Before: Just the Survey Questions; No Graphs Yet!

A couple weeks ago I spoke to Harvard University graduate students about visualizing survey results. This not at all to very survey scale is quite common in my research circles so I’m sharing our ideas with all of you, too.

Here are some of the survey questions that were asked before and after program participation. Don’t worry, the online survey was formatted much more beautifully than this screenshot from the Microsoft Word version of the survey!

After: Stacked Columns for the Ordinal Survey Scales

I’m going to show you two makeovers. In both makeovers, I used stacked column charts to display these ordinal survey scales.

Why did I choose vertical column charts instead of horizontal bar charts? I displayed the patterns over time from left to right across the page. Before results go on the left and after results go on the right, so they get vertical columns (not horizontal rows).

Why did I display percentages instead of numbers? There were more than 100 survey responses, so I converted the raw numbers into percentages. When there are fewer than 100 responses, I display the raw numbers. It’s super confusing to talk about 16.56% of 14 people.

Why did I use three colors? I suggest that you color-code by category because it’s a great strategy for breaking up oceans of information into manageable chunks.

Why did I use different shades of each color? The very knowledgeable to not at all knowledgeable scale is ordinal so I used darker and lighter versions of each hue to correspond to the amounts of knowledge.

Why did I write titles and subtitles? I’m a visual person and prefer reading graphs over paragraphs but some viewers will prefer reading paragraphs over graphs. One of the most common mistakes I see among novices is that they focus so much on creating the graph that they forget about the paragraphs. Your viewers will benefit from having both.

Makeover 1: The Traditional Data Visualization Approach

Before you create any data visualizations, I suggest doing some upfront planning with your colleagues. You’ll want to discuss a few thought-starter questions in advance. For example, who’s your audience? Are you designing your visuals for a technical or a non-technical data audience? Are you making a slidedeck, a handout, a technical report, or some other dissemination format altogether?

Here’s the most important planning consideration in data visualization projects: Are your viewers expecting a story?

Sometimes your viewers will expect a traditional data visualization approach, in which they’ll (hopefully) read between the lines and uncover a takeaway message for themselves.

Other times, your viewers will expect a storytelling data visualization approach, in which you use a dark/light color contrast and explicit text to uncover the takeaway message for them.

Here’s the first makeover, which uses a traditional data visualization approach:

Makeover 2: The Storytelling Approach to Data Visualization

Stacked bar charts are one of the most common ways to display survey results because surveys often include scales like this one.

But, we have to be careful because one page with two points in time, three survey questions, and five options per survey question can get cluttered, fast!

In this version of the handout, I used saturation to guide the viewer’s eyes towards the very and moderate categories. In other words, those categories are dark while the other categories are light. This dark/light color contrast is called a preattentive attribute, and it’s one of the easiest strategies for telling a story with data. Your viewers don’t have to think about it or waste any of their precious mental energy. Their eyeballs are instantly drawn to the darker parts of the page.

In your project, you may choose to draw attention to the not at all knowledgeable category. There are several correct ways to guide eyes with saturation. Your job is to anticipate what your viewers will find most useful. Just use your best professional judgment.

Here’s the second makeover, which uses a storytelling data visualization approach:

Learn More about Visualizing Survey Results

In this blog post, we looked at visualizing how participants’ knowledge about historical events increased after participating in an educational program at a museum.

Here are additional resources that show you how to visualize survey results in Microsoft Excel:

• Agree/disagree scales
• Check-all-that-apply questions
• Demographic data 
• Workshop satisfaction survey questions

Bonus: Download the Materials

Want to explore these graphs in more detail? Download the Microsoft Excel spreadsheet and the Microsoft Word document that I used to create these handouts.

Purchase the templates

The post How to Visualize Pre/Post Survey Results in Microsoft Excel appeared first on Depict Data Studio.

Option A: Axis stretches just past the actual maximum value (30% in this example)

The biggest number is 26%, and the axis goes from 0% to 30%, just past the biggest number in the bunch. Your stakeholders will think the 26% looks enormous because it stretches all the way across the screen. Wow, look at our numbers! Great news! Of course, we’ll have to work on Category D, but we can certainly improve that number! Especially when the other categories are looking so good! Talk to your teammates: is this the message we’re going for?

Option B: Axis stretches all the way to the potential maximum value (100% in this example)

What if the numbers have the potential to stretch all the way to 100%? (The percentage of attendees who said they’d recommend your conference to a colleague, the percentage of students who graduated on time, and so on.) If you’re trying to hit 100%, now the 26% isn’t looking so hot. Argh. We thought everything was going so well! We’ve got so much to improve upon. Where do we even start?! Ask your teammates: is this the right time to risk overwhelming the stakeholders with so much bad news?
My advice: Good facilitation skills are an ingredient of good data visualization. I often begin with the first graph to avoid scaring stakeholders away from action. As we get to know each other  — and they get to know their numbers — I slowly introduce the second style.
When choosing minimum and maximum axis values, what other factors do you consider?

Download the template.

The post The Graph’s Scale: Actual Maximum Value or Potential Maximum Value? appeared first on Depict Data Studio.

Notice anything funny about the height of these bars?

The shortest bar, for 2012, represents 26,000 youth. But the tallest bar, for 2010, represents 27,000 youth. 26,000 and 27,000 are pretty similar numbers, yet they look reeaaaallllly different. The 26,000 seemed way too short compared to the 27,000.

This graph’s vertical y-axis doesn’t start at zero, so the differences between bar heights are exaggerated.

It’s okay to have a non-zero y-axis. But, it must be labeled.

In the after version, I added labels for 25,600, 26,000, 26,400, and so on.

We don’t want to mislead our viewers. We have to be clear that we’ve intentionally truncated that y-axis so that we could zoom in on that segment between 26,000 and 27,000.

Or, another option is to adjust your vertical y-axis so that it starts at zero.

Then, you wouldn’t need to have any axis labels off to the left of your graph.

The tradeoff is that, now, the bars are all roughly the same height… which might be okay, depending on what you want to emphasize. Maybe you’re trying to show that a consistent number of youth were enrolled in the study sample each year. In that case, starting the y-axis at zero would help you out.

Let’s look at those two “after” versions once more.

Option A, on the top right: If you’re going to start your y-axis at something other than zero, then you need to add axis labels.

Option B, on the bottom right: You can start your axis at zero and forego having any labels.

How have you handled truncated y-axes in your projects?

The post Be Careful with the Y-Axis… appeared first on Depict Data Studio.

These are made-up numbers, but inspired by a real project from a research organization I work with.

Before

Can you spot the fatal flaw?

See it?

Go check out the x-axis.

Where the heck are Days 5, 6, 7, 12, 13, 14, and 20?

Oops…

Yep, the analyst accidentally skipped a few labels along the x-axis.

This is an innocent enough mistake.

Most likely, there weren’t any responses to the survey on Days 5, 6, 7, 12, 13, 14, and 20. So the analyst was busy and forgot to manually insert “0’s” for those days in the data table.

This is what Stephanie Evergreen and I described in our Data Visualization Checklist: Axis labels are equidistant means that the spaces between axis intervals should be the same unit, even if every axis interval isn’t labeled.

After

Here’s what that graph should’ve looked like:

Skipping Dates = The Wrong Pattern

Can you spot the differences now?

The dotted line is the incorrect graph. The solid line is the correct graph.

I can’t tell you how many times I’ve seen this same mistake in published research and evaluation reports. The analysts have accidentally skipped days, years, cohorts, and so on.

How to Fix It

The good news: What an easy fix.

Just add new rows or columns to your data table, insert some 0’s, and voila! your graph will have equidistant axis labels.

The post Why You Shouldn’t Skip Dates on the Graph’s X-Axis appeared first on Depict Data Studio.