/ 6 min read /

Data-ink Ratio: How to Simplify Data Visualization

by Holistics Team

Data-ink Ratio: How to Simplify Data Visualization

Companies nowadays are very keen on incorporating data analytics to boost their performance in the market and so should you. Data warehousing and data modelling aren’t innovative concepts anymore. But there is a long way between having the data and properly utilizing the data.

After all the data collecting, cleaning and processing, as we convert the data into charts, we might think that our message is straightforward. But, is it really? Let’s always keep in mind that the visualization is a vital step for it crowns all the data efforts and compels decisive actions inside the organization. So we can’t spare any effort in providing a clear chart to the end-users.

The data-ink ratio, in broad lines, tries to accomplish this exact simplification for any chart. Applying the data-ink ratio, as we will see, will bring not only simplicity to your charts, but also a hint of sophistication.


What Is The Data-ink Ratio?

The data-ink ratio is a concept created by Edward Tufte, famous author in the field of data visualization. Clearly speaking, the data-ink ratio encourages chart creators to examine if all elements in the chart are relevant to the chart’s message. Or in Edward Tufte own words:

“Above all else show the data.”

The data-ink ratio can also be represented as:

Data-ink Ratio
Data-ink Ratio

During the chart creation, you should have a self-critical moment, when you question if all elements displayed are necessary to represent your data and its message, and if not, remove them. “Elements” should be understood as every aspect in a chart: colors, effects, legends, labels, images, annotations, etc.

A quick example of improving a chart with the data-ink ratio approach:

Source: Edward Tufte
Source: Edward Tufte


The benefits of this approach are:

  • Clear message: having only the necessary elements will make your message clearer and easier to consume by your audience. Not everyone is used to dealing with charts all day, so fewer accessory elements means less confusion.
  • Saving time: not only will the readers get the message quicker, but you as the creator will save time avoiding confused users with ad-hoc requests.
  • Saving space: If you need to show multiple charts and numbers in a dashboard or have some KPI at reports or similar, space is a resource. Data-ink optimized charts occupy less space and accept resizing better.

The Two Erasing Principles

Edward Tufte, in his book, The Visual Display of Quantitative Information, states the two erasing principles for a better data-ink ratio.

  1. Erase non-data-ink, within reason.
    Accessory elements that don’t add information should be considered for removal. In this category we have gridlines, colors without meaning or purpose, 3d effects, annotations that don’t add to the chart’s message, etc.
  2. Erase redundant data-ink, within reason.
    When we create a chart, sometimes we try to pack more information than necessary or the software as default adds extra elements. Check for additional data information that can be removed. In this category we have unnecessary legends, labels, excessive information unrelated to the chart’s message, and others.

Practical Examples On Data-ink Ratio

All that sounds very beautiful, but let’s see these ideas in action. Here we present and discuss some examples for the most common chart types. Keep in mind that the rationale used here can be applied for any chart type.

Bar Charts

Let’s say you are a business analyst for a company with multiple stores. You are writing a report and you need to show that the south region is underperforming. Take a look at this initial chart:

Source: Edward Tufte
Source: Edward Tufte

Let’s examine the chart’s elements for erasure:

  • The title of the chart is very telling of its content. “Total sales by region”: we already know that the numbers are going to be about sales and the categories are going to be the regions. This simple statement makes the labels “Sales” and “Region” unnecessary.
  • The Gridlines are used to give the viewer a comparative notion. Here they are overly apparent to the point of being distracting. One common strategy is to make them gray or light gray so they are less noticeable. In this example, we are going to remove them altogether along with the axis ruler. This will be possible by writing the actual values along the columns. The comparative notion will not be lost if the size of the bars are proportional.
  • The colors tried to create a distinction between the regions. In most cases that is fine, but remember that here we purposely want the audience to pay attention to the south region. In this case the different colors are distracting from the message. How about we make a sleight of hand and just color what we want the audience to see?
  • As we saw, the color categories are self-evident, rendering this legend redundant.

And this would be a data-Ink ratio optimized chart:

Source: Edward Tufte
Source: Edward Tufte

Look how the South region “pops” to our attention and makes the message more clear.

Line Charts

Let’s say now that you are going to report that some product categories had negative profits in some months of this year, but  it was a small loss. An initial chart could be like this:

Source: Edward Tufte
Source: Edward Tufte

Main points of interest for erasure:

  • The chart’s title renders the axis label “Profit” redundant.
  • The colored area under 0 line is an overkill, taking too much attention and ink. The same result can be achieved with a simple line, marking a clear distinction between positive and negative profits.
  • The “month” label is unnecessary. When people see a line chart and the month’s names it is already self-evident.
  • The legend can be removed if we use the labels along the lines.

The end result would be like this:

Source: Edward Tufte
Source: Edward Tufte

The same information is contained in less space, making the chart more pleasant to the readers.

Pie Charts

For this last scenario, let’s say that you want to depict that the product category “Office Supplies” constitutes the majority of the orders. For the initial Design, we could have:

Source: Edward Tufte
Source: Edward Tufte

Improvements:

  • Here the actual percentages aren’t so important. The overwhelming proportion of “Office Supplies” speaks for itself. In this example they can be removed.
  • Instead of the percentages, we can use the Product Category labels. In doing so, we can remove the legends.
  • An interesting idea, when possible, is to use doughnut charts instead of a  pie chart. This not only reduces the amount of Total-Ink, but free inner space for other information. In this example let’s insert the title inside of the doughnut chart.

The final result would be:

Source: Edward Tufte
Source: Edward Tufte

This particular design saves a lot of space and is very convenient for dashboards or reports.


Remember: Within Reason

It is important, as Edward Tufte repeatedly states, to do the erasure within reason. The target audience may actually need some extra elements when consuming the chart. Presenting a chart to an internal public in your company is very different to presenting to a general public. If your target audience isn’t familiar with the concepts involved or you have a more complex insight, you should add extra elements to guide them, i.e. arrows or text boxes. While editing your chart, the best advice is to always put yourself in the shoes of the audience.

A good way to double prove your final chart is to reach for a colleague or friend to read the chart. See if they still can fully understand the message after all the erasure has been done and listen to the feedback.


Summary of Examples: Charts Comparison

Source: Edward Tufte
Source: Edward Tufte

Modern BI Platform for Analytics Engineers.

Write analytic code, check into Git version control. Apply software engineering principles to analytics development.

Register Early Interest

"We're moving towards really becoming more data-driven with tools like Holistics where we can enable people to use data, instead of relying on a whole department of analysts, data engineers, BI people to do it for you." - Senior Intelligence Analyst