Small Bites, Big Results: How Analyzing Info in Chunks Can Change Everything.

Elements that work in this chart

✔ Using a stacked column chart

Choosing the stacked column offers more analysis possibilities. We can analyze the share of each category (part-time and full-time workers) and the percentage of all working people (sum of both categories).

✔ Categories’ order

In the stacked chart, the order of categories matters. We can accurately compare only the one closest to the zero line (and the sum of all categories). Putting people working part-time first stresses this group and makes full-time workers a complement category. In our case, focusing on the part-time workers is more insightful.

✔ Providing different perspectives

Including multiple dimensions such as gender, education level, and having children invites a more profound analysis. We can look beyond the single-dimensional split and observe patterns connected to each analyzed dimension.

Elements that don’t work in this chart

✗ Chart orientation

Using columns leaves a little space for the category name. There is no problem with top-level categories, like gender and education level. But the ones closest to the axis (children / no children) are strangely tilted and hard to read.

✗ Chart layout

Theoretically, placing all columns in the same line is a good idea because we can easily compare all columns. But in practice, the original layout contradicts how we analyze the data. We like to compare similar things with each other. Take the category of women without children who obtained the lowest education level as an example. The most natural would be comparing them with:

1. Women with no children on a different level of educational level
2. Men with no children on the same educational level
3. Women with children on the same educational level
4. Men with children on the same educational level

To allow that comparison, we need to regroup the data so it can be easily scanned.

Step-by-step improvements

✎ Switch to the bar chart

Rotating the chart 90 degrees solves the problem of unfitted labels. With this change, we can separate two genders and reduce the number of text elements by half.

✎ Introduce small multiples

The original chart’s layout separates totals from the breakdown by the group, which distances aggregated value from the categories and impedes the comparison. Additionally, all categories are mixed up and hard to distinguish, looking for patterns. The solution is to reorganize them and group them into four charts based on two dimensions — gender and possessing children.

✎ Work on colors and legend

The original chart uses two simplified sequential palettes — one for women and one for men. This is a good choice because there is a relation between full-time and part-time working patterns (we can order them according to the number of working hours). But the order of the colors looks odd. Putting the part-time pattern first suggests it’s more important; therefore, encoding it with a more prominent color is more natural. We can also simplify the legend by removing the squares and coloring the category name.

✎ Improve formatting

Start with repositioning the scale and putting it between groups with and without children. This further separates the two groups and sets the scale closer to both charts. Next, we can replace the total bars with reference lines and make comparison easier. Representing totals as a line helps compare all educational levels with it and totals across small multiples. Lastly, the original title works more like a subtitle, which is where we should put it.