Katherine Bayless: We’re going. Samuel Roberts: Doing alright. How about yourself? Katherine Bayless: , it’s going. It is one of those days, , that’s not a bad day by any stretch, but, , my brain just feels there’s extra, , latency, , it’s buzzy. I’m just … Samuel Roberts: I keep losing track of what day it is. this… the holidays approaching, , it’s just, I… , family’s coming to town one day, and then another day, and I don’t know what… , it’s just… , it’s my mental model is collapsing. , . Katherine Bayless: , we, I just got off another call with, that, the AMS vendor remembers, where they were trying to tell us how invoices work, and I was , it sounds your system just doesn’t support a pro forma invoice, which is clearly what we need, and they’re , , , nobody else has ever asked for that. I’m , , . They fascinate me. But , I’m excited to have this conversation, even though I’m probably a little bit behind on some of the action items that I need to do. Oh, , , I can do the Asana thing while we’re here, if we want. We’ve got those from the thread. Samuel Roberts: … I’m gonna ping Utena, make he’s… Katherine Bayless: , I still owe… I need the S3 bucket with the flat files, which I can do that. do that one today. My delay has just been because I admit they are not very organized on my end, and then I was , I should clean these up a bit first, and now I’m , , I should just send the one set of files and the cleanup can be later. is a common problem for my brain. And then the Git integration thing, I’ll have to set up as . I realized… I started on that path, but then, Ashwini, you had called out, , the service account thing, and I was , oh , probably… probably do need at least one or two service accounts for the team. I requested one at some point, but I’m not if it ever got created. that might be a homework item to get those in place, because now, everything is just some human’s credentials. Uttam Kumaran: How’s everyone? Katherine Bayless: Good. How are you? Samuel Roberts: Alright. Cool. Uttam Kumaran: Doing . I need another coffee. Samuel Roberts: , I was just, I was just pouring more. Uttam Kumaran: I need another coffee. Katherine Bayless: what I should do. Uttam Kumaran: Cool. , what’s a good place to start? Maybe as Sweeney, do you want to start by sharing where we are on the… on the dbt side? And then I want to carve out time to talk about… the SOWs as . , , maybe, Ashwini, you want to start there. Ashwini Sharma: Jaw, , on the DPT side, , second… where are… where is my browser icon? . Let me just quickly show how it looks , ? there is an integration that I said, ? I used my user ID and access token to create this, normally, , what you would see is something this. dbt project, and this is all CTA dataOps, the same repo, GitHub repo that we have, that we are using, ? And any changes to that repo has to be synced, now, it’s, it’s… Manually, , you have to do a pull, and then it refreshes. But the cool thing that you can do from here is, while it pulls. you can, you get this screen, ? Once you have a dbt project, it loads the profile file that we have specified in dbt project. And the profile indicates, , where do you want to generate your models, depending on what area that you are running, ? , for example, , let’s say somebody has pushed a PR into the dev, ? And there are going to be multiple developers who are going to be working on different, different models, different areas, and you don’t want , at one point, you, During the initial phase, , you don’t want developers to mess up each other’s models. And the way they are going to work is all the work is going to happen locally, they’re going to run dbt commands from their local laptop. And when that happens, it creates the models within their user schema, ? , for example, , See if I can… Show you something over here, oh, over here, . , if you see this, Catalog Explorer, ? Dev staging… you can see these schemas that are appended with my name, ? that’s, , I am a developer, I am running things locally, but when things run from, dbt project within Snowflake, everything gets dumped over here, ? And as we move across different environments, , for example, , , 5 developers did some work on dbt. Two of the developers are going to promote it to QA, , where… and in that case, we’ll move to staging. Where we are going to create those mods, and then out of those two, one of them gets a QC clearance, all good to move to production, we move it, and then we’ll create it in the production. happens through this dbt run, for example, , it’s relatively easy, ? You can directly click run, or you can specify some arguments on top of that, and then you can click run, in which case it will run it. And you can do this through tasks. You can, which is nothing but a Quran. And you can schedule, , at what point do you want it to run. The other thing that I wanted to do, which I will be doing, is triggering these things from GitHub, ? … essentially, , when a developer pushes a PR, we need to run some basic tests, ensure that the schema that is generated out of those models are correct. And that should happen via GitHub Actions, which I’ll be incorporating in the repo. As as we need to refresh this repo, , every time somebody is merging PRs into different branches. that will happen through some snow CLI. , I’m working on those things, as as I’m working on that, active members report, which I should be done… done. Max by tomorrow. I’m sorry, I’ve been telling you, I’ll do it, I’ll do it. But , some other things that are coming up in between. Katherine Bayless: Alright, , I’m late on sending my stuff too, … Terminology question, is a DBT model, . What is it? What is it as a unit of code? Ashwini Sharma: It’s, dbt is a glorified templated SQL, ? ? And the way it looks is… this. Katherine Bayless: Are you able to see my screen? . Ashwini Sharma: , we have something called models, ? And the basic unit of model is a SQL file. That translates to a table in the warehouse. ? And then you build on top of those models, ? … , in the extremely low level, we have those raw tables, the reimbursed raw tables, and the first step is to ensure that we follow a standard naming convention, which is clear. Easily understandable and consistent across all different data models, … That is what I’ve tried to do in the staging layer. if you go in staging, , now we only have source for remembers, but eventually we’re going to have a source for Salesforce Marketing Cloud, we’re going to have something for Shopify, and on, ? , that accumulates over here. And if you look into CRM, we had the CRM, sorry. remembers data was, , grouped together into different functional areas. One of them was CRM. , if you look into some of the CRM staging table, , for example, country, what I’ve tried to do here is, . Column names this are typecasted into a… I’m using a snake case, ? And what this does is it will create a table in the backend, ? In this case, it’s a view, not a table, you don’t want to replicate tables. In the staging layer. And it’s just typecasted into something consistent, and it follows the standard naming convention across All the different columns. And once you have the staging layer ready, what you can do is, you can go over to the intermediate layer, where you, again, group it based on, different functional areas. For example, , I’ll create some of the models for CRM, Which is, which is… which is going to be entity-specific, ? , , for example, a customer entity. Or organization entity, and on. And then, in the mods, what we’ll do is we’re going to do the dimensional modeling, ? we’re going to have a DIM customer, or… dim organization, or dim country, or things that. And then we’re going to have facts, which are utilizing this dimension tables. And also, , some of them are not going to be facts, just facts and dimensions. Some of them are going to be reports, , which… we are working on now. , for example, this active membership report is just a report. , what you’ll get in Snowflake is these tables, which will be exposed to analytics. and reports, which will also be exposed to analytics. The other tables can lie there, , in the staging layers and the intermediate layers, but everything in the MART Is what will be used for analysis. Uttam Kumaran: , , all of… but all of this, , all the naming convention, folder, structure, schema is all us. , dbt does not… they have guidelines, of course, but they does not prevent you from not doing that. This is after… Katherine Bayless: just doing dbt work. Uttam Kumaran: And we haven’t… we can… I’m happy to… we have a doc about how we do DPT that I’m happy to share that is… , , mix of, , several different, , Great. dbt programs that exist, , how dbt structures their own dbt project, how, , GitLab does it, and we felt this environment structure, it usually works. The best… the thing we want to enable, and I particularly care about, is I don’t want… engineers to come in and have to think about where does a file go, how do I name stuff, and where can I go find… , I ran something, where do I go find it? Those are all, , extremely taxing things that, , are just, . Really can… can just, . cause a delay in the pace at which you do things in data, we’ve tried to solve a lot of that here. Additionally, given that things are now being built a lot by cursor, cursor-assisted. Having a really clear naming convention structure is, , much better for that, to just follow that, , versus it makes its own stuff up. , that’s, , generally how it is. And then the innovation here that I pushed Ashwini to think about is, , it’s… I don’t… traditionally, we wouldn’t… centralize this in Snowflake, but given… one, , CTA and your guidance on, , hey, as much as we can do in the tools that we have, do it. And, , they started building a tighter connection with dbt. I was , hey, maybe we should go after doing this directly in Snowflake this time. And then we’ll figure out the kinks in GitHub Actions and merging back to Snowflake, , I’m not worried about that, but I’m happy to see that because it reduces having to do it, get another Vendor, , and you get the exact same functionality. Katherine Bayless: , , , , , I’m super excited, to be able to start using this, because, , I… I don’t know if you’ve noticed in the repo, but there’s totally, , the migrations, branch in there, and, , it’s my fumbling attempt at a similar idea of version controlling a database, , at the table and, , data level. this is… this is gonna be interesting. I feel it’s gonna take a minute for my head to, , learn the new vocabulary, as always, ? a really in-depth, , training for Kyle would be good too, since presumably he’ll be one of the people writing a lot of the models for this stuff as we. . But , , I’m… this is cool. This is really cool. Uttam Kumaran: , ideally, , folks Kyle and others come in and they… they work in a couple ways. One, they find, , some intermediate tables that they can use to create marts, or, , eventually they’re able to the data gets landed, and they can build a lineage, but ultimately, the time isn’t really on, , writing that, it’s just, , a debugging. , , oh, this column’s empty, , I need to trace back where it came from. Or, , I need to… I want to get this code reviewed. And , , that’s where having a PR review process is really great. as part of the PR process, we will run CICD checks. Is this… is this running in Snowflake? We can run testing there, … Just getting this now just, , sets us… sets a lot up. For, , the people that are gonna join to start building on top of this, , … Katherine Bayless: , , totally. I’m , I’m looking forward to the day where we have multiple developers, and would. Uttam Kumaran: . soon, ? Soon, … It’s already set up in a way where , , probably by the end of this week. , at least Kyle, we can start to onboard and show him, , how to… how to get access to some of this stuff. , and then he can start pulling… at least pulling stuff from Martz would be great, because then he can start to have requirements for us to develop, … Katherine Bayless: , , , that is definitely the entry point, because I don’t think he has any GitHub, experience. If he does, he hasn’t, mentioned. , , that’s a new model for him to work in, and then I know, , he’s in SQL, and he’s very aware that AI is great at generating SQL, if what you’re looking for. And , , building out some of those muscles, too, . He’s a quick learner. Uttam Kumaran: Cool, that’s… that’s it. , Sweeney, you’re… you should have on your side a report ready this week, at least probably tomorrow or . Ashwini Sharma: I’ll have that ready, ? Uttam Kumaran: Any other… anything else, Catherine? , I want to make that we also… document a lot of this effectively. , we’ll… in the repo, we do have a lot of documentation, on, , name… we’ll have documentation on naming conventions. also we’ll… we should make that we need to also put in, , any Snowflake-related grants that we’re running. But is there, , anything else Catherine, , we should keep aware of there, or, . you want us, , while we’re in this , , initialization phase, to, , save. We’re also happy to, . walk through, , what is dbt and what is Snowflake to, , ever… for, , other people, too, if it’s, , it would be helpful. Katherine Bayless: , , for the team, generally, that walkthrough would be good. , it’s funny, because, , Kyle and Kai are both, , they’re , , we need snowflake training, and I’m , I don’t know, I’ve never used it before either, I just click buttons. But , , , , something that would be useful. , I don’t have… tons of questions that I feel I need to ask on this call about DBT specifically, but… . It takes some time to, , just get familiar with the vibes. Uttam Kumaran: Great. we can… Maybe pivot to talking about those scopes? I don’t know, Sam, maybe do you wanna… Share screen, and then we can… Ashwini Sharma: , go ahead. A question, regarding the remembers data that has been shared to us. At one point, we were thinking, , maybe we should ingest that data into the Snowflake instance that you have, because it’s a third party that is sharing in through Snowflake Share. , how reliable is that data? Is that going to be available to us all the time? As long as we use remembers, or is it, at one point of time when you stop using it? If you do that, ? That’s true. Katherine Bayless: It’s a good question. , they… we have a… , it’s part of our contract with them now, we’re paying an additional annual fee for them to deliver the data this way four times a day. , in theory, if that breaks, then we get to go yell at them, because they are not delivering on their contract. However, it is our intention to get a new AMS at the end of this two-year contract, ? , we were … the team was interested in going out to bid when I started, and I was , can we just wait? Because it’s a big move. , it’s one of our biggest CRMs in-house, and I want to make we get that transition . we signed a two-year renewal that’ll start Jan 1st, and at the end of that, we will be on a different system. Yep. I don’t know what, . Uttam Kumaran: , Ashwini, we should just… you should just keep, , run maybe one backup. Script, and just store it, and then we should just execute that. every month, or something, where we just drop and clone that. , I don’t think you… just continue to use the share, because if there’s an issue. I… I just want to avoid the fact that, , maybe we copied something, or we’re using stale data, but maybe we’ll just keep a copy every month or something. Katherine Bayless: , holding onto a backup isn’t a bad idea. And , , , the monthly backup is not a bad idea, because I have run into this before with, , AMS vendors. They don’t do a great amount of, , changelogging, and sometimes, if you want to be able to see, , granular changelog-type data around, , a membership and, , how often it was updated, you have to be storing that data along the way. , they’re not going to be able to give you every thing that happened to, Roe, which is… An annoying limitation. But , a monthly backup of it is not a bad idea. Ashwini Sharma: , cool. Alright, . Katherine Bayless: Also, I’m a data hoarder. Uttam Kumaran: , I know, that’s how I am. Let’s just save it somewhere. Katherine Bayless: Storage is cheap. Samuel Roberts: , shifting gears a little bit… Share the , … , this is also in the repo, . I’m just looking at it in GitHub, because it’s all just marked down now, but, we can go wherever, but we looked at the… two things we discussed before, which was the Okta being used for all the registration, and then, Shopify. side of things. And , really, , this is just talking about doing some discovery work on that for the Okta side. Specifically, I believe Jay was fine with workforce stuff. It was really just that. This just walks through evaluating different vendors, looking at the different tools that are out there. This highlighted Auth0 and Clark. But, , there’s a number of different ones we could look at. And just… Trying to figure out the best way to start with that? Excuse me, sorry. . Katherine Bayless: Sorry, one sec. the coffee went down the wrong way. Uttam Kumaran: You’re all good. Katherine Bayless: , I will say… Samuel Roberts: in there. , sorry, go ahead. Katherine Bayless: I was gonna say, just to buy you time, Jay was looking into Clark a little bit, and he did seem quite intrigued. His wheels are turning around this, . Samuel Roberts: , , great. , I know we had… we had discussed that one a little bit. The other side of that is, , is it clerk that also does payment stuff, too? that might be a way these two things dovetail together, potentially, but that’s not really outlined here. That would be after we do the work and figure that out, but, , , the only thing that we were wondering here was, Outstanding questions… oh, there’s outstanding questions as , but was there anyone else that would be… Need to be looped in on some of this stuff. … And we talked about, , some customer… , there were some issues, , we weren’t… I wasn’t … who else was… Katherine Bayless: , , it’s one of those things where it’s , there… really, it’s me and Jay, but there should be a different answer to that question, ? , I’m trying to get him to start, , , if you’re the strategy, , VP-level guy, , you shouldn’t be in the weeds on Okta, but at the present, it would be him, … , . , . , , I don’t know if it, , makes sense as something to call out, , in a scope per se, but, , just one of the, , lenses on my brain as we go through this is, , , Jay is very, . he believes that, and I don’t disagree with him, that makes it sound I don’t believe him, but, , he thinks a lot of the trouble with Okta is not, , necessarily the configuration, but just, , the things we ask it to do based on our policies. And I’m encouraging him to also start, , surfacing those things, , not just with Okta, but across the board. , if we have a rule that we made 30 years ago, and it’s causing us to need, , , custom config, or a whole bunch of, , spaghetti code, then, , we should change the policy, or at least revisit the spirit of it, rather than just continue to code around stuff. , I see it as, , my team doesn’t clean dirty data, that’s the, , data stewards owner. his team shouldn’t be responsible for cleaning up dirty policies. , . , … , surfacing some of that through this is, , good. Samuel Roberts: , , … , that would be a big part of, , doing… , figuring out… because I… it still seems odd to me that Okta is the tool that’s being used for however many, I forget what the number was, but… , the 150, , that’s . I was just, . in my experience, that seemed, … but I don’t know what policies are, , coming down from other places. For… but that would be what we would, … Uttam Kumaran: It’s also, , again, if they had to pick… if they had to pick in a jiffy, or if it was, , we’re already using this, just use them, and there wasn’t, , a discovery process, then it’s not surprising. And that’s what I , … for Jay. I want to think about how this process makes him, , either we’re confident that, , we can do this in Okta, it’s not we… we’ll see, , if we were to support this use case in Okta, what changes, ? And, , possible or not possible. We will also explore the other, at least you’re, . for the next 5 years, you, , have a canonical, , we thought about this, ? And , at least you’re , what if next year we’re , what do we… why don’t we decide this? The answer is, , at least that team at one point decided, , they have this document that you can reference, ? , , how we’re thinking about this repo, too, is, , I… the thing about coming in as a new data team or a new member is, , the speed by which you can ramp up or understand why decisions were made. , one, hopefully. allow you to, , not question those if you’re , oh, I understand, , the reason why we did it, but also able to hit your… whatever your first deliverable is much faster than being , I still haven’t been able to get it running locally, or, , couldn’t find this thing, ? … Samuel Roberts: I was hitting that just accessing the GitHub repo with Okta, where… GitHub kept directing me to cta.octa.com, whatever, but I only was able to log in properly through octa.ctta.tech. slash whatever, and I spent way too long, and I was about to message Jay and realized, wait a minute, these URLs are different, and all I did was replace that, just worked, and I was… I don’t know, , there might be something that surfaced here, too, but I… . Katherine Bayless: , . , honestly, , that’s another one, because I, , I know in his mind it’s not, , that big of a deal, but I’m , there are definitely people in this office who get very annoyed when they see it go from screen to screen to screen to screen to screen, ? , that’s not a pleasant experience, ? And … Samuel Roberts: And I was completely locked out, and if I hadn’t. I don’t know what someone else might have done if they didn’t, … if they’re just a little less technical than me, because that took me a while to find that, too. I thought there were internet issues, that was… it was a whole, , it went down a long one, but anyway, that’s, , proving the point here, , that there’s some… Katherine Bayless: , . But , agreed, agreed, agreed. Samuel Roberts: Sorry. , but , , , the idea here is just to put that together, do the evaluation, come up with some recommendations, and then , , with optional implementation here as , . But , definitely take a look at this. Katherine Bayless: One other, , tiny question, which. don’t over-interpret, I don’t know that it necessarily belongs in this, but it occurred to me earlier, , Okta has the dashboard that you land on, . if we do decide to continue using the customer dashboard, it might make sense to, , leverage that. , now, people tend to only hit on it by accident. But if we… if we’re gonna pay all this money to have that level of solution for that big of an audience, , we might as at least see if there’s something useful we can do with that dashboard, because it’s probably the only single touchpoint that you could reasonably expect any audience member to hit. , not everybody will hit ImpactCM or re-members, not everybody will hit… . , , … . Samuel Roberts: the point. , I… and that’s the one where it would got me one way to GitHub, but not back the other way. , it was weird. But there’s definitely… even using that, I’m there’s… there’s all kinds of things to sort out. Interesting, . Excuse me. Alright, that’s, , out of scope, requirements and inputs, I don’t know. , deliverables, the discovery, evaluation, recommendations, I don’t know if there’s more to dive through into all this, if you want to take a look. I don’t know if you want me to read through it or not, but I, … Katherine Bayless: , , , , I can dig in on it. Uttam Kumaran: Maybe let’s go through… can we check out other open questions on this? they’re all at the bottom. Samuel Roberts: , they are. Uttam Kumaran: , , , this was , , what we… Are going to try to answer. , , net-net here, is , , what is this… I also want to see, . How many tickets are we getting? , what are all the current systems that are going through it? you can … , go ahead. Katherine Bayless: Oh , I was gonna… , you can fish a sentence, but I do have a comment on that one. Uttam Kumaran: , and then we scope, , how much this impacts, and then… Talk about, . , , what… what did… we’ll circle back to see, , did we talk to Clerk? Did we talk to these guys? , why… why did we, why not? Looking, and then try to see, , , can we… can we arrive at, , an option or two to try, ? Katherine Bayless: , I was gonna say, , definitely with the system integration thing, there are a handful of the apps that go through Okta, but not everything does, and as much as Jay is , ugh, when I say these things, I’m , everything needs to be behind the SSO. And , , what it is currently is definitely not what it should be. And I’m slowly making headway getting people to want to bring platforms under enterprise governance, and , , the ability to, . Onboard and, , set up new systems behind it more rapidly, even if they aren’t , brand new… , it was easy to do Snowflake, because it was brand new out of the box, ? . But, , if he’s gonna bring Salesforce Marketing Cloud, for example, that’s one that would to go through Okta starting next year. , he’s gonna have to be able to, , preserve all of the roles and permissions that are currently in there, all that, , blah blah blah, . more enterprise governance. The tickets thing, I don’t know if we’re gonna have that data, to be totally honest. , Jay can correct me, but there’s certainly some of it, I’m , with our MSP, and probably some we could pull out of, , his email and Slack, but… , ticketing has not been a strong behavior at the organization, historically. Uttam Kumaran: , that … maybe we can… , the ask here, Catherine, is for you just to look through the phases, and then be , is this… Does this roughly seem fair? Is there stuff that we want to add, or stuff we want to remove? , , out of the end of this, we should be tired of talking about this problem, is my hope. , and we have some clear guidelines. Cool. We can stand, we can talk about the other one. Samuel Roberts: , the other one was the digital asset, delivery on Shopify, which, is very similar discovery, making we’re using the tool thing, , fix it, change it. decide it’s fine as is. Whatever that is, the same idea. Excuse me. But , that’s, , roughly the same thing here. What are the open questions on this one, I wonder? , everything, all the details, Ruby stuff, Shopify Okta, , all that, . , they’re related as , but. Uttam Kumaran: this is where, , , it’s just for us to understand, , what is being sold through there. and give you, , one document on just, , … , , what is… what is Shopify being used for, and then what are other… , opportunities to streamline this. we go through, , pieces of, . sharing digital assets and selling digital assets, , what options are there? But, , this is , again, the… the questions that we want to answer, and, , if we were to pick another thing, , how do we run both, and test, and… , how is the team able to customize this more easily? Katherine Bayless: , , because , , finding out that our sponsorships are being sold through Shopify made me, , oh, , interesting, . it still doesn’t make me feel very, , committed to the Shopify solution, but it does raise my level of curiosity about this and how we’re using it to solve problems, and then, , if it does wind up, , Sam, to your point, that we decide to keep it, make it a little better, but keep it, I’m , , , what else are we selling internally that we probably should then use Shopify for? , I’m all about, , consolidation into, , single lanes for certain types of things, and I have a feeling we are selling lots of stuff piecemeal in different systems that, . Again, if we keep Shopify, maybe push those that direction. Also, one of the folks on our market research team said he can’t understand why anyone would ever purchase our research, which is funny. Uttam Kumaran: , , , the biggest thing in the first phase is, , we’ll get access to a bunch of stuff and be , here’s… here’s, , what’s in Shopify. … , for me, the biggest things here are, , you… can you… should you build your own on top of Stripe? Are there other easy digital asset storefront opportunities that are worth considering? the other piece of this is, . for us to think about, Katherine, is, , this is something you own? Is this, , a mix of you and some people? Or, … , what is the ownership over this? And, , that’s, , really what we highlighted in Phase 3, which is, . What is the ownership model of a tool this internally? And… and then the other piece is, . reporting. , how is this data getting into reporting? , that’s, , that’s the full circle thing here. Katherine Bayless: . , , , the who owns this system question will probably become timely very soon, because the current quasi-owner for it is a very lovely woman on our marketing team who does not want this responsibility and is about to go out on parental leave, … I have a feeling I’ll wind up being the temporary custodian. Uttam Kumaran: When is that? she’s going any day now. , . Katherine Bayless: , , and she’s, , the only person who’s, , got the actual, , account owner status, and , , I don’t… Uttam Kumaran: a question on this is, , when… are people adding new digital assets? Are they go… are they, , going through her to just get it added? Katherine Bayless: , … if they’re… Uttam Kumaran: Because Shopify, there is a lot of configur… there’s, , I’m just… it’s just crazy, there’s a lot of configuration. It’s hard to run a single person Shopify store. Katherine Bayless: , she’s really not even necessarily, … Necessarily, , actively doing or managing. , , … most of the digital assets flow through a probably broken integration from remembers to Shopify. . The sponsorship stuff, I have idea. Casey doesn’t want to have to deal with this, but I can’t just, , say, , , just give it to me, ? I want to give her the, , here is why we will transition this ownership to either my team or the IT team, but… Uttam Kumaran: , I wonder if it’s a mix. Katherine Bayless: I’m a doctor. Uttam Kumaran: , and I wonder if it’s a mix of you owning the platform, but then empowering people to go in and, , update Products, or, , change pricing, or, , , versus… you taking it all over and then just becoming, , storefront owner is, , , that’s… . Because this is a lot of work, , we support a lot of e-commerce clients, and… , it’s just, , a huge system. , it’s just, , too much to have to manage in there. that’s why I don’t know if, , if it’s… if it would be easier to, … again, if we do it on Stripe, and you can have, , an admin panel, and you’re just cha… people are just… all they’re doing is, . changing prices or titles , , I don’t know, I wonder, , if there’s a more… if there’s an easier option, I would have to just look at, , what… how it works now, but… Katherine Bayless: , , , totally. Uttam Kumaran: Cool, that’s both of the scopes here. , Catherine, from you, it’d just be, , confirming that these are, and then I can … try to give you a sense of, , pricing. I… I know that the… I don’t know if any of these are … , is the CES registration one, . We’re trying to solve that before… the CES, or is this , we would to kick off something that maybe there’s… if there are any wins. , , how do you think about that? I assume, , the Shopify one is… A lower priority if you had to choose, but… Katherine Bayless: , definitely the Shopify thing is the lower priority, but, still… If you had to choose, . , , . Now, the registration thing, … , , it’s a… it’s a bit of a rock and a hard place for me, to be honest, because it’s , should we solve it before CES? . Should we really put a solution from me in place before CEOs? Samuel Roberts: That’s what I… , I made the assumption here that it was not… Uttam Kumaran: , and that’s also what I… I made that… I tried to make that clear, too, but… Katherine Bayless: I’m, , I’m just asking again for… Uttam Kumaran: Confirmation. Samuel Roberts: , Discovery, who… , we can… but . Uttam Kumaran: , it’s tough, , I really think that you can make a big… mistake here, , changing some of this, I wonder, … I don’t know. … Katherine Bayless: , I genuinely, probably what we will do is limp across the finish line and just vow to not let it be this way next year. . that might be the most correct… it’s the Batman thing, ? It’s the answer we need, not the answer we deserve. But, I also think, . there might also be some ways to creatively salvage some of the Okta stuff. I was doing a little digging, , the one piece of the immediate use case, which is really nothing to do with authentication per se, but the reason that this was also floating to the top of the pile around CES was that without Okta IDs from the workforce tenant in that registration system, then they’re using email address as the parameter in some of the, , API calls that are going around between these daisy chain integrations, and … in order to not use the email address as the query parameter, we need to have an ID present on every record that we can use, ? the original, , alarm bell that went off was that Okta IDs aren’t in merits. However, we have learned it is Okta Workforce ID, or people who exist in the workforce tenant. don’t have a customer Okta ID in merits, it’s only, , 485 people. To be fair, it is 485 of the most important people, probably. But I’m , , , can we hotfix that? , I can patch those IDs into merits, and then we could use that Okta ID in the query parameter. It doesn’t solve any of the authentication piece, but it does solve that, , security flow. , we might be able to at least solve that narrow piece of the problem with a little bit of duct tape and popsicle sticks, and then, to your guys’ point, , … Do this the way with enough time to not, , over-promise and under-deliver. Fingers crossed. Uttam Kumaran: that’s why it’s also, , if we end up doing this, and we start in January, then at least one of us is also there poking around. It’s , then that makes 3. , or we’re, , one entity, but , , tech… … That may also help if there is, , Just another eyeball, and again. it’s , I know you and Jay are… that’s, , 5 or 10% of your time, … That could be nice to have someone else just look around. Katherine Bayless: . But , I’ll take a look at both of these in detail, I’ll get Jay’s thoughts on them, but , should be able to get you guys feedback pretty quickly. And they look great at first blush, … Uttam Kumaran: , I just want to make that for all of these, I , , I’m , just let’s put everything we’ve heard about the problem, you can always be , , that’s not… we did that, or that’s not worth… instead of… Instead of not having not making it verbose and then missing items, , … , cool. And then , , this week, I want to make that we deliver the first report, and then we’ll also have, , DBT stood up, we have, , Snowflake now running. a good goal, honestly, Ashwini, is maybe we can try to, . I don’t know, , it’s already tight, but maybe we can try to… see if Kyle wants to meet, Take… at least take a look at some of that data we can get them set up in the repo this week. That way… And Casey’s poking around, I don’t know, Catherine, are you guys all out next two weeks? Katherine Bayless: , , with the show, , , some people are bold and they take leave, but for the most part, we’re in the 3 days and then off the 2 for the holiday. Kyle is traveling to Utah, he has a couple additional days of PTO between Christmas and New Year’s, maybe? . But you definitely will be able to catch us next week. Uttam Kumaran: Alright. , great, because, , I’m just, , we’re… we’re just trying to… a lot of people are off, a lot of our clients are off, , the last two weeks, … And then maybe, Ashwini, we can see where we land here, and I would love to just do a little bit of, , a dbt Snowflake onboarding for Kyle, and that way, also, we can record that and leave that in the repo, too. Katherine Bayless: , that’d be awesome. Uttam Kumaran: And then, , I haven’t heard anything back from Polyatomic yet. he said… he mentioned he was gonna try to grab time, I don’t know if that ended up happening. Katherine Bayless: we’re gonna call Thursday at 11, I wanna say, off the top of my head. , we need to get that on books. I sent him over a few things, that I had to hand. I also… I did hear back from the guy who’s doing the event co-pilot, , the little chatbot thing, for CES this year, and he said they don’t have an API yet, but he was very… , he was very kind and responsive, and he’s , we can build one! I’m , that’s , just asking. And he’s , we can send you all of the raw data directly to an S3 bucket. I’m , , that sounds better than an API. Uttam Kumaran: Please, just… we can get anything? Katherine Bayless: , , , . now, we get an email digest. But the other thing of interest that I found that I’ll mention to, the Polyatomic guy on Thursday, is apparently we do get some requests for reports on scanner data at CE. apparently… I don’t… I don’t think Merits has an endpoint for that data, but, , they’ll give us two flat files a day, , which is fascinating to me. But , I need to… I need to do a little, , legwork before the call Thursday and figure out, , , are we getting flat files because there’s truly endpoint, or is there one and we just didn’t, , set it up in the past? Because if we’re getting on-site requests to be, , reporting on that data, then I would much I’d rather have it coming in programmatically than via flat files. And building out a single workflow around one endpoint is not the end of the world, even if it’s not the most, . Forever code. Uttam Kumaran: , that would be helpful. , I would love… I would love to see that data, and then, , , maybe that is something that gets… monetized through whatever the new system is, but… this is the exact work that I did at WeWork, where we took that, and as part of the QBRs for our enterprise clients, there’s a huge section on looking at the swipe data for the people that went into the WeWorks, , their employees, and it led to, , oh, you need to buy… again, it’s all upsell related, it was , oh, you need to buy more space, or you had some people come in in Berlin, maybe you did an office there, , things that. I’m the salespeople, if we could produce them, , a nice dashboard, they would be, . Cool, perfect. This is what I needed to go sell more space and more… more ads, ? Katherine Bayless: I… this is where I’m , ugh, I feel bad for all the things the AI companion hears me say. , the big innovation from our sales team this year is adding another day to the in-person space selection, where people can go stand in a tiny, crowded room and get really angry for a few hours and try to book their space for next year. Uttam Kumaran: Wait, what is the in-person… what is the in-person, what is that? Katherine Bayless: , on-site at CES, you can purchase your booth for next year. you are willing to go stand in a tiny room at the Venetian and get angry, and we also, we bring literal, physical, cabled, networked… we set up a LAN. , we’re gonna spend an additional $12,000 this year to have a few more hours of that be available, and I’m , iPads, guys. iPads. to the booth on the floor with an iPad. We’ll get there. We’ll get there. Uttam Kumaran: , we’ll get there, we’ll get there. . Katherine Bayless: , space selection, , there’s a lot of servers. , it’s. Uttam Kumaran: , we did a lot of space planning. It’s weird, this is, , very similar to some of the stuff we did at WeWork, but… , , we use the key swipe data, we use the badge data to show that, , oh, all your people are getting desks, , they definitely need an office, you should sign long-term space here, you should grow. And, it was an easy story, because without that, it’s, , all , . anecdotal, and the account… I supported the account management team for a while, and it’s hard for them to tell the story, but that is, , one of the key things, and … I don’t know what other data store you can, , the data team can support at CES. It’s , it is the ba- it is , how many leads did you get? Maybe what… Other types of digital engagement story you can tell, but it is the floor. Katherine Bayless: Theta, and … Uttam Kumaran: badge data, and you can say, , oh, , this year we saw that, , there was much more peop… the people that got this type of booth got much more, , maybe you should consider that. I don’t know what… , I’m they’re… they’re very crafty, but that’s… there’s a… that’s probably the only real… measurable data story, ? , apart from just total attendees and, , Potential impression volume. Katherine Bayless: , , , , at present, that’s the upper limit on what we’d be able to do. I have many ideas about where we can go in the future, but , it would mostly be, , traffic, ? And, . . I, I, , … There’s much potential. I did find out, that Merits does do the, , RFID in a badge thing, you can, , track people everywhere, and the woman that runs that team is open to revisiting. we had looked at them in the past, and it was, , prohibitively expensive, but I’m also questioning what prohibitively was, to be honest. Uttam Kumaran: Cause I know there’s some sticker shock troubles, and I’m … But if you can tell the sales story on top of that data, then it’s … And what you can… what we should do is we should ask them… we can ask them for a sample, put something together, and be , guys, if you had a version of this, , what could we sell? Could you sell more space? , if you had, , this , . a CES review doc that showed foot traffic that, , … Katherine Bayless: , somewhere I have a place where I, , sketched one out. Interestingly, the team that’s hungriest for it is the membership team, because they do their sales prospecting largely off of, , CES exhibitors and attendees that, and they’ve asked for this. They’re , we want to know, , if an exhibitor got 3,000 comp badges because of their booth size, where did they go? What did they do? Who did they talk to? all of that stuff, and , , the membership team is really. for sales data, and conveniently, now that we have to remember stuff, , we can start to really knit that together, even if it’s just in Snowflake and reporting for the moment, eventually pushing it back into their system, maybe, but . , they’re hungry. , we’re actively working on building out a member engagement, , report with them. Kyle and Kai are working on that. And some of it is, , just, , did they attend CES? Did they have a booth? But we can get more granular. If you want to see the historical scan data, it is in S3, it’s… , I can post a link in Slack, I won’t remember, but it’s, , under that archive bucket. DBO, or marketing database, and then DBO, and then it’s, , CES scan history or something, but it, it goes back further than I thought, we have the scans for all the sessions going back to 2019, in there. , . , , great. We’re , oh, hmm, not bad. , I’ll figure out how it’s coming in this year, but… . Potentially. Potential. Uttam Kumaran: , perfect. that’s all I had. Katherine Bayless: , I have a couple questions. one of them is a practical matter, that, , webhooks database that I had set up that I could, share the show floor tour survey responses with the guy in Market Research, eventually, I won’t need it anymore, because the show will be over, and he will longer be collecting show tour information. Should I plan to just… delete it at the end of CES, or do we want to move… any of what I have done into the structures we are building. maybe that’s a clumsy way of saying, , it’s a V0 version of a form stack data mart, ? , because that’s what the data is coming through. I’ve just filtered it to be, , the one form that Chris could see it. But I don’t know if that means there’s anything really of true utility in terms of the stuff that we need to build out. Uttam Kumaran: What do you think, Ashwini? Katherine Bayless: You also don’t have to answer it this second. Got it, but . Ashwini Sharma: I don’t know, , any data, let’s just keep it, ? You never know when it is going to. Katherine Bayless: They’re all data hoarders. , it was more just , should I delete, , the webhooks, , database that I created in Snowflake, when we’re done with the show? . , I’m gonna say raw data. Uttam Kumaran: I would just… we should save it, and we store it somewhere in Snowflake Raw, . Katherine Bayless: , we’ll migrate it from the structures. , . Similarly, the role-based access control stuff. I could use some help figuring out, , what the… not policies in terms of, , literal policies, but, , me approaching this, , 150 people work here now, I’m asking Jay every time, , can you give somebody access to Snowflake? probably it makes more sense for him to, , set it up that all new employees, maybe at manager and above. , by default, and then based on, , what team they’re on or what role they’re in, , I’m giving them, , maybe light read roles on production. , that’s the piece that I need to understand, is , how do I… how do I get people into using Snowflake? Uttam Kumaran: Totally, , if we’re at that, if we’re at that point today, then we, we, we typically run, , we have a script of, , how we architect Snowflake role-based asset control. Typically it’s a mix of, , reader-writer roles for various environments. , , you have devs, staging, and marts. , you have read or writer on each of those, and then, for example, Polyatomic just needs read and write on, on a certain database, and then something needs another. , dbt needs just transform roles, and the BI tool just needs… you’re able to do that, and then you can assign reader-writer roles to higher levels. , example, you may have a… a data engineer role, you may have a data analysis, an analyst role, you may just have a CTA employee role that is, , read on just marts and, , nothing else. , it’s very easy for us to architect in that way. The thing to be worried about here is, . One, you don’t want to do direct grants to users. Second, you don’t want to do… you want to have these, . functional roll-ups, it just makes it a little bit easier to be , , I have a user coming in, , they need reader here, they need writer here, and that rolls up for the most part. The data engineer role will maybe be able to create, , pipes and ghemas stuff, but… it also allows us to support service accounts, , , the DPT, , if we were to have a BI tool service account, it would just have read access to marts, that someone can’t come in and, , read from RAW, and then mess up by, , reading something super upstream, ? Katherine Bayless: , similar to the webhooks thing, , my quick and dirty, , get this out the door, I did… I put Chris… into the, , Snowflake analyst role or something that, and then I gave that role permissions on the webhooks thing, all of which I assume will be undone and redone in the correct structures. I just was , I don’t know, this gets you in there now. But , we have people who I would to be able to get into Snowflake to look at, at the very least, , dashboards built on marts. And , Anna, the brother, and Chris, and Erica. are all in there currently, but I’d really… , I’ve got people, , ready to go as soon as I can give them a green light, … Uttam Kumaran: , Srini, we should run through our typical… I don’t know if you ended up running that, but we should… we could run Catherine through what we typically set up, because I also want to talk about warehouses. we set up, , the different warehouses for the different workload types, typically. something for ETL, something for Transform, something for BI, and reporting. And it’s all based… we have this, . typical long script that we just, , run through. we can share that and then make any tweaks, that we want there. But hopefully that will allow a very flexible access structure. And you, you have to worry more about giving… putting someone into… , just granting one role to someone, versus, , oh, this person… , you shouldn’t have to run grants, you can do all this through the UI, , … Katherine Bayless: Add a new user, they have the… Uttam Kumaran: CTA employee role, and then if they graduate, maybe they have the data analyst, and then it , , goes from there. Katherine Bayless: , , , . , I would… if we can do that, that’d be awesome, because there are enough of these people who are really hungry to get in there, and enough things that I could, , really throw together in, , little dashboards, for CES reporting, some of the things that people are going to be asking on site. , not everybody’s gonna be , oh, let me fire up Snowflake, but there are people who will, and I feel if we can meet their needs, this year, then that’s a huge win for us, … . , let’s see, there was another question that I had for… Bold. , , this might… this might be, , a… I’m not how to ask the question, or in a way that it makes sense. , , we have that, enterprise, , AI intranet search thing, Glean. And, , Jay has tried in the past, , connect it to raw data, and I’m , please, for the love of God, do not do that. But what I would to do is give Glean somehow, someway, some visibility into, , the data catalog in Snowflake. Eventually, it doesn’t have to be, , tomorrow, but, , that way, if somebody asked a question in Glean around, , , I don’t know, how many active members do we have in the vehicle tech space, it could try to answer it out of the intranet data But also, say, , oh, if you’re looking for the most current information. you can go into Snowflake and look for this dashboard, ? , , it might not necessarily, . push the data into Glean, but it would direct them to the stuff that we have built. Also open to other ways that that works, but, , I do want Glean to help people find our dashboards in Snowflake. Uttam Kumaran: if we can… if you can add me to either the channel with Glean, or we can create a separate Glean channel, … or we can just add Glean to our existing channel, I would love to test out , its answers now. I haven’t used Glean before, but I would totally love to… , just see what we can do there. one is totally, it wouldn’t be too hard for us to pass Glean information about what is in the warehouse, and I don’t know… that could be pulling live from the Snowflake catalog, that could also be done in… several other ways, I would love to see, , what the ergonomics are for Glean to be able to do that. And then, , just play around with, , if you were to ask different questions, how it prioritizes what’s in the intranet versus, , what’s in here in Snowflake, and… And then also be able to show, , how many questions are… potentially being… , I don’t know what the reporting out of Glean looks , , , can we identify how many questions could be answered? , the big thing I… when we go look at a BI tool is, , I want to measure how many questions… , both how many questions are being reported on by a person that could be used by AI, then how many questions are being asked that… that just, . someone just gave up, or, , they didn’t ask, ? those are, , it’s both taking care of some of the existing questions, and increasing the TAM of questions that, , people are , oh, now that I have AI, I’m not embarrassed to ask this, or, , I can just , , just ask it, , not worry about, . how to ask it, ? , those… that’s how I’m thinking about it, ? Katherine Bayless: , , , totally. I’m not … Glean does have pretty decent reporting, I have not really played with it, I know Jay has, but we’re also… we’re bringing… we have, , a small bucket of hours with their ProServe team, and they’re going to be doing some work with us in the new year around, , adoption and enablement, that stuff, and… Truthfully, the muscle we need to start building is, , clean tends to do . Where it falls apart is, . the SharePoint dumpster fire underneath of it, ? Somebody will build, , a glean agent for a use case, and then share it with their team, and everybody’s , oh, your agent sucks, but it’s , , they just don’t have that, , SharePoint access they needed for the agent to do its job, a thing, ? . Or you’ve got 25 versions of a document, and it has idea, , which one should be correct. There are some ways, to your point, around, , weighting, and , , if… if we If we got into some of those settings, you could probably tinker with it to say, , if people ask data questions, , , lean heavily on this information. Uttam Kumaran: , that’s what I’m interested to see, , what options it has for that. And at minimum, we should… we should advertise, , use Glean. To start to ask questions, and then we can, as a data team, we want people to be asking questions, whether we can serve them. annually or not is, , up to us to figure out. But, , I do want to think about how do we make it more open for people to ask questions. Maybe they… maybe Glean is our first option now. And that way, when we evaluate BI tools, we show, , this is what’s possible in Glean. this is what’s possible in something that’s more AI-native, ? Katherine Bayless: , , . , I hated Glean for a long time. It’s gotten a little better, … . , , . Uttam Kumaran: , I would love to try it, especially, , if, , we’re seeing… we’re able to see what questions people are asking. Katherine Bayless: , I’ll see if, , I’ll see. We can definitely get you guys access. I’ll see if we can also get you, , admin, or if not, I can export. I also… I realize I… I’m Glean is on my system’s inventory. I wouldn’t have put it down as, , P1 or P0, but I’m assuming they have an API we could consume. Uttam Kumaran: I assume , . , even if I can just look in what the reporting is now, … what are we using it for? I can give you a good sense of, . How… what percentage of these questions are, , questions that the data team could have answered that maybe people are trying, ? , and I’m also interested to see, , if CTA is enabling people to build agents, maybe our job is more of, . give them the MCP, or give them the tools, versus, . , it’s great to hear that people are building on top of some of these things, because maybe more of the job is not to, , build an analysts, but, , be , , here’s the tool you can use to query Stove Lake, , you go… , build whatever you need. Katherine Bayless: , , . , people are pretty… they’re pretty resourceful, I will say. , it’s , it’s good and bad. Sometimes I get more nervous about, , oh gosh, , things, , running away too fast, , if somebody , , hacks something together thing, and then I’m also, , I don’t know, better than having a bunch of non-self-starters. But , , we’ve got quite a few people in there who have built agents, some of which are not bad. Often, they tend to have really, really complex branching logic that’s largely unnecessary. I created a, Claude project to generate my, , CES schedule, and Jay wants to migrate it into a Blean agent. , why not? . See what happens. Perfect. , those were the questions I had, , off the top of my head. many things. Uttam Kumaran: I’m gonna send… I’m gonna summarize our meeting, and I’ll send it out. Katherine Bayless: , sounds good. Oh, I was gonna say, for the 2026 scopes, do we. Will you guys do one for just, , the continued, , more. More models, , , . . Because that one, at the very least, I’ll ship through as soon as I get my hands on it. Uttam Kumaran: , that’s, how… how would you want to, , should we… did I just produce that and , , do that Q1? Do you want to bundle all these? , what’s best on your side? Katherine Bayless: Let’s not bundle Okta and Shopify, only because it isn’t 100% my decision to make as to whether or not we do that work. Uttam Kumaran: if Jay is , I do want to do it, but not now, , whatever, that thing. , perfect. Katherine Bayless: in terms of, , for the stuff that I do control the decision rights on, , , I’d really to get something in that’s just, , the 2026 scope. , we’ve talked about, , just even if it’s just Q1, and then we’ll figure out what queues 2 through 4 should look , but just to keep scaling out the data that’s landing, and the marts, and the available. Uttam Kumaran: , great. , that was my question. I didn’t know whether we wanted a bundle, perfect. I’ll… I can have that this week for you to have. And then, the other stuff, , you let me know. Katherine Bayless: , cool, cool, cool. Uttam Kumaran: Alright, thank you, everyone. Katherine Bayless: , thanks guys. Samuel Roberts: Thank you.