Vizible to the naked eye

Everyone loves to hate PowerPoint and other receptacles of bullet points and charts.

But when you are trying to tell a story with data, smart data visualization can keep the narrative moving. The bar is high, however: data viz has to be as simple as possible, without extraneous bits and pieces, and designed to make conclusions easy to draw.

What we do at Spark can sound like a bunch of gobbledygook: multivariate-this, strategy validation-that, and—oh, behold the ‘test matrix.’ So we spend a fair amount of time making the power of our methodology clear.

Visualization of our work is really important. So important, in fact, that at least a few of us keep a book called Better Data Visualizations within arm’s reach at all times.

We’ve learned a thing or two over the years, so we thought we would share a few visualization strategies that you might find handy in other venues.

Make Mine Multivariable

At Spark, we test different strategies side by side with different audiences and measure relative responses. All those strategies and audiences represent variables and fuel the ‘multivariate’ aspect of our work. But what does that actually mean?

You may use charts like scatterplots, radar charts, and heat maps to show correlations between different things. We do, too. Early in our testing process, we need a way to show the variables and values that are being tested. We use a ‘test matrix’ to lay it all out.

Imagine you have a steady customer base that is starting to waver a bit. You want to try reaching some new audiences, and you have hypotheses about several discrete groups that might be interested in your brand if you repositioned it. Maybe they look like this:

Now imagine that you have three ideas for repositioning your brand to appeal to those new potential customer groups. Spark helps you bring each position to life using ads that a portion of each audience will see. What’s the best visualization to convey all of those variables?

The matrix approach helps people understand a key principle of anything multivariate: every combination of variables gets tested. That snarl of dotted lines, highlighted in white above, conveys the ‘fairness’ of testing: every brand position, expressed as two different ads, gets a chance with every audience. That’s an important part of multivariate storytelling.

Hot Results

The next part of the story: the results of all those pairings of variables. Where are the strong matches between positions and ads on the one hand and audiences on the other?

There are a number of different ways to convey correlations. When you have a lot of variables, bar charts can become awkward. It’s hard to see all the correlations at a glance.

Instead, we use a modified correlation matrix that we refer to as a ‘heat map.’ It helps the eye find the points with the highest and lowest correlations. At Spark, we use the idea of a heat map to understand which position appeals to which audience—and to what degree. Here is the exact same data as a heat map:

A heat map instantly shows that ‘surprise’ positioning may be a great way to attract new audiences without alienating current customers—in fact, in most cases, current customers prefer it to current branding. The heat map also suggests that prioritizing Beauty Brand Fans is probably the right place to start. (In reality, we continue with several waves of testing winners to confirm them and make them more effective.)

Heat maps, with thoughtful color-coding, are a great way to make relationships between different variables come to life visually.

Cause & Effect

What is the result of multiple waves of multivariate testing? Again, we need a way to show correlation. In this case, we are not showing the correlation between variables but the relationship between test activity and revenue.

Because we tested a new product in a specific geographic area, we knew that all sales were attributed to our campaigns. In this case, we can see that the first wave of testing generated very modest revenue:

But as we eliminated unproductive ads and audiences and doubled down on winning strategies, revenue generation improved. By the end of testing, we knew the exact parameters for generating steady revenue. Success!

Looking Vizzy

All the charts above are from a real S9 project, disguised for confidentiality. We rely on a core set of data viz that includes all the charts above to power the narrative that emerges when we test strategies for new product development, brand repositioning, and other growth initiatives.

We are always on the lookout for fellow data viz aficionados…if you are one and you have some ideas for us, we would love to hear from you. Email us at hello@sparkno9.com.