@aru_sas I think variable mapping is still one of the very human elements of data. In general, I think more is better in the data collection and structuring phase; any trait that is directly related to your primary key should potentially be included. You can always leave out data in a visualization.
@Houngbo - For very large datasets, I would recommend animation if there is a time series component or a word cloud if it is language based. It is very hard to use a static visualization for large datasets without getting overwhelmed and drill down often can get very granular and leave you 6 levels down trying to look for something,
@dslakeland With Stephen Few, I tend to think of this work as focusing on simplicity. I'm a big fan of simple and practical visualization in general. But it's important not to confuse simple with sophisticated and cutting edge. One can still use animation, 3-D, mapping, or anything else and still have a conceptually simple visualization.
@dslakeland Wayne at least used to have this view of data analysis vs data reporting: analysis is for discovery while reporting is more static. I think the two worlds are coming together and that reporting should have an analysis or interactivity aspect to it. I'll be honest; I haven't spoken with Wayne recently about visualization, specifically.
Second, remember that visualization is a first step we're taking into data interaction. We are just starting to learn how to really use data, so keep your eyes open on how you can take the next steps into data immersion.
@BethSchultz - For keeping a visualization simple, the key is in keeping the view simple and relevant: Rule of 7 and using the appropriate visualization type. Let the end user dig in as much as they want, but don't make their view so cluttered that they can't understand what they see.
Hyoun, I create visualizations, but rarely get feedback on what I create. I get better suggestions from my wife than my customers! Any suggestions on how to force feedback as part of the process? In the corporate world, they are usually just happy to have data to look at and rarely refine the first draft.
Thank you Hyoun for the presentation. Most of the analysis I do can be shown in a bar on line chart. It can be tempting to "jazz up" the charts by using 3d. I will not do even think about doing that again :-)
@pkilaruoud The GIS tools are great for demographic work and as a starting point for coordinate-based visualization. Because they have to render information at such a granular level, they can often also be a good starting point for visual drill downs of information
Hyoun, during the lecture you talked about the importance of keeping the interactivity built into a visualizaiton simple enough. Are there any rules of thumb for knowing when enough is enough and the interactivity is getting too complicated?
@Houngbo - Visualization as a scalable capability requires developing a process to consistently translate data into visualization and it has to take into account: data management, BI, and front-end development. Ideally, you would develop front-end templates for your major use cases rather than simply reinvent the wheel for each visualization
@BethSchultz - different types of visualization should only occur if different types of analysis are taking place. If all analysis is categorical, sticking to a bar-based or word-based visualization is probably OK.
@Hyoun, in your comment to @sungkim you suggest figuring out end user requests for interactivity. That's an interesting point, I think. I would imagine this is less asking them how interactive they want the visualization to be and more about asking the views of the data they'd like -- and then figuring out the level of interactivity from there. Is that about right?
dslakeland - I think the future of visualization is about greater immersion. This is pie in the sky, but consider that we can smell certain scents at a part per billion. Why not use that as part of our understanding of data?
@Sungkim - For time vs perfection, the real keys are to to start with figuring out scope, end user requests for interactivity, and a basic viz structure. I've always worked in enterprise environments where time is of the essence. Also, don't be afraid to hand off a data-ready framework to a front-end developer who can jazz up your work.
Hyoun, one of the questions I have deals with the level of interactivity to build into a visualization. I would think this would sometimes vary by the intended user -- which could get problematic if more than one user type (executive board vs. business managers, for example) are viewing the same visualization. Do we sometimes need to create more than one version of a visualization to provide different levels of interactivity?
The end user like our organizaton still have a way to go to develop great interactive visualization on our own. The "7" idea is a good reminder to keep the category numbers down for viewers to understand better what we're telling in the charts
I certainly grew up with the line and bar charts and tables back in the early days of the Apple computer which at the time we thought was truly amazing as data could so easily be manipulated by novices and viewed via very simple chart designs.
Participants, please note that the audio player for today's lecture will appear in the window above at start time. If it doesn't start automatically, please click the left arrow. If you have troubled with the audio stream, you might try another browser.
Today's All Analytics Academy class will be starting at the top of the hour, at 2 p.m. ET. Joining us will be Hyoun Park, who is a founder and principal consultant at DataHive Consulting. I'm looking forward to hearing what he has to say about creating interactive data visualizations, as I know visualization is a particular passion of his.
So, this is actually a combination of several challenges. One is around the initial development of relevant data access through appropriate data structures. One is around coordinating both developers and analysts to have access. And one is around convincing other stakeholders regarding the value of data experimentation.
I'll talk about some of the pitfalls that prevent dynamic data visualizations from being adopted and how to get consensus towards becoming more data-oriented as an organization. The convincing process and politics can often be more challenging than the actual work!
Full understanding of the capabilities requires full access to production data logs from different sources.
This access remains a constraint to developers and analysts.
Is there any good way to convince how experimenting with the data can be beneficial?