Curious what y'all think about this approaching of highlighting the incremental change in each step of a stacked area by using a different shade. I think it's valuable to highlight in this case, but I'm concerned that people will interpret the darker shades as separate data series instead. Any suggestions for alternatives?

@veltman I would try to avoid using two different greens maybe. also the red/green combo might bring in some color blindness problems

@gka yeah, the colors aren't final, though the choice of two similar shades is intentional (group B and C are related and sort of the opposite of A)

@veltman @gka will this be interactive? You could have different views with each series on the zero baseline (for either the whole series or just the differences). As it is, I'm finding myself still having to do the mental math.

@scott @gka ideally as uninteractive as possible, but a fallback is to potentially show the overall without the shading and then the incremental adds as a set of small multiples below

@veltman @gka yes great idea. The shades make sense to me, but I'm used to reading crazy charts. May depend on your audience.

@veltman also better not use green and red, because... you know why 😅

@veltman perhaps the difference stack can have a different width? it calls to mind error bars kind of, which are like a 1-width stack on a bar

Issue 1: yes, that's tricky. Strong labeling will help. How dedicated is your audience?

Issue 2: stacked bars are a better representation of proportion per group, but not easy to compare incremental group changes.

Issue 3: This emphasizes the size of the pending group, but makes all others more challenging.

Issue 4: what if a group shrinks?

Bottom line: what's the most important take-away message / data points?

@veltman Which is to say, I can't have an opinion about structure unless I first understand purpose.

@veltman Interesting, never seen something like this before. How about using a gradient (from lighter to full shade) to show the "extension" of the bar in the new period? It's not stacked bars, but here's something similar I did recently, to give you a visual of my description (forget the arrows).

@francis that's interesting - I do think I'd feel better about it if I were guaranteed to have space to label some of the squares directly (e.g. "50 existing", "+25 more") but depending on the data they could be tiny

@veltman maybe flip it vertically? the coloured bars growing up would be more "natural" to me. if the part-to-whole relationship is not that central, i might think about three "small multiples" of bars starting each from a zero baseline, then the growth is even more obvious.

@drawingdata interestingly, someone else said the opposite about top vs. bottom!

@veltman interesting, i guess i was influenced by my first (visual) impression that the coloured area looked like stacked bars, but the pending area like one coherent area - and therefore i did not initially connect those two parts into individual bars.

@veltman nice idea. how about opacity instead of hue? newer stuff = full opacity, older = fading away. you'd probably have to bin the time brackets and def. annotate / provide legend

@briromer agreed, in this case it actually is opacity-based - the lighter is same color at 75%

@veltman could you use cross hatching instead of darker colors?

or maybe a black stroke between groups?

@1wheel I've tried a few different pattern fills, e.g., but been struggling to find a pairing that feels like it puts the focus on the incremental and also makes the larger group clear... another kooky idea I had was to put little bars or arrows in the corner instead, but that just wound up looking like cuisenaire rods

@veltman looked better in my head!

bar seem like they should work; but there's not a good place to put 'em. the red ones would work better in lower left, but that doesn't work for the other colors.

@veltman for fun a few months ago, I snuck in a weird way to add context to a stacked area chart here (click on a segment to see)

@joe that's neat! surprised you didn't call it out in the caption or something

@joe I've also been toying with this as an alternative to focus in more on the incremental, since this is a case where groups can't shrink or swap, they just get populated from pending

Sign in to participate in the conversation

A social space for anyone in data, visualization, creative coding, and related arts and research. Share your work, discuss, critique, and ask for help! Come one, come all:

→ creative coders
→ data scientists and visualizers
→ generative artists
→ visual researchers, curators, and critics
→ anyone else data- or visualization-adjacent

If you are curious about the world and creativity, you are welcome here. Learn more.