1. Visualizing Collections Data Why Pie Charts Aren't Always the Answer

2. Disclaimer Data can help tell a great story, but data alone will never tell the whole story.

3. The Deal TRLN E-Journals Shared collection – no overlapping titles between institutions Assessment and decision-making at the local and consortial level E-Books Holdings different for each institution UNC holdings go back to 2005 Assessment only looked at 5 years of BR2 reports (2010-2014)

4. The Underlying Dataset - E-Journals Maintained in Access DB created by NCSU staff Dataset components – title metadata from various sources, 6 years of cost and usage data, 35 columns We layered in additional fields to classify e-journals value

5. Determining Value Likert scale for e-journals good acceptable problematic low value unacceptable Various combinations of usage and CPU for each category

6. Question #1 How do UNC-curated titles scatter across these categories across TRLN?

7. Question #2 How do all the titles in the collection scatter across these categories for TRLN?

8. Question #3 How did the titles scatter across these benchmarks in 2010 vs. 2014?

9. Question #4 How many e-books were used year over year?

10. Question #5 How does the ratio of books used vs available change over time?

11. Question #6 How many books had 5+ chapter downloads, year over year?

12. Best Practices Pies and human perception The Data-Ink Ratio Visual Math

13. Pies Question: How do all the titles in the collection scatter across these categories for TRLN? Issues: Everything is treated as a proportion Values only available via labels Hard to compare areas/angles Gets confusing past 4-5 categories Stephen Few on Pie Charts: Save the Pies for Dessert

14. The World’s Most Accurate Pie Chart http://visual.ly/literal-pie-chart

15. Idea: Accuracy and Human Perception Basic rankings derived from experimentation in McGill and Cleveland 1984 We can use these rankings to assess whether a given graphical form is more or less effective than another at communicating accurately perceived values to the reader. However, context and audience as judged by the designer, can overrule these rankings. Maximizing visual accuracy doesn’t have to be your primary goal. Mackinlay, J. (1986) Automating the design of graphical presentations of relational information.ACM Trans. Graph. 5, 2 (April 1986)

16. Pie Alternatives QualityTitlesPercentage Good81765.9% Acceptable15912.8% Problematic1139.1% Low957.7% Unacceptable554.4%

17. Backgrounds Question: How does the ratio of books used vs available change over time? Issues: Hard to read values on dots Data doesn’t contrast highly with background

18. Data-Ink (Edward Tufte)

19. Stacked Charts Question: How did the titles scatter across these benchmarks in 2010 vs. 2014? Issues: What does the overall height mean? How many titles were unacceptable in 2014? Visual Math Before:

20. Alternatives: Height always encodes single year Slope of each line emphasizes different rates of change

21. Tool Landscape Spreadsheets In-browser tools Business Intelligence Tools Coding Design

22. Spreadsheets e.g. Microsoft Excel, LibreOffice, Open Office Pros: You probably already have it Your data probably passes through it already Secure Already integrated in workflows Cons: Software not primarily designed for visualization Static and local

23. In-Browser General: Plot.ly, Datawrapper, Raw, Timeline.js Mapping: ArcGIS Online, CartoDB Pros: Often easiest, most accessible, quickest Often free or cheap Many tools available Specialized tools like ArcGIS Online Cons: Most subject to change (or disappearance) Inflexibility Specialized functionality Strict data format needs Dependence on other software Too many options Full benefits require a more advanced tool ArcMap QGIS Plot.ly ArcGIS Online

24. Business Intelligence e.g. Tableau, Qlik, SAS Visual Analytics Pros: Flexible, but don’t require much if any coding Point and click interfaces Good support/frequent updates Some free public options Cons: Most expensive IT support for large implementations Business-oriented user communities https://public.tableau.com/s/gallery/fatal-drug-overdose-rates-united-states Tableau

25. Coding e.g. JavaScript(D3.js), R(ggplot2), Python Pros: Generally Free If you have the time to learn it Most flexible and powerful Cons: Multiple languages necessary Need to hire developer(s) Time-intensive http://bl.ocks.org/mbostock/4060954 D3.js

26. Design e.g. Adobe Creative Suite, Inkscape Pros: Most aesthetically oriented Can be combined with other tools Cons: Expensive Not data-oriented ‘Infographic effect’ Static http://icharts.net/blogs/2013/spotlight-interview-unique-approach-infographics-journalism-alberto-cairo Adobe Illustrator

27. Learn more: Theory Practice Edward Tufte: The Visual Display of Quantitative Information (2001) Visual Explanations (1997) Envisioning Information (1990) Colin Ware: Information Visualization: Perception for Design (2004) Stephen Few: Show Me the Numbers (2004) Information Dashboard Design (2006) Now You See It (2009) Alberto Cairo: The Functional Art (2012)

28. Learn more: Tools Tool lists: http://dirtdirectory.org/ http://selection.datavisualization.ch/ Map Galleries CartoDB: https://cartodb.com/gallery/ ArcGIS Online: http://www.arcgis.com/home/gallery. html#c=esri&t=maps&o=avgrating Sample Galleries D3: https://github.com/mbostock/d3/wiki /Gallery Tableau: https://public.tableau.com/s/gallery Plot.ly https://plot.ly/feed/