We need to use data throughout the improvement journey.  It is important that we show the data as effectively as possible to create true understanding of the system, to communicate with and engage the improvement team and wider stakeholders, and to tell the story of the journey – which can be very important to help with spreading a successful change.

What’s your objective in presenting your data?

1. Influence —to make a point or convince people
2. Explore—tap into subject matter knowledge in team and seek better understanding 
3. Inform —summarise data to tell a story, or so that others can use it

Depending on which of these objectives is most important, the relative priority of clarity and accuracy may vary.  Consider the two pictures of a heart.  The left one is clearer (although less accurate) while the one on the right is more literally accurate (although less clear).

For example, if you want to convey a message or seek an explanation, often the pattern in the data is the most important thing to show.  Clarity is therefore more important than precision/accuracy.  Alternatively, if you need people to know the actual values in the data, it may be best to present them in a table.

The human brain has an amazing ability to recognise patterns. If we can harness the power of visual perception, we can make it much easier for people to take useful information from the data we have gathered and analysed. They can then make changes and take other actions for improvement. 

In order to do this, we need to apply what we understand about visual perception ‘through design principles and practices that are aligned with the way people see and think’ (Stephen Few).

Research shows that the perception of  values or difference among them is  affected by the way data is presented. As  the figure shows, in order of effectiveness  key factors are: 

1. Position along a common scale ( [a]) 
2. Position along non-aligned scale ( [a] v [b])
3. Length [c] 
4. Slope/ angle [d] 
5. Area [c] 
6. Volume 
7. Colour [c] 

Other “pre-attentive” (an early stage of subconscious visual processing) attributes of visual perception as described by Ware are: 

• Intensity or hue of colour  
• Form (orientation, line length, line width, size of symbol, shape of symbol, added mark, enclosure)  • Motion (flicker) 

Following a rational set of design principles, based on an understanding of visual perception, will help you ‘let the data speak’.  

Walter Shewhart’s rules for presenting data: 

• Data should always be presented in such a way that preserves the evidence in the data for all the predictions that might be made from the data.
• Whenever an average, range or histogram is used to summarise data, the summary should not mislead the user into taking action that the user would not take if the data were presented in a time series.

Another highly influential figure in the world of Visual Display of Quantitative Data, and author of a book of that name, is Edward Tufte. He has laid out a series of principles including: 

Use the least ink to present the greatest amount of information in the smallest space

• Sometimes graphs can be smaller if the intention is to show a shape rather than allow people to read off values. Small multiples can be useful. 

• Lie factor—practise with substance and integrity: don’t mislead by their way you scale, sample or show frequency. Use equal time steps on time series and show missing data points. 

Other principles to consider include: 

• Clarity —patterns/values a chart depicts should be as easy as possible for a reader/viewer to interpret. In general, simple charts are more effective. 
• Self-sufficiency —aim for simple charts that do not need labels, values etc. to be printed on data series. Using a standard repertoire of charts aids familiarity (they then need no explanation). 

Use different graphic tools depending on which you want to show. There are 5 basic types of data displays 

• Show relative frequencies across discrete categories - bar chart, Pareto chart 
  • Use Pareto chart when you want to clearly show the most common categories

• Show how a continuous variable is distributed - histogram

• Plot data to show a relationship - scatter plot

• Plot data over time  - line graphs; run & Shewhart charts 
  • Use line graph for a simple display over time
  • Use run chart to distinguish random and non-random variation and where you don’t have enough data points for a Shewhart chart
  • Use Shewhart chart to distinguish common (intrinsic to process) and special cause variation, and assess process stability and capability
     
• Plot data by location - map 

Title 

This should clearly describe the chart—including what, where and when in large clear font, e.g. Arial—and allow it to stand alone. 

Axes 

• Use clear labels in readable (same) font.  
• Suppress unnecessary decimal places.  
• Where values are missing this should be shown.  
• Don’t allow percentage charts with a vertical axis that goes past 100%.  
• Simplify the scale—include (e.g. thousands) in title. 
• Dual horizontal axes are confusing and should be used sparingly. Prefer two charts side by side.  

Colours 

• Be consistent & clear. 
• Minimise number of colours used. 
• Consider that there may be colour blind individuals looking at your chart, so use colours with care (avoid using red and green to differentiate). 
• In line charts use different types of lines so they can be distinguished if printed in grey scale. 

Legend 

• Consider omitting the legend and labelling directly. 
• Don’t let legend steal space from the data.  

Layout 

• Don’t put too much data on one chart; use small multiples (same format, same scale).  
• Remove unnecessary borders, backgrounds and grid lines.  
• Consider omitting axis lines on horizontal charts.  • Don’t use 3D graphs unless have three variables. 
• Default orientation should be landscape.  
• People judge variation in slope best if roughly 45 degrees.  
• Ordering of columns/bars improves the display of a distribution and gives a ranked list. 
• Include data source.  
• Avoid distracting graphics or animations  

• For run and SPC charts: 
  • annotate with events and changes so relationships can be seen; 
  • depending on whether appropriate for your audience, say what type of chart (run, I, P, C, U etc) & when updated and by whom; 
  • clearly label means, medians, baselines, control limits; 
  • if using side-by-side charts for comparison use same scales on y axis; 
  • where a time series has missing values for certain dates, or gaps between data vary, the chart should display this. 

Scaling 

• Scale axes appropriately so they don’t deceive—charts show patterns and people may not read the values (Tufte’s ‘lie’ principle). 
• Include an inset with a shortened y (vertical) axis if you want to make a specific point.  
• If you must cut the y axis indicate with a zig zag or two parallel diagonal lines.  
• Avoid large areas of white space unless the intent is to draw attention to stability/lack of variation. 
• For bar charts start the scale at 0 (or just below the lowest bar if negative numbers) and continue just past highest bar. 
• For scatter plots scale to the data so you minimise white space.  
• For run and SPC charts: 
  • the data should occupy about half of the graph’s vertical scale with the remaining half of the scale as white space; 
  • if data can’t go past an absolute value (e.g. percentage values of 0–100%) then the vertical axis shouldn’t either;  
  • always use rounded scale divisions, and whole numbers when possible. 

Default Excel charts commonly violate many of the design principles, but simple changes to the default graphs automatically produced can make big improvements. 

• Make sure your charts are clear and readable – follow the tips above.
• Set the context before you reveal the data.  Using stories to engage people can be really effective.
• Explain the dimensions of the data (axes) and what’s been plotted.
• Interpret the chart explaining the key messages.   It can be helpful to think of  GEE when talking about your chart.  This stands for 

Generalisation Summarise what the chart is saying (and what it is not saying!)

Example Point out data that illustrates the generalisation

Exception Draw attention to any exceptions and explain

Office for National Statistics guide to data visualisation

Data Quackery by Mike Bell

Government Statistical Service guidance

Article on data storytelling