You can read the whole story of a financial report using a Sankey diagram, and by using the same scale across years, you can see visually exactly how the story has changed over time by flipping between the images.
For a more interactive version of these graphs, and to see their other quarters’ results from recent years, visit the NVIDIA Data page.
I plan to post the various new Data diagrams there as they are updated on this site – quarterly results, etc. Consider following me over there. 🙂
As part of that effort, I’ve put together a linktree page as well, with several handy links in one place. The process was surprisingly friendly/usable for a free service.
The results have not changed significantly; in fact, the margin of victory for Mamdani over Cuomo has so far been slowly increasing.
I find it interesting to graph the various editions of the data for comparison. Here is a gallery of the three graphs so far, in sequence. They use the same scale, so the visual differences should be representative of the actual magnitude of the changes.
One other note: the buttons I mentioned below which said “use this one as a starting point” are now updated to say “Edit with SankeyMATIC“, like so:
They do the same thing as before, but I think this design gets the point across better – more like the ubiquitous Edit on CodePen buttons you see on web tutorial examples.
There’s a fun new feature on each of these pages as well:
The top diagram on each page now has a ‘public link‘ that lets you use it as the starting point for a diagram of your own. Just look for this on each page:
Beyond the new make-your-own links, you’ll see some additional changes if you happened to visit those pages before:
The Data pages are now reorganized. Rather than having an extensive directory structure, there is now just one page for all of a company’s Quarterly diagrams and (in the case of Apple) another for its Annual ones. This should be rather less work to maintain, so hopefully it’ll be easier to keep them up to date.
There is now an index of thumbnails at the top of each page, to help give you a broad overview of recent history before you get into the details. (In data visualization terms, it’s a set of small multiples.)
The quarterly results are arranged in rows of 4 so that you can scan up & down to compare the same quarter across different years. This makes it easier to see at a glance, for instance, NVIDIA’s growth over one year’s time:
The scale which is used for every diagram on a page is now listed at the top.
The next question you might ask is:
Why not use the same scale across all of these pages?
The short answer is, the companies are just too different in size. Considering just NVIDIA and Apple:
If you made NVIDIA use Apple’s scale, most of its diagrams would be pretty tiny, making it hard to arrange labels in any readable fashion. (See NVIDIA’s Q1 2024 for an example where the flows are already getting too small to comfortably label, even at the scale I chose.)
If you made Apple adopt NVIDIA’s scale, then Apple’s diagrams would be incredibly tall.
That’s why each page is considered independently in terms of which diagram scale will provide good readable diagrams at a reasonable size.
For fun, here’s what NVIDIA’s biggest-ever quarter would look like if it did use the same scale as Apple’s latest quarter (compare to the NVIDIA image above):
There’s a new Sample Diagram to try out at the top of the Build-a-Diagram page: Journey, which demonstrates how you can produce traceable paths for multiple players across multiple stages by using a specific approach to encoding your data.
It was added in February of this year. The conditions required for this kind of diagram to work are:
Each flow maintains a single color from start to end.
All flows for a particular player must be listed together.
The “Arrange the Diagram” control must be set to “Using the exact input order”.
In the sample diagram, every flow for every player has the same value (1), but that condition isn’t absolutely necessary as long as the other conditions are met.
This diagram style can be used for such things as a ranking chart for team standings in a league (sometimes called a “bump chart”).
This post prompted me to go track down when each of the other sample diagram types was added, for comparison:
