Apr 11, 2019
Braze's Director of Data Jesse Tao and Tech Alliances Manager at Looker Erin Franz graciously break down BI tools and the value of data for the rest of us civilians. They walk through the marketization of data and the power of Looker blocks.
PJ: Hi there. This is PJ Bruno. Welcome back to Braze For Impact, your weekly tech industry discussed digest. And I'm thrilled today to have two very good friends of mine, Jesse Tao, our director of data, the man about data. What's the title Jesse? It's just data person?
Jesse: Well my Slack title is just data stuff.
PJ: Right. So we have Jesse Tao, data stuff-
Jesse: Official title is, Head of Data Strategy.
PJ: Head of Data Stuff, Jesse Tao, and also our very good friend joining us from Looker, that's Erin Franz. Hi Erin.
PJ: Good to have you here.
Erin: Yeah, glad to be here.
PJ: How's the day been so far? You guys have been doing workshops right?
Erin: Yeah. Flew in last night. Just starting the day early, east coast time. Feeling great.
PJ: Awesome. Not too jet lagged yet? You're feeling good?
Erin: So far.
PJ: Hitting our stride. That's what I like to hear. So we're here today to talk about data, about insights. I'm sure as you two know, over the past 30 years there's been monumental strides in what that means to companies, and the value that it can add. So let's start really, really general, where we are today. Erin, can you speak to some patterns that we've since in data, since the beginning of it? I guess from relational, to non relational databases, to the kind of stuff that you work with right now?
Erin: Yeah. I mean, I can speak to ... Since I've joined Looker about four years ago, sort of how the landscape's changed and how we've seen sort of the product evolve with the technologies that have become available. So I think when Looker was founded six years ago or so, Redshift was just emerging as this modern analytical data warehouse. And those technologies didn't really exist before. And what this enabled, was the ability to actually expose large volumes of data across an organization in a way that multiple people could be accessing at the same time, and really using it to make data-driven decisions. Luckily, Looker took a bet on SQL being kind of this language of querying that would scale with all these different technologies that have come out. And luckily, that's been the case. With Redshift, we've also seen other databases like Snowflake and Google has BigQuery, that have really enabled organizations to become data-driven and self-serving when it comes to making decisions based on data.
PJ: And making it more accessible to people like me, like pedestrians, plebs, who just don't really understand kind of the technical side of data. It's like-
PJ: Democratizes it a bit.
Erin: Right. Making it accessible in a way that it's not just accessible to technical folks, to data analysts, to people who understand SQL and know how to code, to people who just want to click and drop and create reports and explore data on their own, products like Looker make that possible.
PJ: And Jesse you work with Looker pretty regularly at this point?
Jesse: Yes, almost every day.
PJ: Almost every day. And I mean we wouldn't call you a pedestrian, you're pretty deep in data, you understand it well enough.
Jesse: Yes I do.
PJ: Why don't you talk to us a little bit about the marketization of data. This is something that's-
Jesse: Yeah, you know, I think today we collect a lot of data. And in my opinion, data has more or less become a commodity now, rather than the hot topic. And what the hot topic of today is, it's insights. Because you're thinking about it, we're collecting a lot of data. We have data coming in from IOT and all these other sources, and most of the data that's collected, isn't being used. So, how useful is something that's sitting in the data of our house, kind of just collecting dust? So, very low value there. The value is from the insights, from actually analyzing the data, getting the data and figuring out what you want to do with it, to drive business decisions. And this is kind of where Braze comes in, and Looker comes in. We're providing the data and also providing the framework and the tools for people who are not using data, to get insights out of it and actually use that data. So I think in terms of the marketing pressure in the industry, we're moving ... We're going to still collect a lot of data, but more of the focus is going to be on how do we actually use that data faster and more efficiently?
PJ: Right. Because if you're not using that data and you're not taking action on it, you're going to be left in the dust, more or less, right? Is that the ...
Jesse: Yes. To put it in Marie Kondo terms, data just sitting there, brings us no joy.
PJ: And you're all about sparking joy.
PJ: All right then. Okay, well what sorts of data, insights are available to day that wasn't available in the past? Obviously this is kind of a big sweeping generalization, but what can we speak to currently?
Erin: I think some common themes that we've seen emerging are, people are collecting data from tools that they are using in their business, whether that's a Salesforce as a CRM's index as a support system and centralizing all that data in one place. So you're not just accessing one data set, not just your transactional data set, but also the data sets that define your whole business and your whole customer journey. So you're actually able to create kind of that 360 view of the customer that we're all sort of striving for, from as many sources as possible. And that's become possible because of these data warehousing technologies that are now available.
PJ: It's all about that 360 degree view these days, isn't it?
PJ: Because I'm still kind of just getting my feet wet with my understanding of the eco system of products right? You have your attribution, you have your CDPs, Braze is in there somewhere-
Jesse: Mm-hmm (affirmative)-
PJ: Engagement. So Looker is the analysis, it's less the visualization and more the business intelligence right? Because I feel like on our call, we talked a little bit, it's not just graphs right?
Erin: Right. Part of it is graphs for sure-
Erin: You need to be able to visualize your data, but much more than that, of course we always say the starting point for Looker is a dashboard, or a visualization. You can really drill into that visualization, see the components that have built that. If you're technical, you can even see the SQL that is being written to the database to supply that result set. And then you can modify that report, you can drill down into the granular level data that's supplying the data for that visualization. So let's say you're looking at event count by day on your application, you can see what those events are just by clicking into one of those data points.
PJ: Gotcha. And that data, that belongs to the company effectively, or that belongs to the user?
Jesse: I have a point of view on that. And before I share my thoughts, I'll preface it by saying I'm not a lawyer, so do not use this as legal advice.
PJ: Okay, thanks for that.
Jesse: I think the data ultimately belongs to the end user, but the company is basically the custodian of that data. Because without the end user, there is no data but without the company, there's not way of collecting or storing that data. So, the company is more or less using, collecting that data on behalf of the user. They're creating some sort of value from it, either from messaging or personalization or just understanding the user a little bit better. Some way of using that data to create insights into the level of value to that user. But ultimately, it is that user's data and the user should own that data. I think that's the point of view that many countries and regulatory bodies are holding as well. If you look at GDPR as well as the upcoming California privacy laws, the focus is really on the end user and their ability to control the data that they collect, the accuracy of it and the right to be forgotten. So, I think there is a common theme where the view point is the end user owns the data whereas, the companies are the ones who are using it to provide value both to the user and to the marketplace.
PJ: That makes sense. And the California protection, that's going to happen at the end of this year right?
Jesse: I don't know the exact timeline. We'll have to refer to our legal team about that.
PJ: Okay. Well we can patch that up later if we need to. So Erin, let's dig more into Looker a little bit. What's the real differentiator for you guys? What do you guys kind of hold up as a torch? This is kind of who we are and what makes us stand out-
PJ: Amongst the other tools.
Erin: I think luckily, the core Looker product has been fundamentally the same since its inception. With the core differentiators being that it's entirely in database. So as we talked about, the ability to access all of the data, down to the granular ... Most granular level that you're collecting it and exposed that across your organization. And the way that we're able to do that while still providing standards governance, so users are not creating their own one-off definitions of revenue, something that's incredibly important to reporting, is through our modeling layer, which is called LookML. So that's where you define all the business logic that your end data consumers will be using, whether by just exposing them to pre-built dashboards, visualizations or having them build their own content. And the way this works, while still leveraging the database, is it's really just an abstraction of SQL, or the language that you're using to create those database investments. And then finally, it's a web-based modern application. So that makes it really easy to share, collaborate and extend into plenty of other users. We have a fully baked API where you can serve data from Looker elsewhere to bring it into the tools where you need it.
PJ: So LookML, you said it's your own language-
PJ: It's built on another language-
Erin: Mm-hmm (affirmative)-
PJ: And so if you know LookML, it actually is useful outside of Looker as well.
Erin: It's proprietary to the product, but it's very ... What you're doing is modeling the components of SQL, which is a core skillset of any data analyst. It really just makes it easier because instead of writing one-off queries, you're writing the components of those queries so they can be reused, by not only the data analyst, but also by all the data consumers.
PJ: Gotcha. Cool. Well let's talk Looker Blocks. This is what I really want to get into because I first heard about it at LTR 2018, because we announced our first Looker Block right?
Jesse: Two Looker Blocks actually.
PJ: Thank you Jesse. Fact checking on the go. Do you want to talk about that? That was kind of a big release right?
Jesse: Yeah, it was a pretty big release because it was still pretty early on in our relationship with Looker but we saw the immediate value pretty early on, so we decided to move quickly in that direction. And I'll let Erin talk a little bit about what our Looker Blocks, but the two Looker Blocks that we released back in November, are based around our currents data export and it focuses on market engagement and user behavior. So marketing engagement on the Braze data side will be things like email sends, push opens and at message clicks, stuff like that. And user behavior includes things like session starts and app purchases, so the behaviors of the users. We take all of that information together to create really useful insights around how campaigns are performing, user retention, if campaigns are improving your driving purchases, things like that.
PJ: Gotcha. And Looker Blocks for those of us who don't actually know the definition-
PJ: Are basically ...
Erin: They're basically templates for LookML. So LookML is a text-base modeling language. So we can model expected data sources upfront. So, data sources that are going to have a common schema, so common tables, columns, fields, within that. We've created a bunch of these for sources that are commonly used across our customers like Salesforce, Zendesk, as I mentioned before, Google AdWords, Facebook ads. The sources we're seeing most often, and then also the sources that we want to model proactively with our partners like Braze.
PJ: Cool. And so these two Looker Blocks, these are the first of many.
PJ: Cool. I mean, do you know what's on deck? Do we know what's coming up or do we want to save that for our next episode?
Jesse: We can save that for the next episode, but I actually want to talk a little bit about why we decided to make these Looker Blocks. And I think it's because we saw in it, the common vision with our product, which is data agility, or what we call, data agility. And that means basically speed to insight for us. As I mentioned before, the value of data is not in the data itself, it's what you can do with it, and how you can actually gets insights out of it. And with Looker Blocks, it acts as a template where we are predefining all the data fields and relationships, and providing those fundamental building blocks for us and out customers to build on top of. So, what would historically take a data engineer or a BI developer weeks, days, potentially even months to model, we do all of that leg work for our customers so they can just drag and drop in those Looker Blocks and be ready to find insights within minutes or hours.
PJ: So that's huge. That's going to save time.
PJ: It's exciting. All right, let's move on down to data tech changing roles. How is data tech ... How is it changing the way people are doing their jobs and what will the change for real expectations be in the future?
Jesse: Sure. Now, I think that people are becoming a lot more data-driven, and thinking about how to both collect and use data in their every day lives. Well not just their every day lives, but every day professional lives. They're using data to not just justify their decisions, but also to understand what the implications are in areas that they may not have seen before. And I think that's going to be a point of differentiation for customers, for our companies, because if you can actually use the data in a very insightful way, you can understand more about your users, your competitors, the marketplace and be able to confidently act in a way that will set you apart. And I think in terms of the data collection, the aspect of privacy is going to be more and more important as well. As I mentioned before, there's GDPR, there's the California privacy laws. I think people are just going to be ... Sorry. I think people are going to have to be more careful about what they collect because in the past, you could collect everything. And now with the privacy breeches you've been seeing at big tech companies, big banks, people have to be careful about both what they're collecting and how they're using it.
PJ: What's the most insightful way you've collected data to make a decision about your life? Putting you on the spot Jesse. I'll start.
PJ: Mine will have to be using Rotten Tomatoes to decide to not watch movies. That's probably it. That's probably saved me several hours of viewing time.
Jesse: Okay. So I actually have a script that I write, that scrapes lottery websites for the winning numbers, as well as the pay out. And I modeled out something where something like Powerball or Mega Millions, the optimal time to buy is a jackpot of around 3.25 to 3.5 million because at that time, there are not so many buyers where you have to split the pot. So you basically maximize your payout that way. So, we have office lotto pools here and I don't really partake in them up until a certain point where I think there's a higher payout.
PJ: I'm going to keep that in mind Jesse. That's a good one. That was a really good one.
Erin: Yeah, saving time, stress. I'm more on the Rotten Tomatoes path.
PJ: Do your homework, do your reviews, leverage the data available.
Erin: I guess restaurants also.
PJ: Yup. Yeah.
Erin: Avid Yelp user.
PJ: I'm a latecomer-
Erin: Not a reviewer but-
PJ: Not a reviewer, right. I'm a voyeur. I hide in the comments and I watch.
PJ: I'm a-
PJ: What's that?
Jesse: A lurker.
PJ: I'm a lurker. I'm a lurker, that's right.
Erin: Yeah, I rely on those people who are letting people know their opinions.
PJ: And I'm a latecomer to Reddit actually. I kind of just joined the bandwagon because I needed information on a certain thing. I was like, wow, this isn't just funny comments, there's a lot of really useful information here. Who knew? Anyways, so what were some trends, some hot ideas in the last few years that really didn't deliver on its promise? What are some current trends or hot ideas you think do have promise in the future? Erin, you want to weigh in?
Erin: Well, getting back to technology here, I think that as companies starting becoming more digital, and they were collecting so much more data and they wanted a place to put it, a data lake, and I think you know, I don't know how long ago it was, but Hadoop technology has emerged as kind of this place where you could be putting all your log data, all of your transactional data, all of this data. And it was easy to collect potentially, but not easy to actually self serve. So you were collecting all this data, but you didn't know ... There was no way to expose it to the organization. So I think that these analytical data warehouses have really filled that void and actually made that possible. And we've only seen that within the past five years or so.
PJ: Can you tell me the different between a data warehouse and a data lake? Because I've heard data lake around this office over the past eight or nine months, and is there a big different that I'm missing?
Erin: I can give the high level and then I think Jesse might want to comment on the more details. But you can think of a data lake as more like a file system. So you're putting all these files of data in this place for storage, but that doesn't make it necessarily accessible to the people who need it.
PJ: But the warehouse, you can actually do more with it?
Erin: Right. In a more performant way.
Jesse: Yeah, I mean the way I would kind of think about it is a little bit more literally if you will. A data warehouse you can image as potentially a physical warehouse that you can just put anything in there. In this case, it's going to be data. And a data lake, you can think of as a warehouse that has a giant pool in it. All that data is kind of just swimming around in a, I wouldn't call it a liquid form, but there's ... It's potentially unstructured, it's very fluid, it's just there.
PJ: Makes sense.
Jesse: And then people can go into that data lake with buckets or whatever tool to extract the data that they need.
PJ: That's a good metaphor. And so data lakes versus data ponds, is there ...
Jesse: There have been some ... I've heard the term data ponds before-
PJ: Really? Okay. I thought I was just messing with you, but I guess I wasn't-
Jesse: No I've heard it before. I don't think we're currently using that though.
PJ: All right, Jesse, hot shot, will data proficiency be a core skill for talent in the future? What do you think?
Jesse: Yeah, I think absolutely. I think here at Braze, and just at other companies, just reading the news, you hear more and more about how companies try to be data-driven. If you just look at our job descriptions, by the way we're hiring, and job descriptions of other companies you see, the requirement of understanding the different data warehouses, technologies, how to use data. A move from Excel to more complicated analytics technologies like Looker for example, becoming more and more popular. So it's absolutely going to be more important in the future. And you know I think for data analysts, that's ... Their role has kind of changed over the recent years and will continue to change as well. I think for the data analysts that I see, it's moving more and more towards a full staff knowledge. So before, you would see people focusing on one element of the data pipeline, whereas analysts today tend to have more visibility over how to bring data in, how to clean it, how to do the app analysis and the visualization, everything. And I think there's going to be more focus on the domain knowledge as well because data and insights out of context, is not going to be terribly useful to the organization. So we need to know how to appropriately analyze and interpret enough information in a way that the business or the end users can actually use. Also, I think in terms of the marketplace, you're just going to see more and more technologies. Some better, some worse than others, within the visualization space. Looker is pretty new, they're a ... I would call them a challenger, again something encompassing the place and they're doing very well. But going further upstream, you're seeing a lot of new database, data warehouse technologies, a lot of new ETL technologies. So I think the data analysts of today and tomorrow, are just going to be more familiar with these technologies and how to use these technologies. And then flipping a little bit to the non technical people, so the end consumers of the data. I think you're going to see changes there as well, especially as data becomes more democratized, and easier to use and consume. We're definitely seeing a trend towards self service. So, drag and drop analysis of data rather than actually going into the data warehouse to write the code and analyze it. We're seeing more sophisticated alerting, so we know when data isn't looking the way we think it should be looking. And that's just going to allow people to move a lot more quickly and more confidently as they try out experiments, they do AB tests and iterate quickly.
PJ: Brave new world.
PJ: Erin, you want to weigh in? What does the future hold for Looker? You don't need to show your full hand. I know you guys have stuff. But anything you want to leave us with?
Erin: Yeah, I think beyond sort of self service, the core BI use case, Looker's really trying to position itself as a data platform. So, not just for internal analytics and reporting, but also serving data elsewhere to other applications, to deliver data where it needs to go, like the action hub integration that we built with Braze. So, basically connecting the dots when it comes to doing analysis and taking action on it. So, building your list of users you want to target a campaign to and not just having to export that and then upload it into a tool, but creating that link directly to that product you're using.
PJ: Awesome. Cool. A lot stuff to look forward to then.
Jesse: Mm-hmm (affirmative)-
PJ: Thank you guys so much for being here with me, and thank you guys for joining us. This has been Jesse Tao, Erin Franz and PJ Bruno. Happy visualizing.