Skip to content

How can we make data easier to read?

A wooden ruler.

Data is just a collection of emotionless numbers, but we can use it to tell a story. And to produce a good chart or graph we must know what story we are trying to tell.


(This is an expansion of a discussion I had with some colleagues at work.)

In the United Kingdom there was a big controversy when the exams were cancelled in 2020 and replaced with teacher assessments and a standardisation algorithm.

The algorithm was introduced to make sure that grades stayed on the same levels as previous years. Many grades were adjusted to great dissatisfaction among both students and teachers.

The best way to understand what was going on was to create some charts and graphs.

My workplace, Tes, which is an important news source for teachers and schools worldwide (and UK in particular), also contributed with many articles on the subject and some graphs were also produced.

These are the graphs:

Two charts. One describing how A and A-star grades would have been 10% above the normal level if they had not been adjusted. The other charts shows the percentages of all grades that have been adjusted: 58.7% not changed; 35.6% were lowered by one grade; 3.3% were lowered by 2 grades; 0.01% were also lowered by 2 grades (this is an error); 0.2% were lowered by 3 grades; 2.2% were raised by 1 grade; 0.05% were raised by 2 grades.
The original charts produced by Tes. Open image

It is quite obvious that these graphs were produced on a very tight deadline because they contain errors:

So, my team had a short discussion about this that ended with: So, how would you have done it?

The below was my answer:

Grade adjustments made by exam boards for all grades: 58.7% unchanged; 2.2% up 1 grade; 0.05% up 2 grades; 0.01% up 3 grades; 35.6% down 1 grade; 3.3% down 2 grades; 0.2% down 3 grades.
My chart showing grade adjustments made by exam boards for all grades. Open image
A and A-star grades as a percentage of all grades: 26.2% in 2017; 26.2% in 2018; 25.2% in 2019; 27.6% in 2020. It would have been 37.7% in 2020 if the grades had not been adjusted.
My chart showing A/A grades as a percentage of all grades. Open image

My process

A tale of two stories

First of all, I separated the charts. The original graphic tells us that there is one story about the A-levels, but I believe there are two:

I thought this would be better understood if I created separate graphics. Then I could give each graphic a succint title. A good chart title helps the reader understand what story you are trying to tell with your data.

Keep it simple

In the same way that writing in plain English makes it easier for readers to understand the writer's ideas, plain charting does the same for data.

That means showing only the minimum needed in understanding the data. Which is delightfully explained in one of my favourite quotes I use for design:

Everything should be made as simple as possible — but not simpler.

(Ascribed to) Albert Einstein

Less colour is the new black

The original graphic uses a wide colour palette together with a variety of shapes and type sizes. Most of them seems to do the job of being "on brand" rather than helping to understand the data. I reduced them to a minimum (while still using brand colours):

The right chart for the job

The grade adjustments were originally shown as a donut chart.

Pie and donut charts are a popular solution for containing data in a limited space. But they are not always the best solution, because:

I thought the best chart for the job would be a good old–fashioned bar chart:

20 to 40 in 60 seconds

The original graph for the A and A distribution shows the level on a y-axis from 20% to 40% percent rather than a full range from 0 to 100%.

I am generally not in favour of shortening the y-axis as it can lead to mis–representation of the data or deliberately misleading graphs.

Showing the full y-axis from 0 to 100% illustrates the level of A and A grades have been stable in the last four years. And even if there would have been no adjustments made to the results it would not be as wild a fluctuation as the original graph (with its abbreviated y-axis) led to believe.

(Update: The Tes chart was replaced with OfQual’s original chart of the same (and the source of the chart above) and is no longer available on

OfQual’s original image very similar to the one that Tes produced with a pie chart and a line graph. The colours on these charts are less confusing but it is still showing confusing legends and labels. And the Y-axis is also shortened on the line graph, which gives the impression of a wild fluctuation.
OfQual’s original chart. Open image

Further reading

Header image by Dawid Małecki