Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Episode #33: Chris Parmer from Plotly
151 Plays9 years ago
In this week’s episode of the PolicyViz Podcast, I speak to Chris Parmer, Chief Product Officer and co-founder of Plotly, the online data visualization and analytics firm. Chris and I talk about Plotly’s tool, their business model, open source data...
The post Episode #33: Chris Parmer from Plotly appeared first on PolicyViz.
Recommended
Transcript
Introduction and Sponsorship
00:00:00
Speaker
This episode of the PolicyViz podcast is brought to you by Juice Analytics. Juice is the company behind Juicebox, a new kind of platform for presenting data. It's a platform designed to deliver easy-to-read, interactive data applications and dashboards. Juicebox turns your valuable analyses into a story for everyday decision makers. For more information on Juicebox or to schedule a demo, visit juiceanalytics.com.
00:00:35
Speaker
Welcome back
Introducing Chris Palmer
00:00:36
Speaker
to the Policy Viz Podcast. I'm your host, John Schwabich. This week on the show, I have Chris Palmer, the chief product officer and co-founder of Plotly to talk about the product that they have and the services they provide. Chris, welcome to the show. Thanks for coming on. Thanks, John. Good to be here.
00:00:51
Speaker
So
Plotly's Origins and Evolution
00:00:52
Speaker
I'm really interested in the sort of work that you guys are doing, but before we dive into Plotly and what it allows people to do with their data and with their data visualization, why ask first if you could just talk a little bit about yourself and why you helped start Plotly and where you guys are today. Sure. So we started working on Plotly about three years ago.
00:01:13
Speaker
And I come from the electrical engineering and math background. And a lot of the people on the Plotly team, especially the core team, come more from like the science and engineering community rather than I'd say like the software design community.
00:01:28
Speaker
And we really started Plotlib to kind of replace and upgrade a lot of the tools that we were using in those fields. So as part of research and as part of industry, we're using tools like Matlab, Matplotlib and the Python community, Origin Pro, which is this old and very powerful scientific charting platform, Excel. And a lot of these tools are desktop-based.
00:01:53
Speaker
And it wasn't until a few years ago with the pace of the web and the performance of web browsers that it was really possible to replace a lot of these data visualization tools with something that could be done entirely in the web browser. And once you start putting these tools in the web browser, you can do all these cool things like you can share interactive charts with each other, which you couldn't really do in a desktop environment because
00:02:22
Speaker
each person that was sharing or receiving the charts would also have to have the same software as you. Sometimes software doesn't even work across Mac or across Windows, but with the browser, it's everywhere and it's on mobile as well. We really started plotly to try to be the new standard for technical, high-performance data visualization on the web.
00:02:47
Speaker
Once we started doing that, we found all these other communities that were doing data visualization. It could really benefit from this community and this ecosystem like journalism, policy, finance, tools, all these communities that are still using things like Excel or data visualization. It could be using something that's a little bit nicer, a little bit more interactive.
Plotly in Action: Real-world Examples
00:03:08
Speaker
So can you talk a little bit maybe about a couple of case studies is a strong word, but a couple of examples of people in these different fields using Plotly and how they've used it. Are they piping their data directly into the browser and then a news organization they're posting right on their website. If it's a research organization, how they're sharing it. So maybe there's like two extreme examples you can sort of share with folks.
00:03:30
Speaker
Yeah, totally. On the super technical side, one example is this research institute in Finland. It's the largest research institute called VTT Finland. Super technical books. And they're using it across their organization and they're mostly Python, Matlab, and R users. And they're using our platform in a way where they do their data analysis and their data acquisition using a programming language.
00:03:57
Speaker
and like Python or R or Matlab. And then they send that data up to a plotly server that runs internally at their company. And those graphs and that data is now published on an internal cloud, and they can now share that across their organization. They're creating super sophisticated visualizations. Some of these visualizations are updating in real time by coming off of instrumentation and things like contour plots, 3D plots, really, really advanced.
00:04:26
Speaker
I should have done
Understanding Plotly's Offerings
00:04:27
Speaker
this to start, but let me step back and ask you to maybe describe the different platforms or different packages people can use for probably the sort of open version, and then there are server versions, as you just mentioned, for this particular place. Can you describe the different levels, I guess?
00:04:40
Speaker
Yeah, so the core of Plotly is a JavaScript crafting library, and that's called Plotly JS. It's built on top of WebGL and D3.js, and it's open source, and we open sourced it this fall. So if you're a JavaScript programmer, you can just use that library, and we have written other
00:04:59
Speaker
libraries in Python, MATLAB, and R that use Plotly.js as sort of the rendering engine. And you can use those languages and those libraries on your own machine, self-hosted, free, all-MIT license.
00:05:14
Speaker
Then there's plot.ly in your browser, so https://plot.ly. And that is a platform for hosting and sharing graphs that are made with the Plotly.js graphing library.
Why Choose Plotly?
00:05:27
Speaker
So we have an online spreadsheet and interactive tools to edit and create these graphs and a whole host of sharing permissions, sort of like Google Docs.
00:05:36
Speaker
That's an online platform. It's like GitHub. It's free for public use. For private use, we have a subscription that you can get on. But then for many companies and organizations, data just can't leave the firewall. They can't use cloud products. So we've bundled up the hosted platform in a way that enterprises can run and install and license on their own infrastructure.
00:06:03
Speaker
So these are the groups who they have proprietary data, they have administrative data, but they still want to be able to create and share interactive visualizations so they can use just in-house basically. Right, yeah, exactly. Okay, and so what have you seen with customers that are using Plautly out in the open, and how do those sorts of organizations differ from ones that might just say, well, I'm just going to create something in D3. A web programmer does do it ourselves. So where do you see that line between those types of groups?
00:06:31
Speaker
Well, I think where we come in is that a lot of people don't have
The Role of Interactivity in Visualization
00:06:35
Speaker
JavaScript expertise, or they don't have a internal team of JavaScript developers and designers that can create these visualizations. So one of my favorite examples is the physics blog on Wired, where Red Allen is using Thoughtly, and he's just using our public Thoughtly cloud, and he can very quickly
00:06:59
Speaker
make simple interactive graphs to illustrate physics concepts and just embed those charts inside his blog, just as he might embed something like the YouTube video. Right. Okay, so my next question is I want to back out a little bit, and this question sort of applies, obviously, to Plowley, but also to other types of tools and platforms and programming languages more generally.
00:07:22
Speaker
Plotly and high charts and D3 and Tableau and everything. So I feel like for a while now people have been using interactivity for sort of the sake of having interactivity. So I build a column chart and I, you know, the user can can click on the different columns and see the number pop up. But that doesn't really give you a lot of additional insight. And so I'm curious now looking ahead.
00:07:43
Speaker
whether you think that behavior is going to change and that the sort of interactivity for interactivity sake is going to decline or plateau or just not be as important. And it's really going to be about interactivity for storytelling, interactivity for analyzing the data, as you mentioned, or interactivity for eliciting more insight as opposed to, here's basically a static chart, but I've added interactivity on top of it. Right. Yeah, it's a great point. And I think that where we'll agree is that
00:08:11
Speaker
a really well-designed chart at the end of the day. You should be able to read it really quickly without having to sort of explore inside the chart itself to gain the insight. The labels should be well-placed so that you don't have to hover over to see values. It shouldn't take the sort of additional cognitive effort that interactivity sort of helps you with.
00:08:34
Speaker
But I think where interactivity is really helpful is sort of like the step right before publishing, and that's like an exploration. If you're looking at a scatter plot and it has 10 points on it, each point represents an experiment or a molecule or a country or something like that, and you have 10 points, and then, sure, you can place a label on each one of those 10 points, and you can read that without any interactivity.
00:08:59
Speaker
But if that's more like 100 points or 1,000 points, or if you're doing a huge experiment in molecular sciences, and that's a million points, then it's really helpful to be able to zoom around into different regions and sort of identify patterns in a really quick, interactive way. So I think it becomes a lot more useful when you go away from small data but into medium-sized data on the order of 5,000 points in larger.
00:09:28
Speaker
And I think that, and also in the presentation layer, there are cases where interactivity can be really helpful if you're, say, plotting a bar chart that has six different series and they're all stacked on top of each other. At first glance, it's really important to see how maybe all of those bar charts, what the cumulative effect of them are as they're all stacked on top of each other, but then it's really nice for you to be able to
00:09:53
Speaker
click through the legend and then just see what the effect of one of those 10 traces represents. So I think for our case, especially we kind of came from this rigorous scientific exploration background, the interactivity was really important for exploration and not as much in presentation, I'd say.
00:10:13
Speaker
Right.
Defaults and Effective Visualization
00:10:14
Speaker
Now that the tool has been around for a little while, are you guys thinking really hard about the annotation layer and making sure that the labels are in particular places, you know, close to the data and that sort of thing so that even for someone who may not truly understand how to, you know, sort of really do interactive design, they can still make these charts and people can still sort of get the basic story out of it.
00:10:38
Speaker
Yeah, absolutely. It's something I spend a lot of time thinking about in that our charts are completely configurable. You can add text and annotations to any part of the chart. You can customize the colors and the layout of every aspect of the visualization. But to make a really nice
00:10:57
Speaker
in the compelling data visualization that's easy to read, it takes some skill and some practice. And a lot of what we do in our interface is making sure that sort of the default settings are in such a way that they doesn't allow people to make charts that don't look great. So I'll go into an example. One of the classic examples is pie charts and visualization. So people want to make pie charts.
00:11:19
Speaker
And pie charts are really terrible visualizations if they're not done correctly because it can be very challenging to compare the different slices of a pie chart if they're not ordered in an ascending or descending way. So the pie charts that we created are by default ordered ascending to descending, so you can very easily compare different slices of the pies.
00:11:41
Speaker
Another challenge with them is labels inside pie charts where if the slices are very thin, then your label shouldn't be inside that slice because it just won't fit, but it shouldn't always necessarily be outside of the slice because that can get crammed too. But the way that we've done automatic labeling allows for labels to be inside and outside of
00:12:01
Speaker
of the pie slices depending on how large the slices are. There's an example where if you didn't have labels, people might make the case for interactivity in that if you're viewing a lot of different slices, the only way to really view the labels of those slices is to hover over each of the values. But in our case, we wrote a labeling algorithm that makes labels look great by default, doesn't necessarily require an interactive layer, and doesn't require the user to be
00:12:29
Speaker
an expert in data visualization or know who Edward Tuftis. Right. So as a creator of a data visualization tool, how do you decide where to draw the line? So you allow users to rate pie charts. Some people in the data visualization community would say, no, don't even let people to make pie charts. But you, at the same token, don't allow people to make 3D exploding donut charts. So as you and your team are developing the tool, how do you decide what to offer and what not to offer?
00:12:59
Speaker
That's a really
Community-driven Development
00:13:00
Speaker
great question. Yeah. And that's interesting to think of, of where that line of really too bad is. Yeah. But you know, I mean, an exploding donut, I mean, that's just like, it's terrible for so many reasons. Right. You know, essentially, it's just all right. So I sort of feel like
00:13:15
Speaker
I feel like the 3D for the sake of 3D, we have almost unanimous consent that that's a poor visualization decision. Pie charts, of course, there's a big debate. But even things like starting from zero on the vertical baseline, that's obviously a very contentious discussion these days.
00:13:32
Speaker
So as you're offering different visualizations, and I'm sure there's others that are sort of more for the scientific community than are for others, how are those lines drawn or is it just an evolution as you create and see what customers want and use and don't use and that sort of thing?
00:13:49
Speaker
Yeah, I think it's more of an evolution and a lot of our development has been driven by the huge user community that has come to Plotly. And, you know, in pie charts is an example of that. In folks, I think without Plotly, we're going to be making those pie charts in Excel and they were going to be terrible. And so I think we can kind of do a service to create what is to be the correct way to draw a pie chart in Plotly.
00:14:13
Speaker
So I think mostly it's been user driven. And I think another interesting part of Plotly is the community aspect to it in that we can provide things like pie charts or stacked bar charts, but then also show our users a lot of different ways that people are communicating really similar data sets using slightly different chart types. So exploding a stacked bar chart into subplots with multiple bar charts.
00:14:39
Speaker
And so I think we also kind of have this duty in our product development, in our community organizing to really showcase our users different ways that they can visualize the same data set and to really showcase the best charts that are created by the community for everyone else. And what
Open Source Transition and Business Implications
00:14:56
Speaker
makes the best visualization? Well, I think that is in some ways sort of a little bit subjective.
00:15:03
Speaker
I want to shift gears for a moment and talk about Plotly going open source, which you mentioned earlier. Plotly opened everything up late last year. I just want to ask you, what was the motivation behind that and how do you view Plotly business model going forward now that the whole tool is open for people?
00:15:22
Speaker
Yeah, so we've always been huge proponents and huge fans of open source, and Plotly itself is built off of mostly open source technologies. And there's this huge trend, especially in the scientific and technical computing world, to
00:15:41
Speaker
work exclusively off of open source technology. So that's been like sort of the general shift from Matlab to Python and SciPy, the shift from S and SAS to R. It's totally the right direction, especially when in sort of science and technical communities where
00:16:00
Speaker
decisions are being made off of the results from these libraries. And you should really be able to look under the hood and feel very confident about the way that your data is being presented or the algorithms that are being used. But for us, you're aware of a privately-invited company, and our core technology was our JavaScript Graph Library. And when we started Plotly, there were a lot of other folks that were offering very similar libraries and licensing them.
00:16:29
Speaker
One of them is like high charts. It wasn't really obvious in the beginning of whether this is software that we licensed and try to build a business off of or whether we go with the trend towards open source and can build a platform around that technology and hosting that provides additional benefits that we can charge and continue and sustain as a company and continue to develop our underlying technology.
00:16:59
Speaker
So it was really a matter of the library becoming stable enough and well tested enough over this last couple of years for us to feel comfortable bringing in a larger community and really seeing the value in a platform around this library and seeing the types of value that we can bring enterprises in a hosted solution that works on premise and feeling very confident about
00:17:27
Speaker
these types of solutions being enough to build a really great company out of rather than just that library. I'm guessing the answer to this is yes. But do you think that that's going to be the trend going forward that more places, be them data visualization platform providers or, you know, statistical package providers, what have you, people are going to demand that these firms open source and that's sort of going to be a driving part of the actual business. Is that actually to have parts of it or the whole thing sort of open court?
00:17:55
Speaker
I really hope so. I think that in the private sector, there's a few different types of companies that are open sourcing technology. There's companies like us that are building technology and we're trying to open source and build a platform around that specific technology. So we build Plotly.js and we're trying to build a platform around that technology. And that's really, that's sort of our core technology and that's our core asset.
00:18:21
Speaker
And I don't think there are many companies that are doing that yet. There's RStudio, there's Mapbox, and beyond that, there aren't that many.
00:18:30
Speaker
Then there are a lot of other companies that are open sourcing their core technology, but their core technology has never been the main value proposition of their company. Some of them are thinking things like Facebook, where Facebook open sources a lot of their core technology that they're using, like React, and that has had incredible benefits. I can see that happening a lot more, where companies are
00:18:52
Speaker
are open sourcing the internal tools that they're using because that really benefits everybody else and they're not necessarily trying to build a company off of that technology. And then there's the third group which builds open source technology and then tries to create a business around the services, around that technology. So that's maybe companies like Docker, historically companies like Red Hat.
00:19:17
Speaker
building a service organization around the open source technologies that they develop. But we don't necessarily want to be a services company like Red Hat. We really want to be a technology and a platform company. And so I think, you know, the answer to your question is going to be sort of like time will tell, like we're going to have to see
00:19:35
Speaker
If companies like ours and our friends at Mapbox and RStudio are able to prove and be really creative about business models that can also develop rich open source ecosystems that everybody can benefit off of. I think there's a way forward and I'm excited the direction that we're going in. It's looking really promising but it's really new territory in the business world.
00:19:57
Speaker
Yeah,
Conclusion and Call to Action
00:19:58
Speaker
absolutely. Well, I'm interested to see what will happen. And congrats on the success you guys have had so far. And thanks for coming on the show. This has been really interesting. Yeah, thanks so much, John. This is a pleasure. And thanks so much for everybody for listening. Hopefully, you've enjoyed today's episode. If you have comments, please hit me up on Twitter or on the website. And please rate the show on iTunes so others can find out about it. And until next time, this has been the policy of this podcast. Thanks so much for listening.
00:20:34
Speaker
This episode of the PolicyViz podcast is brought to you by Juice Analytics. For 10 years, Juice has been helping clients like Aetna, the Virginia Chamber of Commerce, Notre Dame University, and US News and World Report create beautiful, easy to understand visualizations. Be sure to learn more about Juicebox, a new kind of platform for presenting data at juiceanalytics.com. And be sure to check out their book, Data Fluency, now available on Amazon.