Data Visualization Tools

Editor's Notes

• #5: SPEAKER: Karen 3 minutes Tech changing our work – potential to increase reach, quality, efficiency, make our orgs more effective. And we know that in order to succeed we need to assess our needs, choose the right tools, implement them, and make sure everyone is using them well – but that’s not something each of us has to do alone. Pro bono service is one source of help; it can help you take advantage of technology at your organization and to ultimately help you better meet your mission. 1. At the end of this webinar, we want you to feel excited, confident and prepared to use pro bono service to address your organization’s technology challenges. 2. To do this, we’ll look at common tech needs among nonprofits, general ways pro bono can help, and existing tech pro bono programs 3. Then we’ll look at some specific examples of how nonprofits are engaging professional pro bono expertise already to solve their technology challenges 4. – including some new types of projects you might not have thought of. 5. And finally, we want you to feel equipped to take action, so at the end of the webinar we’ll provide some specific next steps you can take to maximize the benefits of tech pro bono service at your organization. Before we dive in, let’s take a moment to get to know each other (poll).
• #6: Data is everywhere– tables, numbers, charts, in everything we do. It can be overwhelming. What do we do to make it less so?
• #7: The answer is not just to make it better looking. Lots of pretty things are pretty useless. On left (anonymous to protect the guilty) something that uses people to represent size… but should we be considering width as well as the height of the people? If no, then proportion is way out of whack (area vs. just height); if so, the space under the arms makes them seem like they’re smaller than they are. Regardless, there’s no data benefit in this being people rather than bars, in a regular bar chart format. On the right, a travesty of a pie chart. Pie charts in general aren’t a good thing unless you’re literally comparing only two or three percentages. Here, dimension is added to a horrible new impact. Look at the pink, orange, yellow, and purple slices of the pie– how big are they comparatively? Turns out that the first three are identical and the purple is more than 2x as big. Who knew. A simple bar chart would make this instantly clear.
• #8: An oldie but a goodie. Important data presented simply and powerfully. How many kids are there in California of pre-school age? How many are enrolled? You can see at a glance powerful facts like– there’s more Latinos in pre-school than white kids. That might appear to be a good thing. But wait– there’s so many more Latino kids in CA that the *proportion* is way under what we’d all likely want it to be. You may look at this chart and feel like it’s not “sexy”, not jazzed up enough. But the point is to make your DATA do the talking, not your graphic design. You don’t try to make your point more clear in a report by jazzing up your font.
• #9: First things first– what are you trying to do?
• #10: The first thing to decide with a viz is what point you want to make… but you may not know it yet. Are you looking for patterns? If so, cycling through a number of formats can be useful to see what trends there are. For instance, here’s a scatter plot that clearly shows a trend between price and age. A line chart is a great way to look for cyclical patterns– there’s perhaps a repeating pattern in the downward spikes on this, but you’d want to dive in deeper. Anyway, this is the stage at which you identify what the data has to say.
• #11: A dashboard would be a classic case here– you’re not trying to change people’s minds, but rather to inform them with the data http://www.dimins.com/wp-content/uploads/2016/01/outcomes.jpg
• #12: Much of the time, you’re presenting a visualization because you want people to draw a specific conclusions. Define them succinctly and make sure your graph/ chart backs them up.
• #13: Once you get beyond the exploration stage, you’re then looking to use the data to communicate with others– and that means you need to think about what you’re trying to say and how to say it
• #14: Whether you’re presenting info in a neutral way or making a point, each visualization should be a clear answers to a question. It’s not just “here’s some data”.
• #15: Make sure your facts are real facts and not “alternative” ones– either by collecting the data yourself or making sure they’re from a reliable source. Include the source in your graphic, so people can judge the credibility for themselves. And don’t pull information from multiple sources together into one– for instance, if I find in one source that “women” are on average 150 lbs and in another that “women” eat 150lbs worth of cheeseburgers, it’s questionable to say that women eat their weight in cheeseburgers on average. “Woman” are likely to be defined differently have different samples. (A similiar type of thing is actually done in advocacy, to sometimes good effect… but it’s not good data practice. You’d want to caveat it a lot. Maybe say that one study says X, another says Y, implying that women eat something like their weight in cheeseburgers I just made up the “meme” image. I thought it was funny– see if it works for you.
• #16: Think through the different ways you could represent your data – likely in your head or on paper rather than with a tool. A tool isn’t going to tell you the right way to go about answering your question– it can only do what you tell it. Often, you’ll need to slice and dice the data differently – for instance, to add a % change column, or remove some years– to effectively answer your question and only the question. For instance, back in the LSC graph we looked at, they’ve choosen to show seven years– enough to show a clear trend, but not so many as to overwhelm. And they’ve also chosen to show a graph of the percentages rather than the actual total. If this showed *number* of probono vs *number* cases, the increase trend would be essentially lost in the noise. In fact, the the raw number of pro bono cases could actually be going down even though % is going up—that tells a different story.
• #17: We’ll look at actual graph types in just a few minutes, but something critical to keep in mind– there’s seminal, accepted research on what’s easiest and most accurate for your readers. (walk though) these don’t immediately translate into graph types, but the top elements would put bar charts and line charts near the top, with pie charts and anything that makes people calculate area -- -or worse, 3-D volume or even curvature– much less desirable.
• #18: “experiment” doesn’t mean “contort your data to make it say what you want”. For instance, here’s a typical one– you can make the same trend look huge, as per the left, by making the axis not start at zero. If it correctly starts at 0, the trend is more subtle– but accurate.
• #19: Credit: Dr. Thomas Jackson, Graphing Charts, Slideshare : https://www.slideshare.net/lyndamk/graphing-charts-dr-jackson-presentation Complex presentations of data can be just the thing if your audience (say, your organizational staff) has the right context. For instance, if they’re looking at a dashboard every week, it makes sense for that to be a lot more complicated than if you’re trying to convey a one time point. The same is true of a complex map. This map of Gore vs. Bush districts is going to be nothing but a bunch of red and blue dots to most people from outside the United State– it doesn’t say hardly anything at all. But to experienced advocacy people, it tells a strong story about Bush’s widespread support.
• #20: Pie charts are evil– nothing they can do that bar charts can’t do better. 3-D is evil. Never use it. Just try comparing the size of social insurance payments from3-D chart chart to another, and then the bar chart. A good graph shouldn’t make you say, hey, look at that nifty graph! It should call attention to the DATA.
• #21: This is an old default format from Excel--- new ones are way better, but it’s a good example. If you can remove it, do– for instance, what does the gray background add? The dots? Labeling every year? Line around the key? We could do better– for instance, putting the key on the line itself. Graphic designers talk about this as “saving ink”– what could you represent with white space as well or better than with something that uses ink?
• #22: To create something useful, it’s critical to have a sense of what might work, and the options that are at your disposal. Keeping an eye out on what visualizations people use can be helpful for spurring your own creativity– but also to remind you that the same types of charts are used over and over and over again for a reason. Because they work.
• #23: https://lafla.org/wp-content/uploads/2016/09/LAFLA-2015-Annual-Report-Final.pdf Just showing some numbers in a graphic way can be a straightforward way to communicate a data point.
• #24: https://www.justice4all.org/2015/01/08/report-achievement-gap-widens/ It’s usually clear cut if you should use a line chart– are the items in the X axis ordered in an important way? By time, for instance. The line chart shows trends – like here, where you can see simple but powerful trends over the last six years.
• #25: If the information isn’t ordered, a bar chart is the logical thing to consider next. They’re a really versatile and useful way to display things. For instance, here a chart from Utah Legal Service’s performance reporting dashboard, showing open cases that haven’t been updated for 30, 60, and 90 days, as well as (with the line) showing the number of cases the attorney is taking on. Randall (not anyone’s real name) is clearly not updating his notes at the same frequency as everyone else…
• #26: The brain is really bad at comparing the areas and angles of pieces of things– so for instance, this style of pie chart works simply to illustrate male vs. female ration, and possibly would work for three races on the left.. But it completely falls down at representing all seven races included here. Even for the little things– compare the latino to Caucasian slices. If you took the numbers off, could you really tell how much bigger the Caucasian slice is?
• #27: Tables can provide a quick summary if you know what you’re looking for, and can be really useful for internal reporting. There’s a reason that no one looks at accounting statements as graphics. You just have to be careful not to overuse them.
• #28: For instance, you see in this example that in general, the higher the “quality of health care”, the higher the “community well-being” is, however those measures are defined. You could see that trend on a scatter plot, but a bubble chart allows us to put in a third dimension, in the size of the bubbles. In this example, it’s not labelled what it is, but it looks like a another health indicator– maybe, say, number of doctors per capita. You can see that trend along with the others. As we stare at this, we can see that this particular chart isn’t great at answering any one particular question– it’s more of an exploring chart than a communicating chart. https://powerbi.microsoft.com/en-us/guided-learning/powerbi-learning-3-7-create-scatter-charts/
• #29: Maps can be a powerful way to illustrate a point, but going beyond the simple can be a pain. This is called a choropleth map; you could also consider a “pins on a map” model
• #30: As someone who does a lot of data visualization (Laura), it can feel boring to me to use a bar chart over and over again, and it’s tempting to move beyond to something more interesting. But it’s not about entertaining myself, it’s about being clear. Using the same graph formats over and over have the benefit that people know how to read it. Your data is rarely so important that it makes sense to make someone figure out how to read a new format.
• #31: So for all of these possibilities– do you need a fancy tool? Or will a tool you almost certainly already have– Excel– work?
• #32: The old standby– we likely already have it. For 80% of nonprofit data visualization needs, Excel is probably going to make sense. Way more of the time that you might assume. We’ll only spend a little bit of time on it, as you’re likely pretty familiar, but if you feel like Excel is meeting your needs perfectly well, you may well be right.
• #33: Pretty much everything is customizable in Excel– it’s really unlikely that you actually CAN’T CREATE a chart that it makes sense for your to make, or you can’t make it look good. All sorts of things you’d think were impossible are doable. However, it’s not easy to use or find at all—you’ll likely need help from tutorials to figure out some of it. It’s dramatically more customizable than Google Drive/ Sheets It took me about three minutes to go from the “fine” graph on the top to the more snazzy one on the bottom, with customized colors, fonts, playing with white space and lines, … but I knew exactly where to find all of that stuff
• #34: A few years ago, Excel was missing some charts that you really might want– scatter plots, say, or histograms. They’ve really upped their game in terms of charts, though– now they’ve really plugged their gaps. I don’t think there’s any chart that I’ve ever created (this is Laura) that I now couldn’t create in Excel. But some might take me some time to sort out, even though I have a lot of experience. Scatterplot: http://www.databison.com/scatter-chart-with-highlight-box-to-group-data-points-in-chart/ Box and whisker: http://www.myonlinetraininghub.com/new-charts-in-excel-2016 Bubble: http://www.s-anand.net/blog/motion-charts-in-excel/ Histogram: http://www.tech-recipes.com/rx/32068/create-a-histogram-graph-in-excel/
• #35: KG: Also: you have more data than can fit in Excel? LQ: This isn’t a likely consideration for a single data set (Excel will handle millions of rows these days)…. But certainly applies if you have more than one. I’ve added that
• #36: So if those reasons are true…
• #37: Has about 30 templates with the model that you choose your template, can see the demo data, and can see exactly where your data should go to make your graph look like theirs. Not nearly as customizable as Excel, but quicker to make something better looking if you want to do something it wants to do. Note th at you can only embed with the free version– you can’t get it out as a image file. You need to pay to do that.
• #38: Tableau is getting a lot of buzz– it’s a very powerful desktop tool that’s specifically designed for allows people who are up to speed with the interface more ability to explore easily, has charts that perhaps look a little better out of the box, and allows you to share has enormously discounted licenses available through TechSoup. Could be a compelling alternative to Excel for the first place to turn to explore and create visualizations– probably easier to use.
• #39: Or if you want to create dashboards that are robustly updated over time, maybe from different sources pulled together, that’s something that Tableau is great at, and Excel is not good https://www.sdbor.edu/dashboards/Pages/High-School-Transition-Report.aspx
• #40: Relatively new entry from Microsoft, specifically taking on Tableau. It’s quite comparable– it allows fairly easy data exploration, has charts that look pretty good out of the box, good for online dashboards and interactive charts. It integrates smoothly with Excel and also works well with other data sets. And it’s free unless you want to do something really complicated. Tableau is getting a lot of buzz, and Power BI less so, but it’s unclear if Tableau has just been around longer. Power BI is certainly free-er https://powerbicdn.azureedge.net/mediahandler/blog/legacymedia/6740.image007_5F00_Fields-Product.png_2D00_550x0.png
• #42: Tablean and Power BI have particularly compelling pricing for nonprofits, but they’re not anything like the only tools in this space. In particular, Plot.ly is talked about a lot. Periscope, QlickView and HighCharts are a few more of the many tools in this niche
• #43: There’s a lot more support for maps than there used to be. Tableau and Power BI in particular offers some compelling support, but Infogr.am will also allow you to do some nice visualizations. Depending on what you’re doing, also remember that you don’t necessarily need a charting tool to color code a map. If you can find a workable graphic file that delineates the areas that you want to color code, any illustration tool – including Paint– will allow you to fill them in. https://canvas.uw.edu/courses/966147/assignments/2752923?module_item_id=5811462 http://www.smh.com.au/lifestyle/health-and-wellbeing/dental-costs-for-your-suburb-revealed--save-yourself-hundreds-by-shopping-around-20170131-gu21gp.html
• #44: So if those reasons are true…
• #45: Pros: what designers use, full-control Cons: cost?, need to know what you’re doing
• #46: https://piboonrungroj.files.wordpress.com/2012/06/rstudio-windows.png These languages support sophisticated analysis of data as well as the creation of visualizations.
• #47: https://anneladyem.files.wordpress.com/2014/11/screen-shot-2014-11-11-at-12-46-55-pm.png