The Zen of Visual Analysis Chris Stolte Vice President, Engineering & co-founder
How do people reason about data?
The Cycle of Visual Analysis
2008 Presidential Election Campaign Expenditures http://www.fec.gov/disclosuresearch/mapappdownload.do
Not just Aha!
advice: Iterate, explore, p and experiment regardless of task
How can products support being one with your data and how can I use them effectively?
How can products support being one with your data and how can I use them effectively? Support the Cycle of Visual Analysis
Supporting the Cycle of Visual Analysis Incremental Expressive Unified Direct
Supporting the Cycle of Visual Analysis : Incremental Allow people to easily and incrementally change the data they are looking at and how they are looking at it click click
advice: Start simple
Supporting the Cycle of Visual Analysis : Expressive There is no single view for all tasks and all data
Supporting the Cycle of Visual Analysis : Expressive
advice: Reverse your thinking start with the data
Supporting the Cycle of Visual Analysis : Unification with the database Query many times and then generate a summary graph. TraditionalReporting Tools versus Query once and then iterate on the presentation of the data. Traditional Visualization Tools Iteratively change the data and image to find the perfect view. Tableau
Supporting the Cycle of Visual Analysis : Unification with the database Leverage the advances in Leverage the advances in fast database technology
Supporting the Cycle of Visual Analysis : Direct Interaction Make the tool disappear. Allow the user to directly interact with the data.
How is this possible? VizQL Imagine if databases shipped with only 5 prebuilt queries
How can products support being one with your data and how can I use them effectively? Generate Effective e Views of Data
Generating Effective Views of Data People shouldn t have to be graphic designers or psychologists
Generating Effective Views of Data An effective presentation of data: Communicates all of the data Communicates only the data Leverages the human perceptual system Is understandable Is interpretable
Generating Effective Views of Data Communicate all of relevant tdt data Bad Bad Good
Generating Effective Views of Data Communicate only the dt data Bad Bad
Generating Effective Views of Data Leverages the human perceptual system Bad Good
Generating Effective Views of Data Is understandable
advice: Keep your visuals simple
Generating Effective Views of Data Is interpretable
advice: Annotate but remain focused
Generating Effective Views of Data How does Tableau support generating effective views of data? Limiting the visual properties to a small and proven set Great defaults Small multiples Show Me and Show Me Alternatives Titling, captioning, and annotation
Generating Effective Views of Data How does Tableau support generating effective presentations? Limiting the visual properties to a small and proven set Great defaults Small multiples Show Me and Show Me Alternatives Titling, captioning, and annotation
Generating Effective Views of Data How do you pick effective colors? A line in HSV corresponds closely to a set of perceptually p equivalent colors but not quite. There is no good programmatic way to pick the colors so we hired a color expert to manually pick colors.
Generating Effective Views of Data To encode unordered data, we vary colors primarily in hue: To encode numeric or ordered data, we vary primarily in saturation and value: We discourage the use of more than 20 colors. Pick colors that can be named so people can talk about them that thatpoint the thecyan pinkishorangein theuppercorner corner We consider color blindness. Compare to Excel XP:
Generating Effective Views of Data Color for communicating data Color for aesthetics
Generating Effective Views of Data Automatic marks picks the best mark based on properties of the data.
Generating Effective Views of Data It is only possible to encode a limited number of dimensions in properties such as color, size, and shape.
advice: Use small multiples to simply and effectively see more data
How can products support being one with your data and how can I use them effectively? Be Easy and Simple to Use
Be easy and simple to use What do people mean when they say easy to use? Learnable Intuitive Efficient Direct Measuring ease of use is difficult: l A chart wizard is easy to learn and yet inefficient and hard to use.
Be easy and simple to use How do we achieve ease of use? Allowing incremental interaction Being direct Providing freedom Knowing and respecting our user Being simple
Be easy and simple to use Provide freedom No templates, no six month IT projects, no modes, no wizards Ask questions of your data and get answers!
advice: Empower your audience share interactive views
Be easy and simple to use Know and respect the user Focus the user interface on the content Support the user s task Allow the user to get into an uninterrupted t flow Provide great defaults
Be easy and simple to use Examples: A single toolbar with a limited set of muted buttons. Minimizing the non data ink. Unsaturated colors for the interface, saturated colors for the data. the user s data
advice: Stay focused on the data
How can products support being one with your data and how can I use them effectively? Generate Beautiful Results
Generating Beautiful Results
Who can Visual Analysis help? Everybody with data that is not getting answers
advice: Don t limit yourself to large data or special projects
advice: Seek outside inspiration
Summary Iterate, explore, and experiment regardless of task Start simple Reverse your thinking and start with the data Keep your visuals simple Annotate but remain focused Use small multiples Empower your audience by sharing interactive views Stay focused on the data Don t limit yourself to large data Seek outside inspiration