# What's a better way to show number of staff at a location on a bar graph?

I'm designing a dashboard that captures temperature at multiple locations (L1-L5). At the same time, I want to show the amount of staff that have been in these locations. Right now I've indicated them in the blue circles. Is there a better way that this amount could be represented in the bar graph below, or is there a better type of graph suited for this?

• What do users do with the data? In what order do they read the graph? Find a location and then the result or find a result/problem and then the corresponding location? Commented Jan 19, 2022 at 10:00
• May I point out that starting at 0 makes no particular sense for degrees Celsius? Do you really never see a heat index below 0 °C? Another nitpick: you repeat "degrees" twice when you say "deg °C". Commented Jan 19, 2022 at 19:58
• are those 2 pieces of information related? Commented Jan 20, 2022 at 22:13
• @JounceCracklePop Of course, the only reasonable starting point would be -273.15 °C :) Commented Jan 21, 2022 at 1:24
• I personally find your current concept substantially clearer (intuitive) than any of the solutions proposed by the current answers. Some of the solutions in the answers aren't bad, but I personally don't find them to be in any way an improvement. Others of the solutions in the answers are completely unclear to me, and would, in my opinion, be a major regression. Commented Jan 21, 2022 at 6:27

If you need to show correlation (or lack of it) between two sets of data like this, a dual-axis chart can do that. Here, the blobs are shown on the right-hand axis

(What I can't do easily is put the coloured bars in the background, which should really correspond to the heat axis rather than be of equal width.)

• I find this confusing in two ways. First, why would you show a correlation between Heat Index and Number of Staff? The use case sounds like a safety thing where no correlation is expected (e.g. how many staff are at risk of heat stroke?), not an analysis (e.g. do our staff prefer to work in hotter or colder areas?). Second, at first glance the blue circles look like they're labeling the height of each bar, not a separate unrelated data point. Commented Jan 19, 2022 at 19:56
• There may be no correlation and the similarity in shape of the graphs may well be coincidental -- but this does show that. Any confusion arising from the placement of the blobs can be dealt with by altering the scale: I just used Excel's default. Commented Jan 19, 2022 at 22:30
• @JounceCracklePop I would have thought they're working in tight quarters and heating up the space. Commented Jan 20, 2022 at 2:38
• @Robbie I think your comment is probably aimed at the question. Commented Jan 30, 2022 at 20:14
• @AndrewLeach So it might be, and what difference might that make? Commented Jan 30, 2022 at 20:24

Maybe simply show it close to but separate from the graph? So both pieces of information are clear to see and understand:

• OK, now I can comment: I would not recommend that. The different parts of the information are quite far apart, there is a probability of the information being overlooked. Commented Jan 21, 2022 at 15:06
• @Gerda I disagree on them being 'far apart,' they are next to each other. Though yes, the probability is valid. It may be solved by using a more prominent color, font, etc. As OP hasn't mentioned how important the staff information is, I treated it as the second and designed so. Commented Jan 22, 2022 at 8:11

I assume that the boundaries of your temperature categories are significant, and related to the purpose of collecting and analysing the data: maybe you have a target to minimise the number of staff working in a 'very high' temperature, for example. In that case the exact temperature measurement may be less important than the numbers in each category. Maybe you could present this as a table:

Alternatively if you want to keep a graphical representation of the actual temperatures, use a bubble chart as user155100 suggests:

The colours, and the bubble size in the case of the chart, draw the eye to the information of most concern. (Note, I've followed the colour scheme from your question, which may not be ideal for colourblind users. Also the horizontal axis labels of the chart should be the location names.)

As was mentioned in a comment, it's questionable whether a bar chart makes sense for showing temperatures because 'zero' is an arbitrary point, unless your values are in Kelvin.

• Location is a categorical variable. Using it as an axis makes it look like it's a quantitative variable. Commented Jan 21, 2022 at 23:08

I actually like your representation, it is clear and easy to perceive and parse. It lacks only a visual reminder of what the number buble represents.
This can be solved by adding an icon to the number bubble to show that you are showing something related to people there. Something like a small stickman - here graphics designers can shine :-)
Of course, the form of the bubble has to change, as the stickman (or whatever) has to go into the bubble as well.

Regarding dicplaying the numbers below the graph: This puts that information very far away from the rest. It will be easily overlooked.

• Or make it a blue person-outline rather than a blue circle. Commented Jan 21, 2022 at 11:55
• Great idea, @Dannie Commented Jan 21, 2022 at 12:20

I would prefer this way of showing temperature & staff together. Though both are different scales, it gives a relative size estimate . High temperature regions are highlighted to grab the viewer's attention.

My only suggestion would be visually convey the staff amount, rather than only as a number. That way you can already process the core message of the data at one glance, before you even interpret all the numbers. Because big = many.

And add a description somewhere, what this icon + number actually mean.

I would suggest that you use a scatter chart. That would allow to have the Temperatures on the left, number of visitors on the bottom, and use colored dots (site names, colored dots, etc., whatever works best for you) as the locations.

• While you can see the correlation between temperature and amount of people, it is hard to spot the result for a particular location. This gets even worse when multiple locations share the same result. So this works best if you want to find problems and then the corresponding locations. But if you want to know how well a particular location did, it is more difficult to read. However, since the OP doesn't state any reading order I consider this to be a good suggestion. Commented Jan 19, 2022 at 9:56

You could try making a scatter plot such as Nathan suggested, with Heat index on the Y axis and Location Labels on the X, then use a Bubble with a number in the center for the count of people in each location. Possibly even scaling the size of the dot based on the Staff counted.

The site Gapminder has a pretty good example:

If it aligns with your goal, you could do the following:

• Show a map or grid or similar
• Display both numbers at each location
• Use visual hints such as color, size, shape, and so on
• Allow the user to choose different sort orders
• Downvoters, please leave a comment explaining why you're downvoting, especially if there's something I can improve. Commented Jan 21, 2022 at 15:33

To minimize the elapsed time it takes a viewer to orient and interpret the graph I would simply encode the staffing level into the color saturation of each bar; deeper the color the more staff.

If it's still important to retain the precise staffing number, do so with a simple text annotation at the bottom of each bar. (Use the full saturation color for this text; the left-most bar should read "19 staff members", while the other bars should just say "16" or "10" and so on. This embeds the legend into the graphic instead of causing the reader's eyes to flick back and forth).

Your graphic is very functional as is - I wouldn't go chasing a complete re-design or switch graph types. Double bar could be good, but is a bit less elegant I guess, and will take a greater elapsed time for the viewer to interpret. And I wouldn't disturb humans' intuitive ordering of the Ls (L1 to L5) by ordering them based on size or anything else (unless L1 -L5 are just example names).