How to Read a Sankey Visualization

n dimensions + 1 metric

Shows the distribution of a metric on several dimensions, with the value of each dimension’s distribution shown on the next dimension.

Example : It’s possible to distribute the “visits” metric on a “types of visitors” dimension where the values could be “new” and “known”, then distribute the “new” and “known” on a dimension “sex” between male and female.

For Example: Let's say you want to track the entire journey of your visitors. Which channels do they come from? Which pages do they land and leave on?

You can clearly see that the majority of your visitors are new.

Sankey New visitors

Most of your new visitors come directly to your website via url, but a significant portion also come from referrals.

sankey referral

Your known visitors mostly come from organic search.

Sankey organic search

All of your organic search visitors accessed your website from, and most of your referred viewers come from, or

sankey google

sankey designmodo

What about landing pages?

Direct and organic search visitors land on /the-company, /the-product or /home pages.

sankey companyUh oh...your referrals are landing on your /404 page.

sankey 404 You must have removed a page from your website which was referred to by an external link. Something must be done about your 404 page and your /511 page because people land on it and leave without ever finding your home page.

Is this the journey you had in mind for your visitors?

In order to play around some more with Sankey Visualizations you can head over to the visualization lair at

How to read Leaderboards


The leaderboard displays a metric over a dimension in a way that allows you to easily categorize each dimension value’s “performance” in terms of the metric value.

Leaderboard Visualization Captain dash

In the example above, we can see the amount of revenue distributed over different dimension values (Points of sales). The dimension values are displayed in descending order, from the highest metric value to the lowest which allows us to “stack them” against each other. Travelling down the visualization, we can see that Shanghai is the point of sale selling more, followed by Berlin, then New York, etc.

The general statistics displayed allow us to determine the significance of each dimension value by comparing it to the Total and Average revenue per point of sale.

In order to play around some more with leaderboards you can head over to the visualization lair at

Two dimensional Leaderboards

These three different leaderboards essentially offer a different view depending upon how you wish to present your data.

Grouped Leaderboard

Grouped Leaderboard Captain Dash



The Grouped Leaderboard further segments the revenue by town into “Point of Sale A” and “Point of Sale B”, which allows for a more specific breakdown of information. The two values are grouped together vertically with the superior value appearing first.

Stacked Leaderboard

Stacked Leaderboard Visualization Captain Dash


The Stacked Leaderboard combines the two values together, which grants a view of each metric value as a “part of the whole.”

Superimposed Leaderboard

Superimposed Leaderboard Visualization Captain Dash


The Superimposed leaderboard displays the total amount of sales on a bar, with a specific segment highlighted for identification. In this case, we can visually pick out the “online sales” from the bar of total sales.

In order to play around some more with two dimensional leaderboards you can head over to the visualization lair at

How to Read Stacked and Bar & Line Graphs


These add a dimension to the bar chart, represented by a group of bars or one « cut » bar on the dimension of the horizontal axis.

Stacked mode allows you to effectively observe the evolution of the sum of elements of a dimension, over the dimension of the x-axis (example: time), and gives you an idea of the distribution of elements within that dimension.

Grouped mode compares the parallel evolutions of each element of the dimension, which is shown by a group of bars.

grouped:stacked bars Captain Dash

Let's say you want to split your visits over time by type of referral. You want to see on a day-to-day basis if your visits are direct from search engines or from other websites to which your website is linked.

At first glance it’s clear that visits from search engines vary little over time, as opposed to direct visits or referrals, which vary quite a bit more and display an almost parallel evolution.

If we switch to stacked mode, we can see that the proportion between direct visits and referrals is regular over time.

What does this mean? That your visitors already know how to access your website, whether it be by typing the URL or because they know where to find a link?

In order to play around some more with grouped/stacked bars you can head over to the visualization lair at



The bar & line visualization grants the comparison of many different elements of the same dimension on two common metrics and is effective in exhibiting a time evolution. This method allows you to see the correlation between two metrics.

bar & line captain dash

Let's say you want to see if an ad campaign is effectively translating to greater web traffic and a higher number of visitors viewing your content. Adding a line representing your website bounce rate shows you whether your paid visits are efficient or if your visitors just leave your website as soon as they arrive.

At a glance, we can see that even if your paid visits vary dramatically, your bounce rate remains relatively consistent over time. You can still see, however, that the slight curve in the bounce rate correlates to the evolution of visits over time. The higher the amount of paid visits, the higher the bounce rate (and vice versa).

So is your campaign worth it? Oh, and keep an eye on your bounce seems to be increasing over time.

In order to play around some more with bars & line graphs you can head over to the visualization lair at

How to Read Bar Graphs and Cumulated Bar Graphs


The bars allow the comparison of a large number of different elements within the same dimension, sharing a common metric. Effective for showing a time evolution, that is to say where the dimension is time and each value of the dimension is a date.

Let's say you want to visualize your website's visits over time. Here, the dimension is time:monthly so each bar is a month and the metric is visits so each bar height is relative to the number of visits for this month. 2015-03-31 13-05-45

At a glance we can see that there’s seasonality in our database, with almost regular sinusoidal variation. At the same time, we can see that 2013 outperformed 2012. As we take a closer look, we see that our best months in both 2012 and 2013 are February, March and April. Moreover, we can quickly spot that 2013 outperformed with a gap consistent with other months. We can also note that September was consistently positive with significant audience peaks, as compared to appalling performance in August of both years.

So, the question that we need to answer at this point is - what does this mean for our business?

In order to play around some more with bar graphs you can head over to the Visualization Lair at .


This visualization allows us to follow a progressive increase as the value of each bar is cumulated. In other words, each bar represents the current value added to the value of previous bars. This means that the added value each month can be easily seen in the difference between the value of the current and past bars. Progression of bars that appear constant signifies little increase in value, while steep inclines indicate that the value of the metric has increased dramatically from the previous period.

Let’s say that the dimension is time:monthly and the metric is the amount of Facebook fans. 2015-03-31 13-06-07In that case, the bar length represents the total size of the Facebook community and the different between each month represents the amount of new fans. We can then identify the months in which community size increased significantly. The community size increased more significantly between June and July 2013 in comparison to July and August.

Here the questions we can answer are - Was your Facebook post content different during the former period? Were you more active online?

In order to play around some more with cumulated bar graphs you can head over to the Visualization Lair at .