, then your audience has no way to determine the value of your points at a glance. Ensure your axis labels and ticks are detailed enough to be meaningful, if you include them.

Colour

There are multiple ways to use colour in a point chart. You can add entirely new information – add multiple points to each line and distinguish them by colour. You can also use colour to create groups – if each line is a country, you can use colour to group them by continent (so long as they are separate dimensions).

Colour is somewhat effective at encoding non-numeric information – for example, a different colour for actual, budget and forecast points. However, it’s ineffective at encoding numeric information – for example, a lighter shade for smaller quantities and a darker shade for larger quantities. If you want to show an additional numeric dimension, consider a different chart type.

See here again if you want to know how well colour encodes numeric and non-numeric information compared to other design elements.

Use colours that contrast well with the background and each other. Since the points in a point chart are small and spaced-out, subtle differences in colour will be harder to see.

Two charts side by side demonstrate the importance of a high-contrast colour palette in point charts.

The point chart and the bar chart each use three colours, two of which are very similar. In the bar chart, the large, adjacent areas make the similar colours distinguishable. In the point chart, the small, separate points mean the colours look the same. — Point charts need distinct colours – close colours may be hard to distinguish in a point chart.

Colour Vision Deficiency

To maximise the effectiveness accessibility of your chart, ensure your colours are colourblind-appropriate. About 4% of the general population has a colour vision deficiency, and your organisation may print greyscale copies of your chart. Here are a few ways to ensure your colours still work in these situations:

Choose a pre-made palette: the Okabe-Ito palette, ColorBrewer palettes and other sequential palettes are made with clear, colourblind-suitable data visualisation in mind.

Ensure your colours have some light-dark variation: this way, colours are distinct for all types of colourblindness and for greyscale printing. Even if you aren’t colourblind, the human eye is three times as sensitive to light-dark variation than it is to variations in hue alone, so your charts will be more precise.

DIY: the iWantHue tool automatically generates distinct colour palettes. Set the luminance range wide and choose to improve for colourblindness. Check the colour comparisons below to ensure your colours are distinct enough. The tool is also a fascinating use of the k-means algorithm, but I digress…

Test it out: once your chart is ready, put it into a colourblindess simulator and see how it goes. Red-blind and green-blind are the most common types, and monochromacy simulates greyscale printing. Viz Palette is also an excellent resource to test your palette across chart types and colour vision deficiencies.

Rotate Your Point Chart

The item lines in the PAW documentation’s point chart go vertically, but many point charts in the wild have their items along horizontal lines. Despite what the IBM documentation claims, you can rotate your point chart accordingly.

This can make your chart easier to read. The information flows left-to-right, top-to-bottom, the same way as written (Roman-alphabet) text (p. 57). There are multiple other reasons you may want to rotate your chart:

You want a list-like appearance,
Your items have long names, or
Your chart area is tall or narrow.

Follow these steps to rotate your point chart:

Turn on edit mode and select your point chart.
Select “Properties” on the top right. Go to “Visualization”, then “Chart”.
Turn on “Transpose”.

Two point charts side by side show the effects of rotating a point chart. The charts are the same: a series of number lines show the number of units sold by machine type by location. There is one difference between the charts: the number lines go vertically on one chart and horizontally on the other.

The horizontal chart has a list-like appearance. The long location names are easier to read, and there is greater space between the lines. The horizontal arrangement of information also mirrors written text in many alphabets (e.g. Roman, Cyrillic), so the information flows better for viewers with this background. — A point chart reads differently when rotated.

The Mirror

The point chart has long been underused and misunderstood. Despite decades of academic backing, it remains in the dark. However, its number line-style graphics are versatile, intuitive and clear. The point chart is a mirror for your data. It allows for reflection. When you shine a light on it, the light spreads. Illuminated, data-driven decisions move through your organisation. All you need to do is shine the light.

Need Help with Point Charts?

Please reach out if you need any assistance with PAW visualisations – we’d love to help!

The Point Chart, PAW and Beyond – Into The Light

History of The Point Chart