 Good morning. It's 11 a.m. Eastern Standard Time and it's time for the first functional group update for Mel Tano formerly known as BIS OPS My name is Jacob Schatz and I'm on the Mel Tano team. I'm going to share my screen and To start off I wanted to Okay, give a thumbs up if you can see my screen Great. Thanks. So to start off I just wanted to introduce Mel Tano Formerly known as BIS OPS because it's possible that you don't know what Mel Tano is So I would highly suggest you go over to the read me It explains a lot. We spent a lot of time making it really nice and readable But I'm going to give you a quick TLDR. So Mel Tano is a framework Which is for analytics business intelligence and data science and I want to give you a quick example of The whole flow of Mel Tano So here's Chloe. She's a talent operation specialist and She works on recruiting at kit lab and she needs to know how much time it takes us to go from resume to hire and so Luckily there is for lever which is the SAS product that we use for Keeping track of hires in that whole process through hiring lever has an API and So we write some code that's going to basically vacuum up all of the lever data And it's going to take all that data and it's going to store it Into a database on the Google Google cloud platform Which is using cloud sequel with Postgres and that data gets sent into our data warehouse, which is in the Google cloud platform Sorry about the vacuum in here And from there we run dbt jobs And basically all that does is that segments the data aggregates it and fiddles with it and the whole purpose of that Is so that once we have that data we can put that data into a visualization called looker and So once we have that data into looker then we can finally answer Chloe's question of how long it takes to get from an actual resume to hiring someone and So what are the Mel Tano goals the Mel Tano goals? There's a few of them is to meet the data team's needs. So Taylor is on the data team and Mikael Me Yanis and Josh are on the bizops team and so Taylor says hey, we want to know the number of employees from bamboo HR and So he says can you create a bamboo HR ELT so that we can extract load and transform that data from bamboo HR? And so Mikael says sure hold on. I'll be right back. Let me create that So we want to apply software engineering principles to the data science world So in the data science world I'm gonna walk over here. Well, I keep talking to quiet down the sound of the vacuum So in the data science world, they might pass around data Using Excel spreadsheets and you know that the Excel spreadsheets might go. Oh my gosh in a couple different versions You know, you might be passing around different versions and so you what you really need is an end-to-end integrated solution Which is what we use in GitLab. We have an end-to-end integrated solution So we can apply these software engineering principles of version control continuous integration and open source To the data science world to make the process much easier. So first of all, we need to solve our own problems In Montana, we're gonna use convention over configuration and It's going to allow GitLab to make decisions from all parts of the company Marketing sales product and engineering and we're actually going to be able to make these decisions in GitLab So what are we currently working on our milestones follow the GitLab milestone timeline, so We're currently on milestone zero point three point zero and there's a link to the boards where you can see Two boards you can see the bizops team boards and you can see the data boards So when we say that we're following the data team needs the data team has all their needs listed out on boards and with the label bizops those are Things that the bizops team is going to handle so from a Standpoint of what have we done so far? We've created Extractor loaders and transformers ELT's from our Keto lever Zendesk bamboo HR Is a work in progress GitLab's a work in progress net suite is a work in progress and sales force is a work in progress And we have more that are in the pipeline if you go check out the boards The GitLab and the net suite one are like 95% done. So there's already a lot of data that we can pull from this The team is Josh me Yanis and Mikael and Those are the things that we're currently working on and I wanted to make sure that we had plenty of time to open it up for Any questions, so let me Go check out the questions And from a legal standpoint Jamie We have to do a process called pseudonymization which Is a hard word to pronounce but that means that we're taking the data that we can't Show and we're making sure that it's anonymized. So for example with the GitLab Thing we're sucking down all the data from GitLab But when there's a user ID we anonymize that so that we can't directly associate it with the user and so If you go check out in the bizops In the meltano repo you can see a merge request for the GitLab ELT and you can see there's a YAML file That shows all the fields that are going to be anonymized and so that we can configure that over time So ELT stands for extract load and transform. That's where we take The data from one of these APIs and we pull it in we extract it We load it into our own database our data warehouse and then we transform it using a Tool like dbt and then it goes into looker and you're able to see Visualizations of the actual Transformations and you can answer questions Right ELT versus ETL I'll let the data science people argue over that Does meltano have a data in a standardized format like RDF When you say Data in a standardized format there might be other people here that are better to answer that question. What do you mean Lucas? Yeah, I mean the data you will be put just put it in in looker Or do we have like data dumps that we could analyze or play with what do we do with the data here? Right, so the data comes from let's say the lever API and we're pulling in all that data Anonymizing it where it needs to be anonymized and once it's in the data warehouse It's in a database then technically you can query it And you can kind of answer your questions from there but once what we do is we take and Taylor is much better for answering this than I am but we take dbt which is a tool that will transform the data and Make it so that it's suitable to go into looker. Tell me if I'm saying this right Taylor because this is a yeah Now you're doing great. Keep going. Thanks so it's the dbt dbt is going to Run over that data and it's going to fiddle with it so that we can then visualize it in a tool like looker But it could be another tool But right now we're using looker. I Was just asking because the company I work before they worked in that space. So yeah Right, does that answer your question at all? More or less Taylor do you have a better answer? Well, I will say Another pitch from the looker business user training on Monday the 21st But yeah dbt just we can transform the data using just a bunch of nested and connected sequel queries and then looker sits on top of that or any really data visualization layer and Can connect to the different tables and visualize it in myriad ways and Jamie is asking how she can get in on this so legal data can be tracked talk to the data team talk to Taylor and right Taylor and Talk to Taylor and I think you can go from there Yeah, and you can I'll put in the chat here. You can put Issues for visualizations and data you'd like to see and play with in there and we can prioritize from there And I'll be giving a FG you at some point here in the next few weeks. I hope so right Clement is asking how far are we from a 1.0? Do we know what that would entail so we don't know what that would entail Right now right now our only goals is to solve the the data team needs and by doing that We're asking lots of questions and we're solving lots of problems but we're putting off You know a lot of the Creating you know things from scratch and we're using a lot of existing solutions right now To do this stuff where it makes sense But for the ELT's or ETL's but the ELT's for example singer is a company that Can make a lot of these ealtips obviously it's not going to be a LT for get lab And we're not going to wait around for them to create that so we're creating our own ELT in that situation If that makes sense and Clement basically We're treating get lab as a customer number one to make sure that we can meet our needs with the tools And also we have sort of a pressing need to have Answers to questions garnered in data, you know, for example like how effective is a marketing campaign, you know How one's gonna hire someone and things like this, right? And so we're trying to make sure we can we can get that data get those answers So you can make better decisions and better informed decisions as soon as possible and and and so kind of you know As we get time on the way from from getting the data and getting a process and getting it visualized We're also going through to to make sure that the process is easier to use and easy to get started with And so there's a couple of projects around there around, you know, how do we have a data model Given that a lot of the ELT sources have custom fields and you know How do we handle the task of mapping custom fields sort of a standard data model and things that nature and so they'll definitely be work to It's kind of the next phase is to sort of you know kind of trying to manage that problem and to make it More a general purpose tool and easy to get it started with then then right now where you have to sort of customize a lot of things So we're kind of looking for A lot of the kind of ground foundation work we don't even probably queue to and then we can sort of start tackling more of these Sort of customer or two and three and four Troy thereafter that makes sense and I Courtland is saying how do you see Meltano the data team and UX working together? Do you mean from a from a what is the data that UX would like to look at or how are we going to create a UX for Meltano? More the the former so I guess so what I was was Getting that was like to to what extent do we see these as like closely related or like as UX seen as a Customer of the data and intelligence that would be provided or is it going to be incorporated into how we do UX? Any thoughts or or plans there So so I can maybe take a first crack and then of course Jacob and any other team can chime in so I think the UX teams is looking at some other tools amplitude is one which is sort of purpose built for collecting a lot of this moral of like the web event data For usage and funnels and things like that that tends to have a somewhat different data structure than perhaps like more of the standard sequel structure that we're working with and so I Think right now they're kind of going in parallel. I think we absolutely will want to try and make sure we have a common place Try and pull some of these Results that we get in from sort of the web tooling whether it's amplitude or something else And and make sure we can you know represent those looker, but I think right now I think I try and make sure we pick a tool that's great for you. Excellent great for the front-end side They collect those metrics and you know make sure that's working. Well, and then we can pull it in to To the meltano and kind of looker visualization process You know from that point if that makes sense, right? I'm trying I think invent our own thing there And ensure a shoe-huntered in right now Makes sense. Thank you Brendan has a question also Obviously, this is a lot different than our typical open source play. Do we have any concerns about that since looker? our data sources Etc. Are not open source said you might be the one to answer this The open source sorry, I didn't get the open source question. Yes, so the question is Brendan says this is a lot different than our typical open source play Do we have any concerns about that since looker our data sources, etc. Are not open source? Yeah, so This is intended to be an all open source solution So if you look at the meltano read me, you'll see that our long-term goal is to replace looker with jupyter help jupyter lab and Super super set visualizations Looker is kind of an Interim pragmatic solution because the first the first goal of this project is to do whatever the data team needs in the data team needed Visualization and good visualization now we looked at a couple alternatives to looker and they were missing features So we're using Looker, it's the only proprietary thing in the stack other data sources I don't know exactly what you mean with that, but I think that all the the tabs that we use all the importers are open source so I meant like the data like like a sales force or whatever we're pulling data out of you know I am another part of that answer makes a lot of sense. So yeah where the data comes from It's like get lab is open source But you can make proprietary software with get lab like we don't care where that the source whether the source of the Data is proprietary that we're not requiring companies to go all open source just like we use net suite which is proprietary Sure, that makes sense. And yeah, I hadn't seen that in the read me yet. So thanks to the And then Lucas thanks said Lucas is saying are we going to try and do a de anonymization attack to see if we properly? anonymize Yeah, right now I for the anonymization I aired on the side of Extreme caution, so I just basically over anonymized I anonymized to start off everything. I figured it's better to start with a You know Just start on the safe side and then and introduce things as we can reveal that they are okay You won't get much data from a completely anonymized data source, but We'll clear things as we go. I think also to follow on to that We're trying we're trying to you know first not collect PII in the first place, right? So some of these data sources like that sweet, you know have have PI in them And we're trying to be kind of into that when we actually build the ELT and so we just control What data we get in the first place and if there's reasons, you know, I think that the get lab ELT is the first one. We're actually doing like anonymization We're still starting with data that is open, right? So so we're only pulling in you know get lab dash org data, which is already available. We're not tracking confidential issues And so, you know, it's basically stuff. That's already Available out there in the world. So we're trying to You know avoid that as much as possible without having to worry about Anonymization if at all if that's at all feasible and PII stands for personally identifiable information just in case anybody's wondering Yelp says super set looks super cool. What's stopping us from using a short term might have just missed this just joined Yeah So I think you know, it didn't have all of the chart types that we were looking for to replicate for our own internal needs. I Think that was one of the main ones I think we're also has pretty nice integration with the get workflow Which has been very helpful. Yeah, I was on mention that as well. It's pretty neat that Essentially, you can make changes and looker and it gets version controlled and you get merge requests back into get lab And it's a pretty neat workflow as far as how you manage that and how you share things Yeah, one of the reasons of using looker is learning from it They've done an outstanding job of Combining like the data life cycle with like the DevOps life cycle using kids as one of the first ones So I think we're gonna learn from all the good things they did and it will help us to make it that product awesome Does anybody else have any other questions? Not really a fully formed question, but I figure I'll just share anyway, but Amplitude looks to have very interesting applications to trial to paid conversion Core to paid conversion retention expansion You know one paid plan to a higher paid plan upgrade Are we are we thinking of amplitude that broadly or are we focused On a more kind of limited use case in terms of how we're evaluating it So I don't I love this is that A question for for the UX team or or for you for what I've seen And maybe victor's better to answer this because I know he was kind of heading up some of that The initial rollout is probably going to be around some simple events specifically within I think the merge request view And then I'm planning on having conversations with the open mark to kind of figure that out. I mean, I would I would think we'd want It events everywhere where it makes sense and can be useful as long as we're asking good questions that have, you know Good hypotheses about what's what's happening But yeah, maybe that's that's I'll thought I was not super focused on that stuff right now. I think Yeah, I can add some color there Taylor and Cortland. Uh, yeah, and but We can pretty much think about them as separate projects for now And we're just having we're being careful that We don't want to block ourselves in the future So at the outset right now amplitude would be focused on the needs of say the UX team the product team in particular for gilab.com And essentially we're we're paying for You know as quickly as possible getting insights from our users and gilab.com and so we can iterate on the product so that that's um, hopefully a very relatively quick win and You know requires us, you know writing some code on the front end And getting that data and using amplitudes tools in particular say, you know Similar to say google analytics or heap or any other tools where the data lives on their service Since you log into their system and you use the great visualization and charts and so forth there directly So there's very little setup on our part Uh, whereas you what you've seen with update today. There's a lot more custom work that's required now Um In the future there's I can see a number of use cases, you know more the obvious ones is with Uh amplitude, you know replacing google analytics for example on the marketing pages and making more of the integration On the amplitude side, but even further out How can we take that data from the amplitude side and merge it with a lot of the data that we've been talking about in today's call And yes, we verified that yes, they they allow us to do that. It's a you know common workflow You can export that data Via api or on mass. Um, and so That's not something we're we plan to do right now, but it's available for us. Um, and it's we're in talking with amplitude. It's um It's something that a lot of companies do so Not our focus right now, but we we were flexible and it's extensible that way and so when those When we need to have those conversations will be set up very well And as you can see taylor is well aware of of those of our amplitude efforts as well So we're not we're definitely not working in silos. And lastly, let me circle back up front with amplitude part of that plan is that Their customer success team will be engaged with So they'll teach us how to use amplitude in very much the same way that you know Taylor set up a call for us with looker on monday I can only presume that these organizations have have customer success teams just as we do And helping us be successful and using the product correctly Logging the right events sending the right events and so forth. So Um, just a quick update there. Yeah, I I want to add my thoughts here My hope is that if at least for amplitude wants to set up We everyone can just go in there and play with it It's it's very accessible in the sense that you don't have to write any sequel. You could just click around um And i'm fully intending that everyone will have access to that within within github ink to just Click around and explore and see how people are using our product and which features and how they are using it It's then it's really cool. I am looking forward to it awesome Does anybody else have any other questions? or comments or concerns Great, if not, everybody have a great day and we'll see you on the team call Bye