Check out a free preview of the full Building Custom Data Visualizations course:
The "Exploratory Visualizations" Lesson is part of the full, Building Custom Data Visualizations course featured in this preview video. Here's what you'd learn in this lesson:

The process of the creation and utilization of SVG data visualizations at Netflix outlined in this section. Shirley gives sage advice that has been gained through working with the complex problems that have been encountered.

Get Unlimited Access Now

Transcript from the "Exploratory Visualizations" Lesson

[00:00:00]
>> Shirley Wu: Finally, I want to cover a little bit about exploratory visualizations today. Today, we've been talking all about storytelling, suppository visualizations and what it takes the process to build those. And so Susie and Elijah over at Netflix helped me out with this part just because I don't have as much experience with that, with exploratory data visualization.

[00:00:28] But this is what they do day in and day out at Netflix, and they gave me some really great insight. So the very first thing the process that they go through whenever they're trying to build a new exploratory data visualization is they'll have an initial meeting with their stakeholders to figure out the most important questions that they're trying to ask and solve for, and then they'll actually meet with their data engineers.

[00:00:55] So that already is pretty different from our process today. Because unlike the sunny data sets that we use today, they have a dynamic live data set. And so they need to work with data engineers to kind of figure out how can they get the data such that like the data that's most important for answering the questions come back faster from the data base.

[00:01:22] So they'll have a meeting with the data engineers. And after that, they'll mark up their ideas in sketch or they'll sandbox with semiotics which is a library that Elijah has written to make prototyping data visualizations really, really quick. And that kind of comes from their experiences interfacing with stakeholders and having to, in that meeting with the stakeholders kind of prototype and really quickly change between visualizations and see if that helps answer the stakeholders's questions and that also helps them see the shape of the data.

[00:02:02] And finally, they'll go and prototype and iterate with the stakeholders until they get something. I think that the stakeholders then can then take in, use in their explorations and try to answer their questions. And so I have some advice from the two of them. So I asked them what are some things that you might right into quite often across all of the different projects you've done.

[00:02:35] And so one other things they said was, it's a different data source than I think when it's expository and it's often very often SQL queries. And so when you're designing your visualization, you have to take into account and plan for the queries that might take longer. And so it's especially important if you have interactions that are filtering or aggregating, or something.

[00:03:02] If those filterings require additional SQL queries or something and they might take a few seconds to ten seconds or whatever for the results to come back, try and plan around that or plan for that. Either that or try and make a lot of interactions, or try and build a lot of your design around the data that comes back quicker.

[00:03:29] And when they're talking with their stakeholders often time the questions that they might ask them is what is the business question that you're trying to answer? And also oftentimes, they'll talk about metrics that they're comparing and how do those metrics fall into a decision that they're trying to make?

[00:03:51] And they also talked about the level of aggregation that's most effective for that decision-making and they gave the advice of get it to about seven things that are granular, and meaningful enough. If not, the top ten of a default. I think when they say that, it means that a lot of stakeholders will want to just get everything.

[00:04:15] They just wanted to see every single thing and it's our job while we're talking to the stakeholders to kind of get to the bottom of exactly which parts of the data are the most important to answer those questions, and I thought this one was really interesting. A lot of these advice aren't code at all.

[00:04:36] A lot of these advice are soft skills, communication skills. And they said it's really, really important, especially in the organization that doesn't really have data virtualization and doesn't really actually trust data visualization yet and they think of it as this flashy thing like why would we need this flashy thing?

[00:04:54] Gain trust and credibility within the organization. Because oftentimes, if we're coming into an organization trying to bring and communicate the value of data visualization, what we're competing with oftentimes is tables of data like tables of Excel data. And that's really detailed and that's really hard to read, but that's what they're used to and they like the fact that they have access to absolutely everything.

[00:05:18] And we want to kinda convince them like no, actually you don't want access to absolutely everything. Let me show you this way or a this way to kind of just narrow in on the data that will help you answer the question. And finally, I like this one a lot.

[00:05:33] People will get to a state that you never designed for and they will error out. So try and think through as many of the edge cases that you can, but they'll probably always just get to a point that they were like how did you, what? What were you thinking.

[00:05:48] I before I went freelance, I ran into this a lot too. But that also goes back to into building trusting credibility within the organization, try and account for as many edge cases as you can.