The Problem That Broke 5 Datavis Tools

Or how I tested data visualization tools to find the best one for conveying insights.

In our datavis community, people often write and talk about comparing data visualization tools and how much they can do, but I want to talk about something else.

I needed to do something here one story based on data for students, and it would be good if they could immediately repeat it in some simple tool on their laptop.

I started to doubt which instrument would be best to show it in. And I started to go through the options.

It would seem that the task is super simple! But no. Every good tool has limitations, which, as it turns out, simply do not allow it to be solved!

I will tell you about all my ordeals in order, and at the same time I will show you how to build a diagram in these tools (this case will not work in educational projects anyway, I value the nerve cells of my students).

So, given

A test dataset (generated by ChatGPT) on employee performance and how that performance declines over time with the company.

We have information about the department, efficiency, duration of work, number of days off, and working hours of several employees.

To see efficiency problems we need to show the relationship between two numerical variables. Easy! Scatterplot (aka scatter plot) will solve this problem for us in no time!

But! We need to not only construct a diagram, but also design it in such a way that the conclusion is clear. This is where we will have problems…

So!

We order sleeves and jump into the first tool we come across, what could go wrong?

1. Testing DataWrapper

I really like this one program. Works online, registration is simple, free, gives you an interactive chart, which you can give a direct link to and show to someone, or download as a picture.

Besides, DataWrapper itself looks after the style, color and fonts, and everything is already well-designed by default. You don't even have to bother much with headings and annotations – fill in the fields on the side, the program will do everything beautifully. But, as it turned out, this guy doesn't know how to fine-tune…

But let's go in order.

1.1 Loading data

Everything is very simple and convenient here: you can paste by copying, you can download any popular format or connect to an online table. Candy!.. Unfortunately, the program cannot aggregate, that is, you need to give it ready-made data points, it will not add anything itself (well, as far as I know, if someone can – tell me this tricky sorcery).

1.2. Check and describe

Here we specify that all variables are of the required type and can transpose the data if desired. But the most useful thing at this stage is that we can rename the column names if they have technical names or, as in my example, in English.

1.3. Visualization

And here is the most delicious stage! We finally try to assemble our scatter diagram. We select it in the list and!.. The program plays tricks and shows us an arial chart. Well, that's not scary, we go to the point Refine — this is where we will set up the schedule.

Here we have to choose which variable will be responsible for which axis, and we will already get something quite similar to a scatterplot (scatterplot).

Now let's add the department variable – in color. Later I abandoned this because it interfered with the perception of information. I wanted to add data by department – because the departments have different situations, and this is part of the conclusions on this data. But for this I need to get 4 such graphs, and DataWrapper cannot give me this at once.

But uploading data four times, processing it separately, and then collecting it in another product… is not an option.

Here is our colorful preparation ready – and the most valuable thing is, of course, trend line! It is precisely this that makes it possible to understand that something is wrong with efficiency. Before, we could only guess about the problem. Now we can clearly see it.

1.4. Adding texts

And now we are able to convey some conclusions with the help of this graph. As you can see, in the section Annotate You can write a title, some conclusions, notes, provide a link to the data – which is very cool for a publicly available diagram (but I won't do this for the example).

1.5. Result

What didn't work:

• No split into multiple charts

• You can't highlight one data point beautifully (you can do it ugly)

2. Testing Excel

I consider Excel to be the most democratic datavisual tool with a low entry threshold. Firstly, many people have it installed, and many work in it. And surprisingly, the chart settings in it are also quite good.

…Spoiler: up to certain limits.

DataWrapper is certainly not bad, but I was still haunted by the idea of ​​breaking the data into 4 graphs. So it's time to dive into the settings of the MS product!

2.1 Making a schedule

Since I already had the data in Excel, I didn't need to import it anywhere – how convenient! I'm increasingly hoping for this option! And so, having simplified the table, choosing in the section Insert the type of chart I need, I get a graph.

2.2. Design of the schedule

I need to somehow show that the efficiency is falling on the chart. But it is not very clear from this diagram whether anything is falling anywhere at all. I google how to add a trend line, and easily add it via a plus sign in the chart design.

Overall, it looks nice. But remembering the desire to get several graphs, I start digging around to figure out how to make it simpler here.

2.3. Graph on pivot tables

Having realized that I need to somehow interact with the data to see these 4 graphs (of course, you can make 4 separate tables, but the automation is so-so), I think to use pivot tables and graphs based on them. Not the most convenient tool, but still can do a lot.

Well. No luck, having made a pivot table and tried to insert a graph or through Insert or through PivotChart — I come across Excel refusing to work with me.

After banging my head against the wall for a while, trying to get four graphs through filtering a regular table or somehow bypassing the problems, I give up on this… Sadly. Maybe someone knows how to set it up better here? Or is this really a limitation of the system?

Well, I couldn't divide it into sections in Excel, although the tool is otherwise quite customizable, you can even format a graph and annotations and make color highlighting (even DataWrapper isn't very good at this). But alas, there are some problems with the scatterplot specifically.

What didn't work:

3. Testing Power BI

I also love this Business Intelligence tool, you can make great ones in it dashboards and it has a pretty user-friendly interface, it's free up to a point, and it's great at manipulating data! Mmm!

So, of course, I decided to try to do something on it. However, I must say right away that I didn’t really hope for success, because although the example might turn out to be good — Power BI has a pretty decent chart setup — my students definitely won’t have it on their laptops. And installing such heavy, serious software is not an easy task, and it doesn’t work on Mac — sad, right?

But I'm already excited, so let's test it too!

3.1. Making graphs

Power BI interface is not much different from Excel, it is quite easy to fill it with clean data (button Excel in the main menu to help you), and then we get a list of variables on the right and with the help of the block Visualizations we work with graphs.

Power BI also has a Russian version, by the way.

So, we choose a scatterplot, we choose the necessary variables, X, Y and departments in Legend! Beauty! I can also add, for example, time off in the circle size, but I never found any use for it. So in the final version it is worth removing all this.

Beauty! Everything is there, I'm almost opening the champagne.

Because it is quite easy to split this visualization into 4 charts in Power BI – copy the chart 4 times and specify the desired department in the filter for each one. And you can place them on the sheet as you like, so that you can then create a nice report.

Look at the beauty that comes out. The colors are no longer a jumble, and we finally see the 2nd insight – not all departments suffer equally! The Innovation department is not subject to the destructive influence of time.

3.2. Analytics

There is one small detail left. I certainly see that in three of these four graphs there is some trend towards a decline in efficiency, but it would be great to highlight it.
And here we hit the wall, because it is for scatterplot in Power BI no trend line! There are various other wonderful lines – constants, averages and others, but not what I need. In Power BI, there is a trend, apparently, only for time series. Alas, we cried, gathered ourselves, let's move on!

3.3. Adding texts

And yet, the tool for presenting the findings is also good. It has slightly less control over graphics than Excel, but it has much more filtering capabilities, and it is easier to manipulate the data.

So I still formatted this example into an almost finished version, adding several textual layouts.

If there were normal trends, it might look something like this.

In conclusion, I will note the disadvantages:

• Lack of analytics (no Trend Line tool for scatterplot)

• No customization of individual data points

4. Testing Tableau

As soon as there are problems with Power BI, where do we go? That's right, to another BI tool, to its faithful competitor Tableau! I have long forgotten about students and the low entry threshold, now it's a matter of honor!

In fact, the entry threshold for these packages is not that high, but it still takes time to install the software, so for educational purposes, Excel is of course good – which everyone has, Google Sheets and online tools.

But I know that Tableau can definitely do trend lines, so I boldly load the data set there! Let's go!

4.1 Loading data

Loading data into Tableau from Excel is also quite simple, if you have a flat narrow table without any formatting issues, then everything will be ready right away. Be careful with dates – they can be in American format (M/D/Y) by default, even if you had everything set up well in Excel.

At this point we can also do something simple with the data – rename columns, make a small pivot, check the format. So far everything is decent, and I don't have any dates here.

4.2. Visualization

It's pretty easy to make a scatterplot – just take the variables from the list on the left and place them in the rows. Columns And Rows — and your variables will become the Y and X axes respectively. In our case, these are Experience and Efficiency.

By default, your data will be aggregated (an icon will appear SUM in front of them), this does not bother us in any way, because we have unique data points – it does not matter whether we average them or add them up.

The next step is to add a level of detail – then instead of 1 big dot we will get a dot for each employee – for this I throw the employee's last name in the section Mark — Details.

Now we see something that looks very much like a scatter plot.

4.3. Analytics

Great! A couple more tricks – I'll replace the rings with circles in the section Marks having changed Automatic on CirclesI will add the number of days off to the size of the circle (Size in the section Marks) and quickly to the section Analyticsto get the coveted trend line!

Excellent. She is here.

Another cool thing about Tableau is that when you hover over a trend line, you can see the value. P-valuewhich is important to us if we want to understand how reliable the conclusions obtained are.

4.4. Group of graphs

Now I will finally do what I have wanted for so long – I will show trends on different graphs for each department.

And it seemed like I got what I needed. But I didn't really like the horizontally glued graphs. I added color and transparency to the circles, but I still wasn't happy with the result.

The thought of endlessly adjusting grid lines and indents between diagrams made my jaw ache. And then there was the fuss with utility fields to put the graphs in a 2×2 grid.

I love Tableau, but the line formatting is just awful.

So, having realized that there is an opportunity to work here with a group of graphs, but I don’t really like it, I decided to make a picture with texts and annotations for all the data without dividing it into sections.

A nice feature of Tableau is that you can add annotation labels around your data points – you can see this in the example below.

But I didn't try to design this poster anymore, because I wanted more flexible management of individual department schedules. Tableau gives the opportunity to implement my wishes, but I didn't want to spend the time that is necessary for this.

I'll write down the cons:

5.1. Testing RAWGraphs + Figma

Well, this experiment has long gone beyond the scope of a test assignment for students. And most of the tools above solved the problem in one form or another. I lacked flexibility, so I decided – I want flexibility – I need to go to a lower level of detail – to vector graphics!

Beautiful and incredible RAWGraphs — makes it possible to assemble a complex diagram using a template, download it in SVG format, and then process these primitives in any vector editor! Magical, right? I use this combination to create data art projects, because behind all the beauty in such projects there is still data, and a clear layout is quite an important condition there.

RAWGraphs, like DataWrapper, is a free, browser-based tool, so they work on a similar principle.

5.1 Loading data

After the start page, we get to a long sheet, where from top to bottom we fill in the necessary parameters – at first we can simply copy and paste our data into the window, or fill it in one of the popular formats! But not in all.

You can load tabular (TSV, CSV, DSV) or JSON data.

So I had to convert my excel to CSVfortunately, this is done in a few seconds. Then we check whether the formats are correctly perceived by the program and scroll down.

5.2. Selecting a visualization

Next we are greeted by a rather extensive list of complex diagrams that we can try to build on our data. “Try” — because some diagrams require a very complex data format and a certain number of variables of a certain type.

But I have no time to be distracted by tasty Bump charts and cute chord diagrams, I choose Scatterplot! Here it hides under the name Bubble chart (bubble diagram) – this is the name of a complicated version of Scatterplot, when the size of the circle also codes some variable. Not all software packages separate these two types of diagrams.

5.3 Assigning variables

The most difficult part of working with this program is that if RAWGraphs doesn't like the format of your variable even a little bit, it will refuse to work with it. Leave.

But my situation is quite simple, so I can easily put my 3 main and 1 additional (time off) variables in the corresponding boxes. And notice – there is a section right here Serieswhich will help me separate the graphs into different departments!

I add signatures with employee names (to check the data later, if anything). Here you can also add the display of all the necessary data labels. But I don't need this.

5.4 Visualization

Let's move on – here it is, hour X, scrolling a little lower we immediately see the result!

As terrible as mortal sin!

Somewhere off-screen, the glamorous DataWrapper faints.

But the graphs are nicely arranged – not too close together, but at convenient distances.

In this window on the left we can improve everything a little, adjust the colors, axes, bring everything into a divine form a little… But we will not do this. Because instead we will go to the very bottom of the screen and with one button export this “fish” to SVG format, so that you can then finish it in a great vector editor (free and online) Figma!

5.5. Finishing in Figma

Here we can change the colors of objects, add complex design elements and work with vector almost like in Adobe Illustrator.

We can place the file there simply with the mouse — Drag&Drop. We get this beauty.

Now we will ungroup all of this many times, remove the background, change the fonts to decent ones (Inter for example), make the axes not so black and rename their names. Remove the names that I added here completely in vain.

We put a white background and insert the texts we are already familiar with.

It looks pretty nice, despite the rather bad source, and it's all done pretty quickly.

But here we clutch our hearts! Trend lines! They are not in RAWGraphs!

All in vain. Despite the wonderful fine-tuning, complete freedom to work with forms and prototype, we fall right at the finish line.

Alas, alas…

Of course, I have already come to terms with the fact that no tool can fully cope with this task (and even if it does, it will cost me a lot of time and nerves – like Excel and Tableau).

So I went to extremes – I took the charts from Tableau, made them a background for these charts and copied the trend lines from there, adjusting the scale of the diagrams. It was a bit humiliating, but I managed it.

Because I would like to receive at least something ready after all these ordeals!

Here's a look at the final result, the product of three data visualization tools.

By the way, note that we don't really need the sizes of the circles here, and I don't tell you anywhere what they mean, so you either need to add a legend or make all the circles the same. Considering that this requires going back to RAWGraphs or drawing the legend by hand, I certainly can't do that for this experiment.

I also ended up removing the color of the departments, realizing that I needed color here to highlight important points – the abnormal department and problem areas.

So all the circles became gray. This is a very cool technique, by the way, I recommend it to everyone. First, desaturate the diagram to shades of gray, and then add one accent color.

6. The final battle of instruments

Well, let's sum it up.

All of these tools are great and I love working with them, but unfortunately, none of them could solve this problem easily and quickly.

Still, I will note the pros and cons of each tool as I see them:

1. DataWrapper — beautiful and convenient, for small stories on charts, but you will not be able to customize your charts beyond the necessary minimum, which the system will do for you. But it will do damn well! It really will not help you with data. Free, browser-based, can be interactive.

2. Excel — everyone knows him, everyone loves and hates him, he has a wide functionality for working with data and visualizations. True, you can face limitations here and there. But there you can customize the color and style of each individual figure! Almost like in Figma. So let's not throw this guy off the ship, just take into account that he had some bad luck with Scatter later. The downside is that Excel costs money, but almost everyone has it at work.

3. Power BI — is generally good in everything and is the envy of everyone, even more convenient with visualizations than Grandpa Excel, even more convenient with data and Siltfraitz! But where did it lose the trend line, I still can’t figure it out… I also consider this a minor feature, without throwing this wonderful tool out of my arsenal. Although fine-tuning the elements here is difficult. There may be problems with selecting individual columns and data points. Solvable, but not always. There is a free version!

4. Tableau — a very good product with fine-tuning of everything in the world. Limited in working with data, there is a free version, which recently learned to save files to the computer (which is great!), but for some reason it can’t save a picture. And I wrote about the disadvantages of fine-tuning above — when you need to compose graphs with each other — it is a bit cumbersome. Although I still believe that it can solve my problem. Just.. Not today… and not tomorrow…

5. RAWGraphs free, online and very simple. High requirements for the data format, gives you a prototype of a complex diagram, and then do with it what you want. Let me down by the lack of analytical tools. Alas. But then what beauty can be with this prototype in Figma do! All data artists and data journalists love and appreciate this product.

In fact, all the suffering lasted literally one evening, and I wrote the article the evening of the next day, because I was full of emotions!..

I hope this text will be useful for those who work with data visualization.

I looked at it here not all toolsbut only those that I use often. There are several dozen datavis tools in total!

For example, many others programming languages can be good in fast charts. But believe me, there are enough of their own problems and there are serious problems with storytelling and formatting. Who wants to try to solve this problem in their tool – I am attaching a link to the data source! Throw your versions in the comments, and write what problems you encountered.

And if you teach me how to solve the problems in Excel or Power BI – which prevented me from finishing their examples – I would be especially grateful!

Thank you all for your attention, if you are as touched by the topic of data visualization and the need to present some conclusions with their help as I am – I run a channel on this topic – Chatting ChartsI'll be glad to see you.

Thank you for reading to the end!