Data graphing is an information system.
The input is factual tabular data that is truthful and unambiguous. The output is the graphical encoding of this data that imposes a subjective interpretation on this truth that inescapably introduces a certain degree of decoding error.
The information produced by this interpretation depends on the accuracy in decoding the graph. Therefore, unlike the data, the graphical output is subject to an estimation error from the user decoding perspective. This error can be quantified in terms of how accurately we decode the key objective due to the encoding choices and the effect of our limited visual perception.
Jacques Bertin was the first to develop a cohesive theoretical framework for data graphing, and collected his thoughts in the impressive 1967 volume on the Semiology of Graphics. This work is the main source of inspiration for the Graph Workflow model. One of Bertin’s most important contributions to the theory of graphics is the separation of quantified information (data) from the method that is employed for its visualisation, hence the description of data graphing as an information system.
Tables to graphs
Humans have not evolved to extract information by reading data tables. We have evolved to understand visual information.
Everyone must have heard the proverb that “a picture is worth a thousand words“, or a variation thereof. This is wise proverb because it reflects a biological reality. It is claimed that humans dedicate more than 50% of their brain’s neuron activity to process visual information. We process nearly 10 million bits of visual information per second, and pre-attentive processing decodes information with high accuracy within 250 milliseconds (Healy and Enns, 2012).
As an example, consider the following table of iPod volume of sales during 2002-2015:
Give this table your full attention and try to answer the following questions:
- What information can you decode from this table?
- How quickly can you decode information?
- How much confidence do you have in the accuracy of the information that you have decoded?
- Will you remember this information after 10 minutes? How about after 1 day?
Now, repeat the same exercise but now looking at the following graph that encodes the data in a visual manner (right-click to open image in a new tab and enlarge):
We can now decoded information not previously seen in a table. Specifically, we can see the full shape of the so-called ‘life cycle of a product’, from its development stage with nearly non-existent sales to its growth period, saturation point and decline. We can also see this peculiar hedgehog pattern, where the spikes represent the December shopping spree periods when people used to buy iPods for presents.
Importantly, it took us only seconds to decode this information, and we feel absolutely confident about the accuracy of the information. And I bet you that even after a week you will still remember this shape of iPod sales in great detail.
This analysis demonstrates the power of data graphs particularly by comparison to trying to decode information from tables.