Data Visualisation

Audience Feedback from Presentations given by the DataViz Working Party

During 2019, the DataViz working party have given several talks to various IFoA audiences. These built on the presentation that the working party gave at the IFoA Life Conference in Liverpool in November 2018. We thought it helpful to summarise the audience feedback that we have received and the types of questions that have been asked.

We have presented at meetings and gatherings of actuarial societies in Scotland (Stirling, Glasgow, Edinburgh), England (Leeds, Leicester), and Switzerland (Zurich) - to approx. 200 people in total.

Presenting the aim and scope of our Working Party to fellow actuaries has been a great opportunity and rewarding experience for us. We have particularly enjoyed the multitude of interesting and relevant comments and questions from the audience, and would like to share and discuss these in this blog post.

The comments/questions have been structured into three categories:

1. more general comments/questions related to the format or output of the Working Party or the interaction of the WP with other bodies;
   
2. specific visualisation-related comments/questions, which often came in the form of suggestions for a future blog post on a particular topic;
   
3. comments on already published blog posts.
   

1. General comments/questions about the Working Party

Actuarial Syllabus	We've had a comment that data visualisation should be part of the actuarial syllabus, and the question whether we will try to push for this.
Collaboration	A person has asked what engagement we have had with other members of the profession, or actuarial firms. We were also ask to liaise with other practitioners as actuaries don't have all the answers. We are in a regular exchange with the profession. We will soon also post a guest contribution from an American actuary who is not a member of the WP. Note that our blog is public, and we are happy to discuss questions and problems with anyone interested.
Promoting the blog	We were asked to promote our working party via Facebook or LinkedIn. Output of the working party - We were asked what the final output of the working party will be. Our main output is the blog, and if everything goes as planned, the content will eventually be transferred to the profession's website.
Supporting the profession	We were asked if we could help the profession with its public publications, e.g. communicating mortality improvements to the public.
User feedback	Somebody asked how we would know that our examples are "better", i.e. whether we have asked for user feedback.
User support	It was suggested that we could allow users to upload a dataset that they want to visualise but are having trouble with.

2. Visualisation-related comments/questions

General comment	The questions below contain many good suggestions for future blog posts. We will try to cover at least some of these topics in the coming months.
Animations	We have been asked by various people whether we could add a blog post on animated graphs. This is planned, as we also recognize this as something increasingly important
Visualisation software	Multiple people have asked to share information on the use of R, Tableau or Power BI. We have already produced blog posts about visualisations created with R, and we will try to comment on the use of Tableau and Power BI.
Visualising stochastic results	We have been asked multiple times to add a blog post on the presentation of stochastic results. This is in the pipeline, and we hope to have a blog post published relatively soon.
Visualisation of maps	A person has commented that a blog post about geographic-mapping data would be helpful.
Reliability of data/analysis	Someone has asked us whether we could comment on the reliability of the data/analysis/conclusions being communicated.
Good and bad practice	It was suggested that we share our thoughts on good and bad examples of visualisation that we commonly see in actuarial work.
Research into thought processes	Someone has asked us whether we are aware of relevant research into thought process / how the mind reads and analyses information that supports the kind of principles we are suggesting in our blog posts.
Well-known vs. novel visualisation types	We have been asked to comment on the importance of considering what the reader is used to - a simple chart that a reader is used to seeing, compared to an innovate chart that the reader isn't used to seeing.
Data labels	It was suggested that we comment on the use of data value labels in charts.
Colour schemes	It was commented that red/green colour schemes don't print well in black and white, and that we should advise on other possible colour schemes (e.g. red-blue).
Blind users	Someone asked how one could help blind users to understand a chart.
Questions one didn't have	Good visualisations can sometimes answer questions one didn't have, and we have been asked to give an example for this.
Other visual clues	Somebody has commented on the effectiveness of visual clues beyond colour schemes. This is a topic addressed in the 2014 SIAS paper of Julian and Paul (see Resources section of the Blog).

3. Comments on published blog posts

Client recommendations post	It was commented that far too many recommendations are shown, and that the example would work better with c. 15
Mortality improvements post	It was commented that the colour scheme was not ideal, and that we should consider CMI's best practice advice on heatmaps.
Correlation assumptions post	Someone commented that the example is very complex at start and still complex at the end. It would be helpful to have an example where we distil something complex down to something much more simple.