Meeting Title: Snowflake Governance and Fivetran Strategy Date: 2026-02-11 Meeting participants: Uttam Kumaran, Katherine Bayless
WEBVTT
1 00:01:47.460 ⇒ 00:01:49.779 Katherine Bayless: Hey, sorry, I am…
2 00:01:50.150 ⇒ 00:02:01.410 Katherine Bayless: was working with, Allie from HR to go through the resumes for the data engineer, and so we had, like, 4 left, and I was like, I’m just gonna be rude, and I’m gonna say, wait 5 minutes, and I’ll finish these real quick.
3 00:02:01.410 ⇒ 00:02:16.630 Uttam Kumaran: No, no, that’s fine. Mornings, I feel like, are slow, and then everybody wakes up, like, and asks for stuff, so I… these days, I’m trying to get ahead and, like, in the morning, be like, okay, like, these are their to-dos, like, for just the teams that I’m on, so…
4 00:02:16.680 ⇒ 00:02:17.250 Katherine Bayless: Yeah.
5 00:02:17.980 ⇒ 00:02:26.670 Katherine Bayless: That’s kind of my thing, too, is, like, the mornings, I, like, I get, like, a few precious hours of, like, people leaving me alone to get things done, and then once it’s, like, 10 o’clock, it’s like, forget it.
6 00:02:26.670 ⇒ 00:02:41.079 Uttam Kumaran: Yeah, I… I just, like… and I’m trying to do a lot more, like, personal research into a lot of different AI things, and all of that happens after 6pm, like… but the problem is, like, every… every… all the other smart people at the company are off.
7 00:02:41.580 ⇒ 00:02:54.149 Katherine Bayless: And so I’m like, I really could just use health with this one thing, and I’m, like, testing something, and I’m like, this person’s gonna wake up to, like, 100 messages, like, cracked out at, like, 9.
8 00:02:54.150 ⇒ 00:02:58.570 Uttam Kumaran: Like, I got stuck here, but it’s kind of working, and…
9 00:02:59.190 ⇒ 00:03:17.370 Katherine Bayless: Honestly, it’s like, it’s just hard to resist, though. Like, I don’t know, because I did the same thing, and actually, on Saturday, I was in Snowflake, like, playing around building that dashboard for the guy on the sales team, and I, like, happened to pop over to, like, take a look at something else, and I noticed that there was, like, another Streamlit app that had appeared in the.
10 00:03:17.370 ⇒ 00:03:18.230 Uttam Kumaran: Okay.
11 00:03:18.520 ⇒ 00:03:29.419 Katherine Bayless: Hmm, and so I… then I looked, and I saw Kai, who was active, like, logged in, like, 6 minutes ago, kind of thing. And so I slacked her, and I said, are you totally playing with Snowflake on the weekend? And she’s like, yes.
12 00:03:29.420 ⇒ 00:03:30.829 Uttam Kumaran: I was like.
13 00:03:30.830 ⇒ 00:03:31.999 Katherine Bayless: Like, have fun.
14 00:03:32.000 ⇒ 00:03:34.740 Uttam Kumaran: I know, it’s so hard, it’s just so hard, cause, like.
15 00:03:34.950 ⇒ 00:03:43.119 Uttam Kumaran: Friday comes, so these weeks, the last 3 weeks, there’s another guy on our team, Mustafa, who does a lot of AI work, and so…
16 00:03:43.660 ⇒ 00:03:56.969 Uttam Kumaran: I’m, like, I have all the ideas, they have a lot of the execution, and I’m, like, trying to bring us together. So I’m like, Fridays after work, can we just, like, spend as much time as you feel like it, just, like, pairing on stuff?
17 00:03:56.970 ⇒ 00:03:57.620 Katherine Bayless: Yeah?
18 00:03:57.620 ⇒ 00:04:04.670 Uttam Kumaran: Because I’ll be like, oh, I wonder if, like, I wanted to build a Slack assistant last week to test some things, I’m like, we’re working on… and I’m like.
19 00:04:04.820 ⇒ 00:04:13.379 Uttam Kumaran: I just… if you just be next to me, and then I’ll absorb, and then we’ll also work on, like, end-to-end testing for me that I can, like, start to ship things, but…
20 00:04:13.910 ⇒ 00:04:18.090 Uttam Kumaran: Yeah, that’s, like, the best part of my week, I feel like, for the most part.
21 00:04:18.290 ⇒ 00:04:20.260 Uttam Kumaran: That, like, 3 hours where I’m like.
22 00:04:20.380 ⇒ 00:04:25.539 Uttam Kumaran: just, like, trying things at Cursor. I have, like, 5 different, like, codecs things going.
23 00:04:26.110 ⇒ 00:04:27.340 Uttam Kumaran: It’s fun.
24 00:04:27.340 ⇒ 00:04:30.519 Katherine Bayless: Yeah, yeah, yeah, it’s like, what we got into this for, it feels.
25 00:04:30.520 ⇒ 00:04:38.999 Uttam Kumaran: Yeah, yeah, that’s what I was like, yeah, this is, like, what I… I feel like this is the part of the job that’s fun these days, is, like, is a lot of the AI work, so…
26 00:04:39.000 ⇒ 00:04:41.230 Katherine Bayless: Right? The rest is all politicking.
27 00:04:41.230 ⇒ 00:04:42.520 Uttam Kumaran: Yeah.
28 00:04:43.780 ⇒ 00:04:58.180 Katherine Bayless: Yeah, okay, well, so, speaking of data and fun parts, so, for the polyatomic thing, so, like, I definitely… I still think it’s a great tool, but I’m like, okay, I know we’ve had these conversations before around, like, I’m trying to get people sort of familiar with how much.
29 00:04:58.180 ⇒ 00:04:58.650 Uttam Kumaran: Yeah.
30 00:04:58.650 ⇒ 00:05:06.900 Katherine Bayless: what’s it gonna cost? And this one’s, like, a big sticker up front, and so my brain goes into, like, okay, I need to pitch this, like, how do I tell this story, right?
31 00:05:07.450 ⇒ 00:05:11.449 Katherine Bayless: Then I started kind of thinking about, and this is gonna be just, like, a brain dump, and then…
32 00:05:11.450 ⇒ 00:05:12.030 Uttam Kumaran: Yeah, yeah, yeah.
33 00:05:12.030 ⇒ 00:05:26.050 Katherine Bayless: I started kind of thinking about, okay, also getting ready to answer questions, like, if they’re like, okay, $50,000, what do you get? And I’m like, well, they’re gonna build it for us. And it’s like, okay, so we’re actually paying a lot of money and hoping that they will indeed.
34 00:05:26.050 ⇒ 00:05:33.830 Uttam Kumaran: Yeah. Which I do trust, but it’s like, that is totally a question that’ll come up, I think, as part of this.
35 00:05:33.830 ⇒ 00:05:45.080 Katherine Bayless: Then I was thinking about the monthly active rows thing, and then I got kind of down the, like, rabbit hole of, like, okay, if we think about where this is, like, now versus where it’s going.
36 00:05:45.890 ⇒ 00:06:04.860 Katherine Bayless: I mean, yes, right now it’d be nice to land some data in Snowflake and have a little bit more stuff to play with, and to get the Marketing Cloud Connector in place, but, like, as we get later into the year, we’re gonna need to be operating, you know, kind of almost more like a true, like, middleware-y kind of vibe, like, if we’re gonna actually try and integrate some of these CES tech platforms.
37 00:06:05.110 ⇒ 00:06:10.839 Katherine Bayless: Which would require a couple additional custom connectors to be built,
38 00:06:11.060 ⇒ 00:06:22.960 Katherine Bayless: And so, yeah, so then I was like, okay, well, okay, Catherine, are you trying to build, like, you know, end service bus out of an I-Pass? Because that’s not really how this is supposed to work. And then, yeah, so I…
39 00:06:23.850 ⇒ 00:06:29.449 Katherine Bayless: I don’t know. And part of me also… the other thought I had was, really? The thing that we do…
40 00:06:29.950 ⇒ 00:06:37.060 Katherine Bayless: the piece that I do really want to solve that’s on the board is that Marketing Cloud connector. Yeah, okay. That, I think.
41 00:06:37.180 ⇒ 00:06:40.379 Katherine Bayless: I wondered if, they might consider, like.
42 00:06:40.860 ⇒ 00:06:49.360 Katherine Bayless: can we buy one connector and, like, commit to a smaller amount of usage up front? Then we can kind of, like, get a feel for, like, working with them.
43 00:06:49.360 ⇒ 00:06:49.870 Uttam Kumaran: Yeah.
44 00:06:49.870 ⇒ 00:07:06.440 Katherine Bayless: have that one built out, and then I think by the end of the year, we would probably scale up to where he was looking at initially, but it gives me a little bit more time to socialize the value of the tool, because not for nothing, but it’ll wind up being more expensive than probably Snowflake and AWS, right? Which I think is probably.
45 00:07:06.440 ⇒ 00:07:09.559 Uttam Kumaran: So this is usually… yeah, so this is…
46 00:07:10.350 ⇒ 00:07:12.749 Uttam Kumaran: Yeah, so I think I have two thoughts. One is, like.
47 00:07:12.750 ⇒ 00:07:13.300 Katherine Bayless: Yeah, yeah, go.
48 00:07:13.300 ⇒ 00:07:14.810 Uttam Kumaran: We recommend them
49 00:07:15.650 ⇒ 00:07:20.979 Uttam Kumaran: But I would say we… the two options that we always put up is, like, Fivetran or that.
50 00:07:21.220 ⇒ 00:07:22.210 Katherine Bayless: Like…
51 00:07:22.210 ⇒ 00:07:25.300 Uttam Kumaran: Where I’ve liked them as a data person is, like.
52 00:07:25.720 ⇒ 00:07:41.889 Uttam Kumaran: I’m now really confident that they can build the… the nth connector. Like, we… and we haven’t had any issues where they’ve built something, it hasn’t worked, and then they’re, like, we’re, like, in a queue. Usually, if it doesn’t work, within, like, the next day… so, like, I have a lot of… I gotta get that’s hard to,
53 00:07:42.740 ⇒ 00:07:49.000 Uttam Kumaran: that’s hard to, like, deliver that faith to other people, right? So that’s one piece. Second piece is that,
54 00:07:49.390 ⇒ 00:07:52.650 Uttam Kumaran: confidently, they will be cheaper than Fivetran. Yeah.
55 00:07:52.790 ⇒ 00:07:54.140 Katherine Bayless: Yeah, yeah.
56 00:07:54.140 ⇒ 00:08:00.950 Uttam Kumaran: Yeah, so… So that’s one piece. Second piece is Fivetran will most likely not build at the same pace.
57 00:08:01.120 ⇒ 00:08:01.630 Katherine Bayless: Right.
58 00:08:01.630 ⇒ 00:08:06.619 Uttam Kumaran: But there’s… there’s a couple options here we could do. One is…
59 00:08:06.850 ⇒ 00:08:14.349 Uttam Kumaran: if… if we feel like speed is paramount, we should just go ahead and hook up the Fivetran Snowflake Marketing Cloud.
60 00:08:14.650 ⇒ 00:08:17.609 Uttam Kumaran: Take advantage of their 14-day free trial.
61 00:08:17.890 ⇒ 00:08:19.550 Uttam Kumaran: backfill everything.
62 00:08:20.040 ⇒ 00:08:28.380 Uttam Kumaran: And you could sort of just decide after that point. Like, so if I’m thinking about, like, okay, how do we hack this a little bit? One is Fivetrade is a pretty generous free trial.
63 00:08:28.600 ⇒ 00:08:31.669 Uttam Kumaran: We can hook up the connectors that we need, run the backfill.
64 00:08:31.840 ⇒ 00:08:35.600 Uttam Kumaran: And then, get a sense of the volume, get a sense of pricing from them.
65 00:08:35.890 ⇒ 00:08:42.020 Uttam Kumaran: you can have a two-dual ETL strategy. Like, it’s not, like, the worst thing,
66 00:08:42.590 ⇒ 00:08:47.930 Uttam Kumaran: And we can always move to, like, polyatomic later.
67 00:08:48.300 ⇒ 00:08:53.659 Uttam Kumaran: So… That is, like, my first instinct, is, like, maybe we should just, like.
68 00:08:53.940 ⇒ 00:08:57.509 Uttam Kumaran: hook that up, get all the Salesforce stuff landing.
69 00:08:57.840 ⇒ 00:09:03.510 Uttam Kumaran: Get a sense… they’ll… they’ll give you a clear understanding of, like, how much… how many rows are coming in.
70 00:09:03.810 ⇒ 00:09:04.230 Katherine Bayless: Yeah.
71 00:09:04.230 ⇒ 00:09:10.209 Uttam Kumaran: We have a… we have a… we have a relationship with them, too, so I can see what we can do on discounts.
72 00:09:10.850 ⇒ 00:09:18.710 Uttam Kumaran: Really, the risk there is, like, again, they will be expensive in the long term, and they may not build the nth connector.
73 00:09:19.080 ⇒ 00:09:19.900 Katherine Bayless: Right.
74 00:09:19.900 ⇒ 00:09:24.930 Uttam Kumaran: But, like, they… yeah, so that’s, like, that was my… kind of my gut instinct in that, like…
75 00:09:25.340 ⇒ 00:09:30.339 Uttam Kumaran: that may be a good path, so that you don’t have to spend… I would rather, if you’re gonna get budget.
76 00:09:30.600 ⇒ 00:09:33.239 Uttam Kumaran: to focus on Snowflake.
77 00:09:33.430 ⇒ 00:09:39.880 Uttam Kumaran: Because that’s, like, the most business-facing stuff right now, versus if you go talk to people about ETL.
78 00:09:40.430 ⇒ 00:09:41.080 Katherine Bayless: Right.
79 00:09:41.080 ⇒ 00:09:48.590 Uttam Kumaran: like, they might glaze over, and then… and then they’re like, what is… why? We just got you… we just got you data stuff. You’re like, no, no, that was ETL.
80 00:09:48.790 ⇒ 00:09:49.300 Katherine Bayless: Great.
81 00:09:49.960 ⇒ 00:09:53.710 Uttam Kumaran: So, from, like, from the politics standpoint, like, that’s my…
82 00:09:54.030 ⇒ 00:09:57.919 Uttam Kumaran: that would be probably, like, my perspective, is, like, I hear you, like, you don’t want to use
83 00:09:58.330 ⇒ 00:10:02.960 Uttam Kumaran: a couple of the cards that you have right now on ETL. It’s, like, kind of a waste.
84 00:10:03.110 ⇒ 00:10:04.740 Uttam Kumaran: Of a good card.
85 00:10:05.080 ⇒ 00:10:12.099 Uttam Kumaran: But it’s important, so, like, there are some ways… there are some ways, I feel like, around this, but yeah, kind of curious what you think about that.
86 00:10:12.470 ⇒ 00:10:36.000 Katherine Bayless: Yeah, I mean, I think, yeah, you’re totally spot on. And it’s funny, your Notion doc had that kind of nod to it, too, where it was like, you know, don’t… don’t go all in on the enterprise tool when you don’t really have enterprise needs yet. I mean, the, you know, reality is, right now, for the foreseeable future, Salesforce Marketing Cloud, we can absolutely integrate. The rest of it is gonna be flat files for a while, just because, like, there’s, you know, the whole politics and socializing and all the things. Yeah.
87 00:10:36.000 ⇒ 00:10:41.539 Katherine Bayless: The Petran idea is really compelling. The reason being.
88 00:10:41.540 ⇒ 00:10:49.120 Katherine Bayless: I have a suspicion that the way we have been using Marketing Cloud is gonna be not the easiest
89 00:10:49.120 ⇒ 00:10:56.160 Katherine Bayless: to, like, get the functionality we need. I could be totally wrong, I’d be delighted to be wrong. But I know that, like, Marketing Cloud…
90 00:10:56.280 ⇒ 00:11:10.300 Katherine Bayless: I mean, obviously, Salesforce, you know, it knits nicely together with the actual CRM, but everything else, you know, good luck. But we’re using, like, data extensions versus, like, lists or groups.
91 00:11:10.300 ⇒ 00:11:21.419 Katherine Bayless: And I think the way the platform is going, everything is supposedly kind of trending that direction of data extensions, but I think the API has historically had more support for lists than groups.
92 00:11:21.660 ⇒ 00:11:31.199 Katherine Bayless: One nice thing is that, at the end of the day, the marketing team, all they care is that they see something they can click on that says, you know, active members, right?
93 00:11:31.200 ⇒ 00:11:32.069 Uttam Kumaran: Yeah, yeah.
94 00:11:32.070 ⇒ 00:11:42.259 Katherine Bayless: an email to it. They don’t really care what it is on the back end. So, I think Fivetrain could actually give us a chance to, like, just… I mean, honestly, might even help us develop better requirements for Polytomic.
95 00:11:42.260 ⇒ 00:11:43.060 Uttam Kumaran: Yeah, yeah.
96 00:11:43.420 ⇒ 00:11:50.559 Katherine Bayless: The other thing that would potentially help both Fivetran, sort of, like, MVP and with Polytomic is
97 00:11:51.420 ⇒ 00:11:58.079 Katherine Bayless: if I think, again, about, like, integrating versus just, like, you know, greedy data girl,
98 00:11:58.080 ⇒ 00:12:13.270 Katherine Bayless: we can get the data out of Marketing Cloud via that FTP, so we can… I mean, there are jobs actually set up to do it now that, like, schedule certain, you know, exports, like unsubs and opens and clicks and stuff, and they just do, like, a batched file out to the FTP,
99 00:12:13.270 ⇒ 00:12:27.949 Katherine Bayless: So we could just grab it from there, and then we at least wouldn’t need to use the monthly active rows allocation for, like, collecting that data. It’s probably more important that we’re sending contacts across, like.
100 00:12:28.030 ⇒ 00:12:29.629 Katherine Bayless: More frequently and consistently.
101 00:12:29.630 ⇒ 00:12:34.120 Uttam Kumaran: Yeah, and let me even, like, let me see if I have…
102 00:12:35.470 ⇒ 00:12:40.200 Uttam Kumaran: like, an example of, like, a 5 train instance I can show you.
103 00:12:40.200 ⇒ 00:12:40.690 Katherine Bayless: Yeah, yeah.
104 00:12:48.430 ⇒ 00:12:50.890 Katherine Bayless: Slack JBAC.
105 00:13:08.940 ⇒ 00:13:12.000 Uttam Kumaran: Yeah, so this is, like, we will get,
106 00:13:12.120 ⇒ 00:13:21.160 Uttam Kumaran: this is for, like, a very, very, like, old client. It’s no longer using Fivetran, but we’re… this is, like, kind of what you’ll see. So you’ll see, like.
107 00:13:21.350 ⇒ 00:13:25.690 Uttam Kumaran: For each connection, the number of rows, and like, for example, if we looked at, like.
108 00:13:25.820 ⇒ 00:13:30.509 Uttam Kumaran: Google Ads. You’ll get a breakdown of the MARB per table.
109 00:13:30.510 ⇒ 00:13:32.289 Katherine Bayless: And we can easily be like.
110 00:13:32.520 ⇒ 00:13:38.509 Uttam Kumaran: We’re not using the… minute-by-minute, like, breakdown table. We should turn that off.
111 00:13:38.640 ⇒ 00:13:44.900 Uttam Kumaran: So, the nice thing about Fivetrain is, like, we can go… Kick off a new connector.
112 00:13:45.150 ⇒ 00:13:49.809 Uttam Kumaran: We both will get a 30-day trial on the platform, and a 14-day
113 00:13:50.220 ⇒ 00:13:52.749 Uttam Kumaran: Like, no fee on any new connector.
114 00:13:53.130 ⇒ 00:13:57.090 Uttam Kumaran: And so, maybe it’s worth us just kicking those off for a couple of them.
115 00:13:57.560 ⇒ 00:13:58.010 Katherine Bayless: Bye.
116 00:13:58.400 ⇒ 00:14:03.850 Uttam Kumaran: You know, basically as many as we can just kick off that Fivetran supports.
117 00:14:04.170 ⇒ 00:14:06.009 Uttam Kumaran: That’s what I would just suggest.
118 00:14:06.260 ⇒ 00:14:08.700 Uttam Kumaran: And I was just, like, kind of looking at, like.
119 00:14:08.910 ⇒ 00:14:13.539 Uttam Kumaran: Kind of looking at what other integrations are in there.
120 00:14:14.120 ⇒ 00:14:26.299 Katherine Bayless: Yeah, actually, you’re right, I meant to explicitly say that, too. For benefit of the robot, who will type up these notes later. Like, I do actually… I… you make a really good point around, like, use it, dump in all the… Yes, event, yes.
121 00:14:26.480 ⇒ 00:14:30.429 Katherine Bayless: Dump in all the data we want, and then even if it’s not on an ongoing basis.
122 00:14:30.430 ⇒ 00:14:34.369 Uttam Kumaran: Well, that’s also… I think it’s also fair for some clients, I’m like.
123 00:14:36.760 ⇒ 00:14:49.349 Uttam Kumaran: like, is the cost savings worth fighting for, or is the benefit of having this now? Like, so that’s the thing, is… I recommend… we would just recommend both, but I… I’m sort of, like, I always want to have options.
124 00:14:49.430 ⇒ 00:14:59.469 Uttam Kumaran: It’s not that I don’t… I have beef with these guys, they’re just support is not as good. So, for some clients where they want, like, a true, like, they want the vendor to be on the phone.
125 00:14:59.550 ⇒ 00:15:02.510 Katherine Bayless: And I know there’s, like, some weird connectors.
126 00:15:02.510 ⇒ 00:15:08.550 Uttam Kumaran: We usually push them that way, because they just have a great, like, relationship, typically. But…
127 00:15:09.450 ⇒ 00:15:16.589 Uttam Kumaran: Like, it may be easiest for us to just turn this on, we land it, we deliver wins, and then…
128 00:15:16.960 ⇒ 00:15:24.850 Uttam Kumaran: That allows you to, like, pocket another couple cards to then see, like, whether it’s worth getting the budget, whether we can or whatever.
129 00:15:25.000 ⇒ 00:15:33.240 Katherine Bayless: Yeah. So… What is, is Fivetran sort of the same concept of, like, platform fee, and then.
130 00:15:33.240 ⇒ 00:15:38.960 Uttam Kumaran: Yeah. So they have, like, a kind of a… I mean, they have this, like, baby free tier.
131 00:15:38.960 ⇒ 00:15:39.880 Katherine Bayless: For us.
132 00:15:39.880 ⇒ 00:15:47.090 Uttam Kumaran: Yeah, there is, like, a standard… there is, like, a standard tier, and then, there basically is, like,
133 00:15:47.640 ⇒ 00:15:50.069 Uttam Kumaran: Some type of, like, connector-based…
134 00:15:50.070 ⇒ 00:15:51.170 Katherine Bayless: Price.
135 00:15:52.630 ⇒ 00:15:58.499 Uttam Kumaran: They have changed this… they basically change this, like, every 6 months, so it’s… it all kind of always changes.
136 00:15:58.510 ⇒ 00:15:59.839 Katherine Bayless: But.
137 00:16:00.350 ⇒ 00:16:01.650 Uttam Kumaran: Yeah, like…
138 00:16:02.350 ⇒ 00:16:10.649 Uttam Kumaran: Basically, what we’ll do is we’ll just click everything, run it, and then after, like, 14 days of every connector, it’ll give us an estimate of, like.
139 00:16:11.040 ⇒ 00:16:13.060 Uttam Kumaran: What the daily run rate is.
140 00:16:13.340 ⇒ 00:16:13.920 Katherine Bayless: Yeah.
141 00:16:13.920 ⇒ 00:16:16.849 Uttam Kumaran: And then we’ll… it’ll give us, like, the cost estimate.
142 00:16:17.010 ⇒ 00:16:23.539 Uttam Kumaran: For that. Like, otherwise, the product is… is totally fine. Yeah.
143 00:16:23.990 ⇒ 00:16:34.490 Katherine Bayless: Yeah, and I mean, truthfully, like, I mean, I know it sounds silly, it’s, you know, penny-wise and pound foolish by definition, but I really think, like, if it is possible.
144 00:16:34.510 ⇒ 00:16:37.699 Uttam Kumaran: to do a pay-as-you-go thing. Yeah.
145 00:16:37.700 ⇒ 00:16:56.340 Katherine Bayless: point. It’s like, then I can build the narrative around, like, these are the wins we are delivering, this is, you know, the cost increasing as we continue to crush tech debt. And then, I mean, actually, to a certain extent, it also does give us the, like, you know, if we really do start to boil that frog, we’re like, well, Polytomic will be cheaper, we can migrate out to them, kind of a thing.
146 00:16:56.340 ⇒ 00:16:57.000 Uttam Kumaran: Yeah.
147 00:16:58.500 ⇒ 00:17:01.899 Uttam Kumaran: And, like, when you go to Polyatomic, you’ll have all the numbers.
148 00:17:01.900 ⇒ 00:17:03.020 Katherine Bayless: And so you can just.
149 00:17:03.020 ⇒ 00:17:05.779 Uttam Kumaran: I basically hand them the numbers and be like, what is it gonna be?
150 00:17:06.560 ⇒ 00:17:15.979 Uttam Kumaran: Yeah. And then you can… you kind of have leverage, you can kind of just basically be like, we’ll… you can just kind of put some leverage on them, like, we will switch our business over if you can give us X price.
151 00:17:17.869 ⇒ 00:17:18.720 Katherine Bayless: So…
152 00:17:18.960 ⇒ 00:17:22.929 Katherine Bayless: I mean, we have… we have a few other… we have a few clients that are in the same boat.
153 00:17:23.369 ⇒ 00:17:27.030 Uttam Kumaran: And it just sort of depends on…
154 00:17:28.260 ⇒ 00:17:30.870 Uttam Kumaran: Like, kind of the speed, and then…
155 00:17:31.430 ⇒ 00:17:44.400 Uttam Kumaran: it just sort of depends on, look, like, everything, there’s cost risk, but also, like, revenue risk. And so, in this situation, it’s clear that this is more of, like, this can actually get us more visibility and get us some wins faster.
156 00:17:44.490 ⇒ 00:17:59.800 Uttam Kumaran: I don’t think the tech debt is, like, is as… is nearly as bad as, like, if we were to go with, like, a different BI tool or go with a different warehouse, that is, like, a really big migration, versus switching these is not that bad. It’s getting easier, basically, over time, so…
157 00:18:00.240 ⇒ 00:18:09.189 Katherine Bayless: Right, I mean, it makes sense, because the data is going to be consistent, no matter which tool you’re really using to land it, and so once you build out the pipelines against what the endpoints are like.
158 00:18:09.190 ⇒ 00:18:18.780 Uttam Kumaran: Yeah, and because we model… because we set up dbt in, like, this raw staging way, we just swap out the raw, like, and we can just map the columns.
159 00:18:18.950 ⇒ 00:18:19.390 Katherine Bayless: Yep.
160 00:18:19.390 ⇒ 00:18:23.009 Uttam Kumaran: So there’s not, like, upstream of that, there’s not much risk.
161 00:18:23.390 ⇒ 00:18:24.880 Katherine Bayless: Yeah, exactly.
162 00:18:25.560 ⇒ 00:18:35.119 Katherine Bayless: Yeah, okay, well, let me… let me take some time and kind of dig around in Fivetran, and see… because I… I am curious about this Marketing Cloud kind of config, and…
163 00:18:35.120 ⇒ 00:18:43.099 Uttam Kumaran: Yeah, so I was just looking… yeah, if you just type in, fivetran, Marketing Cloud… oh yeah.
164 00:18:45.910 ⇒ 00:18:53.950 Uttam Kumaran: They have, like, kind of, like, all the… both the setup instructions, and if you go to this, you’ll see, like, the whole ERD.
165 00:18:54.390 ⇒ 00:18:56.379 Katherine Bayless: Yeah, oh, nice, okay, okay.
166 00:18:57.100 ⇒ 00:19:00.309 Uttam Kumaran: They’ll have, like, some, like, gotchas, like, notes here.
167 00:19:00.310 ⇒ 00:19:05.329 Katherine Bayless: Oh yeah, look, it even says data extensions. You can sync data extensions from your…
168 00:19:07.170 ⇒ 00:19:09.669 Katherine Bayless: Oh, interesting, I see him go in the other way, but yeah, okay.
169 00:19:10.420 ⇒ 00:19:13.280 Uttam Kumaran: And then they have… they also have, like, you can send…
170 00:19:13.520 ⇒ 00:19:16.120 Uttam Kumaran: They also have the reverse ETL option, too.
171 00:19:16.320 ⇒ 00:19:20.149 Katherine Bayless: yeah, I mean, I think, yeah, it…
172 00:19:20.750 ⇒ 00:19:29.340 Katherine Bayless: I mean, that would kind of be one of the, trade-offs with the cost, right? It’s like, okay, if 5TRAN is a little bit more spendy, then maybe, yeah, for the taking the data.
173 00:19:29.340 ⇒ 00:19:29.960 Uttam Kumaran: Yes.
174 00:19:29.960 ⇒ 00:19:42.969 Katherine Bayless: Marketing Cloud, we would just do the FTP and feel like that’s fine. We don’t need that data to be super high fidelity, but pushing the data into Marketing Cloud, that’s where the value, the win is, so yeah. Yeah. Okay, okay.
175 00:19:43.400 ⇒ 00:19:47.130 Katherine Bayless: This is good. Let me dig around with this a bit,
176 00:19:49.170 ⇒ 00:19:53.249 Katherine Bayless: Yeah, but I think this is probably the right way to pitch it.
177 00:19:53.520 ⇒ 00:19:54.140 Uttam Kumaran: Okay.
178 00:19:54.560 ⇒ 00:19:55.200 Katherine Bayless: Okay.
179 00:19:55.380 ⇒ 00:19:56.290 Katherine Bayless: I like it.
180 00:19:56.490 ⇒ 00:20:03.719 Katherine Bayless: Oh, and then I need to take a look at the Shopify thing. I still haven’t actually reviewed that, but, need to take a look at that, get that in the way.
181 00:20:03.720 ⇒ 00:20:10.170 Uttam Kumaran: The other… the other piece is, like, we wrote up a bunch of docs yesterday on, how we’re gonna do…
182 00:20:10.460 ⇒ 00:20:12.979 Uttam Kumaran: like, governance in Snowflake?
183 00:20:12.980 ⇒ 00:20:13.510 Katherine Bayless: Hmm.
184 00:20:13.510 ⇒ 00:20:16.260 Uttam Kumaran: So, I have, like, a bit of a larger…
185 00:20:16.490 ⇒ 00:20:19.360 Uttam Kumaran: doc on that that I can,
186 00:20:19.900 ⇒ 00:20:22.930 Uttam Kumaran: that I can share. Like, kind of what we’ve done is both do, like.
187 00:20:23.220 ⇒ 00:20:26.240 Uttam Kumaran: I mean, maybe I can just… I can just share a bit of it right now.
188 00:20:26.900 ⇒ 00:20:33.090 Katherine Bayless: Actually, I would love, like, a little debrief of the stuff you learned talking to Snowflake about the AI stuff.
189 00:20:33.090 ⇒ 00:20:35.760 Uttam Kumaran: Oh, yeah. Well, I mean, it, like…
190 00:20:35.980 ⇒ 00:20:44.249 Uttam Kumaran: they’re always… anytime I call them, it’s, like, slightly disappointing, like, they’re… they’re just very slow, so, like, I had to spend the first part of the call basically being, like.
191 00:20:44.560 ⇒ 00:20:48.260 Uttam Kumaran: I’ve used Snowflake for, like, almost 10 years, like, get to the… please get to the point.
192 00:20:48.260 ⇒ 00:20:48.830 Katherine Bayless: Hmm.
193 00:20:48.830 ⇒ 00:20:50.569 Uttam Kumaran: And they’re like, well, what is your, like.
194 00:20:50.730 ⇒ 00:20:54.059 Uttam Kumaran: who’s the stakeholder? Like, who is the stakeholder? And I’m like.
195 00:20:54.660 ⇒ 00:20:58.500 Uttam Kumaran: Dude, they’re like, and so what is your involvement? I’m like, dude, just like…
196 00:20:59.980 ⇒ 00:21:07.500 Uttam Kumaran: I skipped the, like, what are my orders today, like, I’m interested in semantic understanding, I’m interested in, like.
197 00:21:07.610 ⇒ 00:21:17.249 Uttam Kumaran: integrations to pull data out of, like, get to the meat of it, so they sent me a bunch of things. I mean, basically, one is, like, they have a lot of ways to add semantic context into.
198 00:21:17.250 ⇒ 00:21:17.710 Katherine Bayless: Hmm.
199 00:21:17.710 ⇒ 00:21:23.930 Uttam Kumaran: Snowflake, and they actually sent me… Some docs to read.
200 00:21:24.240 ⇒ 00:21:28.140 Uttam Kumaran: That… let me go find, like, an example.
201 00:21:28.630 ⇒ 00:21:34.119 Uttam Kumaran: So, one is, like, there’s…
202 00:21:34.750 ⇒ 00:21:36.290 Uttam Kumaran: They kind of sent me this…
203 00:21:36.640 ⇒ 00:21:46.720 Uttam Kumaran: case study a little bit about using, like, the semantic view autopilot, which basically, like, is able to use your queries and other assets to, like.
204 00:21:47.360 ⇒ 00:21:49.109 Katherine Bayless: I mean, like…
205 00:21:49.620 ⇒ 00:21:59.369 Uttam Kumaran: semantic understanding of things, and so there is a, like, guide on basically using what’s called this, like, semantic views for Cortex analysts.
206 00:21:59.370 ⇒ 00:22:00.560 Katherine Bayless: You can give it…
207 00:22:00.600 ⇒ 00:22:06.360 Uttam Kumaran: a bunch of… You know, documents, and then it… it, like.
208 00:22:07.030 ⇒ 00:22:14.439 Uttam Kumaran: We’ll basically run through and, like, create a lot of semantic views, and then gives you the ability to, like, set up, like, metrics, relationships.
209 00:22:15.160 ⇒ 00:22:15.800 Katherine Bayless: Description.
210 00:22:15.800 ⇒ 00:22:24.849 Uttam Kumaran: trends, and so… My thinking here is, like, twofold. One is, like, in the repo itself.
211 00:22:25.570 ⇒ 00:22:31.420 Uttam Kumaran: I’m thinking we maintain… This, probably, as, like…
212 00:22:32.350 ⇒ 00:22:40.520 Uttam Kumaran: I mean, I’m kind of, like, in between two places. One is, I’m like, okay, should we maintain, like, in dbt, you can have, like, YAML on metrics and dimensions.
213 00:22:40.770 ⇒ 00:22:44.239 Uttam Kumaran: Yeah. So one is, I’m like, should we write it there, and then potentially…
214 00:22:44.350 ⇒ 00:22:47.540 Uttam Kumaran: just… should we store it there, and then write it to Snowflake?
215 00:22:47.660 ⇒ 00:22:50.709 Uttam Kumaran: So that’s… the flake is really not the source of truth, like…
216 00:22:50.730 ⇒ 00:22:52.820 Katherine Bayless: Right. The repo is a source of truth.
217 00:22:53.150 ⇒ 00:22:56.509 Uttam Kumaran: And… and have, like, basically write a couple scripts that, like.
218 00:22:56.770 ⇒ 00:23:03.299 Uttam Kumaran: when you ship a new PR, for example, and you add more metrics, we do two things. One, you’re not allowed… like, we basically…
219 00:23:03.690 ⇒ 00:23:09.540 Uttam Kumaran: block the PR until, like, appropriate descriptions and things are added. Second is,
220 00:23:10.440 ⇒ 00:23:13.200 Uttam Kumaran: like, when a PR goes through.
221 00:23:13.820 ⇒ 00:23:16.909 Uttam Kumaran: maybe it’s on a weekly basis, or it’s on PR,
222 00:23:16.950 ⇒ 00:23:34.069 Uttam Kumaran: something goes into Snowflake and, like, writes the semantic stuff. The trouble with any type of documentation is that you need, like, a sort of a two-way thing. You need both the ease to write it, and then you need some type of mechanism to keep it updated. So the writing piece is where I’m like.
223 00:23:34.470 ⇒ 00:23:38.540 Uttam Kumaran: someone go- like, I don’t know whether it’s gonna be best for someone to, like.
224 00:23:38.830 ⇒ 00:23:43.939 Uttam Kumaran: I don’t know whether everybody in the company is gonna… or developers, are gonna be able to…
225 00:23:44.570 ⇒ 00:23:52.530 Uttam Kumaran: use the Cortex CLI to do this, so I’m like, maybe we configure the repo in a way where, as long as people are good at adhering to, like.
226 00:23:52.760 ⇒ 00:23:55.720 Uttam Kumaran: YAML definitions for new tables.
227 00:23:56.550 ⇒ 00:23:58.740 Uttam Kumaran: then that will all just end up in Snowflake.
228 00:23:59.290 ⇒ 00:24:01.300 Uttam Kumaran: And then also a mechanism for, like.
229 00:24:01.570 ⇒ 00:24:12.469 Uttam Kumaran: hey, I have, like, a PDF that I want to get put in context, or I have a bunch of transcripts, like, maybe we, again, allow just a space in… in the repo for people to upload those.
230 00:24:12.810 ⇒ 00:24:21.310 Uttam Kumaran: and then we facilitate how that gets into Snowflake. I mean, Snowflake will always push you to just, like, start and do everything in there, but…
231 00:24:22.510 ⇒ 00:24:28.419 Uttam Kumaran: I don’t know whether, like, what’s the benefit, because if, yeah, if we can do all this through CLI, then, you know, yeah.
232 00:24:28.420 ⇒ 00:24:44.489 Katherine Bayless: No, totally. I think, yeah, 100% agree, like, build it out in the YAML in the repo, and then let Snowflake benefit from it, but that also keeps us kind of a little bit more tool agnostic. Not that I have any intention of migrating off of Snowflake anytime soon, but it’s at least, like, it’s still…
233 00:24:44.490 ⇒ 00:24:45.080 Uttam Kumaran: Yeah.
234 00:24:45.080 ⇒ 00:24:53.210 Katherine Bayless: our code versus, yeah. The other thing I’m thinking, too, I love
235 00:24:53.630 ⇒ 00:25:01.190 Katherine Bayless: I love the idea of formalizing the metrics that way, and actually I can share some of the stuff that I’ve already kind of started, like, I have a…
236 00:25:01.610 ⇒ 00:25:05.840 Katherine Bayless: somewhere I parked, like, a table of, like, you know, kind of, like, metric name, other metric.
237 00:25:06.180 ⇒ 00:25:31.080 Katherine Bayless: Because then people, you know, call them different things. And then, like, the business definition, and then, like, the actual set of, like, SQL filters. And actually, similarly, I was just kind of playing with it in the UI, so when I gave, Dave Hennessy the Snowflake access the other day, and he was talking to Cortex Code, and, like, it was pulling back more data than it should have, because it didn’t know that, like, we only want verified, and, like, some of these other things, so I gave him, like, you know.
238 00:25:31.080 ⇒ 00:25:32.930 Katherine Bayless: Copy-paste this at the beginning of your prompt.
239 00:25:32.930 ⇒ 00:25:42.589 Katherine Bayless: for the moment, but then I put it in the table description, and then for the relevant columns that were in those critical filters, I put, like, metadata in the columns for it, too.
240 00:25:43.040 ⇒ 00:25:49.740 Katherine Bayless: the first pass after that, where I asked it a question, it did not do anything with it, but when I said, like, you know, check out the metadata, and then.
241 00:25:49.740 ⇒ 00:25:50.480 Uttam Kumaran: Oh, okay.
242 00:25:50.660 ⇒ 00:25:56.610 Katherine Bayless: it did actually respect all of it pretty well, and so, I think, yeah.
243 00:25:56.800 ⇒ 00:26:01.229 Katherine Bayless: I also think this is a great place to, like, lean on Kai, and I mean, she’s…
244 00:26:01.230 ⇒ 00:26:01.660 Uttam Kumaran: Yeah.
245 00:26:01.660 ⇒ 00:26:17.140 Katherine Bayless: willing to… I mean, she’s pretty good at SQL, she’s definitely, you know, quick learner, and, like, the business analyst piece is, like, totally, you know, go figure out what this definition is. And so I think, yeah, if there’s some sort of, kind of, informal pipeline for the moment where it’s, like.
246 00:26:17.140 ⇒ 00:26:26.479 Uttam Kumaran: Yeah, so maybe it’s more of, like, enabling her to do the net new docs, and then update, and just figuring out, like, what’s the, like, what she would need to do that.
247 00:26:26.740 ⇒ 00:26:28.810 Katherine Bayless: Yeah, yeah, yeah.
248 00:26:28.810 ⇒ 00:26:29.450 Uttam Kumaran: Okay.
249 00:26:29.670 ⇒ 00:26:30.390 Katherine Bayless: Yeah.
250 00:26:30.790 ⇒ 00:26:32.439 Katherine Bayless: I think that would be really cool.
251 00:26:33.860 ⇒ 00:26:36.700 Katherine Bayless: The other thing we could probably leverage is…
252 00:26:36.960 ⇒ 00:26:45.790 Katherine Bayless: for, like, you know, to kind of get a head start, is I have all of the old, SQL code and process documentation and, you know, blah blah blah. Oh, great
253 00:26:45.800 ⇒ 00:26:56.980 Katherine Bayless: from the old data team, and so, like, even though I would not want to have AI think that this is how we want our SQL code to look, it, you know, it does sort of give you that window into what the definitions were by.
254 00:26:56.980 ⇒ 00:27:05.629 Uttam Kumaran: Yeah, we should just put, like, an archive or folder in the repo, and you can dump it there, and then I’m gonna basically bootstrap the first version of the docs.
255 00:27:05.720 ⇒ 00:27:09.740 Katherine Bayless: Yeah. Right? Using, like, partly our… some of our transcripts.
256 00:27:09.910 ⇒ 00:27:16.270 Uttam Kumaran: It’s gonna… it’ll infer some things from the structure. It, of course, will, like, use web and, like, kind of understand the type of business.
257 00:27:16.400 ⇒ 00:27:19.489 Uttam Kumaran: So, like, I will basically… because what I want to do is, like.
258 00:27:19.830 ⇒ 00:27:26.840 Uttam Kumaran: Bootstrap the first version to make sure that, like, anything, like, all of these different pieces we have versions of.
259 00:27:26.840 ⇒ 00:27:28.909 Katherine Bayless: Then, it’s more of, like.
260 00:27:28.910 ⇒ 00:27:34.760 Uttam Kumaran: Because then I can figure out the update flow, and then I’m like, okay, now we can go and, like, make sure it’s right.
261 00:27:34.870 ⇒ 00:27:41.129 Uttam Kumaran: And, like, we’re confident it’ll go get updated, and, like, that’s when I want to, like, pass it to Kai, because I don’t want her worrying about, like.
262 00:27:41.840 ⇒ 00:27:44.309 Uttam Kumaran: the mechanics as much, yeah.
263 00:27:44.520 ⇒ 00:27:55.040 Katherine Bayless: Yeah, no, perfect, yeah, totally. I was gonna say, though, but, like, so I took all that stuff, and I, like, I dropped it into OpenSearch to, like, make a bedrock knowledge base out of it, and so, like, I don’t know if the fact that.
264 00:27:55.040 ⇒ 00:27:55.670 Uttam Kumaran: Oh, great.
265 00:27:55.820 ⇒ 00:28:08.660 Katherine Bayless: vectorized as useful at all. Also, I think there might still be a branch in the repo floating around that’s, like, legacy knowledge base or something, where I had started down this kind of, like, yeah, path, and then I, you know, got distracted and never picked it up.
266 00:28:08.660 ⇒ 00:28:11.939 Uttam Kumaran: Yeah, I guess it sort of depends on, like, how,
267 00:28:12.920 ⇒ 00:28:16.469 Uttam Kumaran: how much it is. I guess it… it… If it’s, like.
268 00:28:16.700 ⇒ 00:28:21.330 Uttam Kumaran: couple hundred files, it’s not that crazy, but I can also try to point to that.
269 00:28:21.710 ⇒ 00:28:22.040 Katherine Bayless: Yeah.
270 00:28:22.040 ⇒ 00:28:24.150 Uttam Kumaran: You know, to search through that, so…
271 00:28:24.400 ⇒ 00:28:33.120 Katherine Bayless: If you go in S3, there is a bucket that has all of it in there that’s, I think… actually, I have it here, because I was just… I put the chatbot data in.
272 00:28:33.120 ⇒ 00:28:33.919 Uttam Kumaran: Oh, nice, okay.
273 00:28:34.000 ⇒ 00:28:40.159 Katherine Bayless: Yeah, so, yeah, CTA, DataOps Archive Marketing Data Team Docs.
274 00:28:40.430 ⇒ 00:28:41.010 Katherine Bayless: Okay.
275 00:28:41.010 ⇒ 00:28:41.569 Uttam Kumaran: Let me take a…
276 00:28:41.570 ⇒ 00:28:56.860 Katherine Bayless: like, this is everything? And I think this is actually where you were looking at the Power BI files, too. So it’s that same bucket. But yeah, so this has all of the various and sundry, things that that marketing team was maintaining, and so it’s like.
277 00:28:56.980 ⇒ 00:28:58.200 Katherine Bayless: From a, you know.
278 00:28:59.210 ⇒ 00:29:06.710 Katherine Bayless: from a value standpoint, I think there is a lot of information that could be gleaned just around, like, you know, what the business rules are and that kind of thing.
279 00:29:07.490 ⇒ 00:29:10.050 Uttam Kumaran: Okay, yeah, let me just note that down to look today.
280 00:29:20.810 ⇒ 00:29:21.420 Uttam Kumaran: Okay.
281 00:29:21.950 ⇒ 00:29:29.709 Uttam Kumaran: Yeah, I think the other… only other piece that I wanted to walk you through was,
282 00:29:31.410 ⇒ 00:29:34.340 Uttam Kumaran: We sort of started, let me just…
283 00:29:34.690 ⇒ 00:29:37.030 Uttam Kumaran: So I don’t know if I can hide this, but I sort of
284 00:29:37.180 ⇒ 00:29:41.189 Uttam Kumaran: finally, like, wrote up, I think, a little bit of a bigger doc on, like, how we’re doing
285 00:29:41.530 ⇒ 00:29:53.980 Uttam Kumaran: sort of the role-based access control. And, like, the net-net is, like, we were just missing some, like, I think, just some language around different roles, so we both have, like, action… access roles and, like, functional roles.
286 00:29:54.050 ⇒ 00:29:57.650 Katherine Bayless: And the access roles roll up into the functional roles.
287 00:29:58.170 ⇒ 00:29:58.700 Katherine Bayless: Okay.
288 00:29:58.700 ⇒ 00:30:02.610 Uttam Kumaran: we can always add more functional roles, right? So we have, like.
289 00:30:02.790 ⇒ 00:30:07.650 Uttam Kumaran: Streamlit Creator, for example, will have Roll ProdMars Read, plus Streamlit
290 00:30:07.820 ⇒ 00:30:11.889 Uttam Kumaran: date, whatever, that DB, read, write. And so…
291 00:30:12.100 ⇒ 00:30:20.589 Uttam Kumaran: I kind of worked a little bit on, like, showing that dependency diagram, but mainly the thing to note is that, like, it all starts with these, like, found functional roles.
292 00:30:20.730 ⇒ 00:30:23.770 Uttam Kumaran: Yeah. Or these, these, these access roles.
293 00:30:23.880 ⇒ 00:30:30.830 Uttam Kumaran: that always roll up. And so, every environment needs, basically, Read, write.
294 00:30:31.040 ⇒ 00:30:35.340 Uttam Kumaran: And then those roll up. And my initial thing was like, okay, is this… is this, like.
295 00:30:35.460 ⇒ 00:30:37.989 Uttam Kumaran: too much isn’t noisy, but I actually think it’s, like.
296 00:30:38.260 ⇒ 00:30:47.490 Uttam Kumaran: it’s just best to do it that way, and then that way, the functional roles are the ones that you can tweak. You’re like, well, I want all the data modelers to have access to, like, write to dev.
297 00:30:47.800 ⇒ 00:30:53.999 Uttam Kumaran: Maybe they write to staging, or maybe not, like, only the CICD can write to staging, they can just read.
298 00:30:54.130 ⇒ 00:30:59.300 Uttam Kumaran: And then they can only read production, right? So nobody can go in and, like, drop tables and things like that.
299 00:30:59.980 ⇒ 00:31:04.049 Uttam Kumaran: A lot of that’s in here, and then we also talk about, like, the service accounts.
300 00:31:04.500 ⇒ 00:31:08.569 Uttam Kumaran: In terms of, like, what roles they have, and, like, what the use case is.
301 00:31:08.750 ⇒ 00:31:13.549 Uttam Kumaran: And then, there’s a… we have, like, kind of the entire script.
302 00:31:13.750 ⇒ 00:31:17.499 Uttam Kumaran: to… Basically, like, execute this end-to-end.
303 00:31:17.680 ⇒ 00:31:25.329 Uttam Kumaran: So, I’m gonna just put a version of this into… I’m gonna… I’ll just have a… put a version of this into our repo.
304 00:31:25.370 ⇒ 00:31:26.879 Katherine Bayless: And then…
305 00:31:26.880 ⇒ 00:31:31.429 Uttam Kumaran: I’m just gonna do a pass and make sure that our snowflake
306 00:31:32.210 ⇒ 00:31:39.869 Uttam Kumaran: matches this, like, so I’ll… I’m just gonna do another pass and just change some roles, and then I’m gonna drop also, like, some of the Snowflake, test
307 00:31:39.980 ⇒ 00:31:44.279 Uttam Kumaran: Like, the out-of-the-box, databases for, like, yeah.
308 00:31:44.280 ⇒ 00:31:46.729 Katherine Bayless: Yeah. And that webhooks one, you can get rid of that one.
309 00:31:46.730 ⇒ 00:31:47.410 Uttam Kumaran: Okay, okay.
310 00:31:48.350 ⇒ 00:31:56.389 Katherine Bayless: Yeah. No, this is awesome. I think, yeah, it’s actually… the dependencies graph is super helpful for my mind.
311 00:31:56.500 ⇒ 00:31:57.630 Katherine Bayless: Yeah.
312 00:31:57.630 ⇒ 00:32:00.820 Uttam Kumaran: Yeah, I don’t know if I can make it bigger, yeah, so…
313 00:32:01.030 ⇒ 00:32:10.169 Katherine Bayless: Just knowing it’s there, because it’s like, yeah, now I can kind of get that sense of, like, what are the primitives, and then what are the actual, like, yeah, yeah, yeah, yeah, I like this, I like this.
314 00:32:10.420 ⇒ 00:32:12.579 Uttam Kumaran: Yeah, and that’s why also, like, I mean, we could…
315 00:32:12.960 ⇒ 00:32:18.029 Uttam Kumaran: Would actually layer on the users, but again, we’re gonna… we’re gonna have more than just, like.
316 00:32:18.980 ⇒ 00:32:23.690 Uttam Kumaran: People that are engineers, modelers, and so it helps to just be like, okay, well.
317 00:32:24.490 ⇒ 00:32:32.329 Uttam Kumaran: We should… when we add a new… when we have a new, functional role, we should think about, okay, what access roles is this tied to?
318 00:32:32.430 ⇒ 00:32:48.289 Uttam Kumaran: And then that way, like, you can drop and change these all you want, because as long as, like, we don’t affect these, we’re sort of good. And these ultimately have all the future… they have, like, future grants and current grants on the appropriate environment.
319 00:32:48.630 ⇒ 00:32:52.520 Uttam Kumaran: And so, like, There’s no… you won’t have to run…
320 00:32:52.780 ⇒ 00:32:59.399 Uttam Kumaran: you shouldn’t have to run anything but, like, grant these roles to these new roles, ultimately.
321 00:32:59.400 ⇒ 00:33:00.460 Katherine Bayless: Right, right.
322 00:33:00.550 ⇒ 00:33:01.390 Uttam Kumaran: Yeah. Yeah.
323 00:33:01.600 ⇒ 00:33:02.910 Katherine Bayless: No, that’s perfect.
324 00:33:03.260 ⇒ 00:33:04.950 Katherine Bayless: Yeah.
325 00:33:05.500 ⇒ 00:33:21.909 Katherine Bayless: Yeah, I think, honestly, having the, like, the prod read and then the streamlit one totally, for the moment, is perfect. This afternoon, we have a meeting with the international team, where I’m going to show them a little bit of Snowflake. I think they’re gonna be very eager, but also very unlikely to, like.
326 00:33:21.910 ⇒ 00:33:30.929 Katherine Bayless: think like a data person, so I’m… I’m kind of curious to see if, I’m curious to see how that will go.
327 00:33:30.930 ⇒ 00:33:39.329 Katherine Bayless: They might be the kind of, like… they might help me figure out, like, what is this other sort of, like, maybe super tightly scoped role need to be, but right.
328 00:33:39.330 ⇒ 00:33:39.950 Uttam Kumaran: Yeah.
329 00:33:40.230 ⇒ 00:33:42.560 Katherine Bayless: This is perfect. Okay.
330 00:33:42.820 ⇒ 00:33:43.830 Katherine Bayless: Yeah, yeah.
331 00:33:44.350 ⇒ 00:33:55.399 Uttam Kumaran: So I should have, like, this in there, and then I should basically go and have this all applied, like, this morning. Nice. And then I’ll add a couple, like, for example, there’s things on, like, how do you do, like, key pair auth.
332 00:33:57.040 ⇒ 00:34:01.090 Uttam Kumaran: Again, this is just, like, It’s just annoying, and it’s a one-time thing.
333 00:34:01.090 ⇒ 00:34:03.019 Katherine Bayless: Yep. For everybody to…
334 00:34:03.210 ⇒ 00:34:08.940 Uttam Kumaran: to start using just better security. So there’s a couple things around that, and then I’ll just make sure all of that gets in there.
335 00:34:09.560 ⇒ 00:34:16.789 Katherine Bayless: Yeah. No, that’s awesome. Those sorts of things make a huge difference, because, yeah, like, you need them, like, once every time you get a new laptop, so you never.
336 00:34:16.790 ⇒ 00:34:21.420 Uttam Kumaran: Yes, and it’s really annoying to do this. It doesn’t make… it’s like…
337 00:34:21.719 ⇒ 00:34:27.180 Uttam Kumaran: Yeah, and a lot of people will just get tripped up, like, you know, so… Yeah. Yeah.
338 00:34:27.770 ⇒ 00:34:34.119 Katherine Bayless: There’s something else this made me think of… Hmm… It’ll come back to me.
339 00:34:35.929 ⇒ 00:34:43.749 Uttam Kumaran: And then Awash and I are about to meet, actually, to talk about identity stitching. I think Ashwini has some PRs out for some of the data modeling changes.
340 00:34:43.750 ⇒ 00:34:44.310 Katherine Bayless: Okay.
341 00:34:45.659 ⇒ 00:34:46.319 Katherine Bayless: Nice.
342 00:34:46.320 ⇒ 00:34:51.999 Uttam Kumaran: And then… yeah, if you want to give me, sort of, the green light, I can set up a Fivetran trial.
343 00:34:52.550 ⇒ 00:35:01.229 Katherine Bayless: Yeah, yeah. I think what I want to do is just kind of, like, wrap my head around it, but yeah, I think that’s probably the right way to do it, is to just kind of, like.
344 00:35:01.780 ⇒ 00:35:04.989 Katherine Bayless: Use that as our reconnaissance phase, but yeah.
345 00:35:06.050 ⇒ 00:35:06.830 Uttam Kumaran: Okay.
346 00:35:06.830 ⇒ 00:35:07.370 Katherine Bayless: Okay.
347 00:35:07.550 ⇒ 00:35:12.189 Katherine Bayless: Cool. Have we talked about that I’m gonna be out for a few days?
348 00:35:12.610 ⇒ 00:35:13.900 Uttam Kumaran: We did not talk about that.
349 00:35:13.900 ⇒ 00:35:18.589 Katherine Bayless: Okay, I’ll be in tomorrow, it’s just Friday, and then out,
350 00:35:18.820 ⇒ 00:35:25.720 Katherine Bayless: well, at least Monday, Tuesday, Wednesday, next week. I did put Thursday, Friday down as PTO, but I have a few weeks that I will gravitate back.
351 00:35:25.720 ⇒ 00:35:26.190 Uttam Kumaran: Okay.
352 00:35:26.210 ⇒ 00:35:41.480 Katherine Bayless: So it’ll be at least, like, the Friday and Monday meetings with the team, but I just, I won’t be there. But I’ll make sure that, I think since Kai is kind of serving as our, you know, de facto project manager sort of thing from Master, I’ll ask her to set up the invites, but…
353 00:35:41.870 ⇒ 00:35:51.180 Uttam Kumaran: Yeah, we’ll still… we’ll still sort of send updates today and tomorrow, so, like, if we want to hop on anytime, just let me know. Yeah, yeah. Otherwise, yeah, we’ll keep Friday and just kind of run through usual stuff.
354 00:35:51.430 ⇒ 00:35:58.169 Katherine Bayless: Okay, cool. Oh, and I need to add you guys to the meeting… well, actually, that’s right, I need to ask Anna to add you guys to the Thursday meetings with the membership.
355 00:35:58.170 ⇒ 00:35:58.900 Uttam Kumaran: Yes.
356 00:35:58.900 ⇒ 00:36:09.569 Katherine Bayless: Yeah. Yeah, this afternoon, we’re gonna do a deep dive with them, to do kind of, like, a For the brave, like, this is how SQL works, just to get a sense of, like, have I now scared you? And you’re like, no, no.
357 00:36:09.570 ⇒ 00:36:09.990 Uttam Kumaran: Okay.
358 00:36:09.990 ⇒ 00:36:19.640 Katherine Bayless: wait for the reports, or are you like, tell me more? And so, I think that’ll be kind of interesting, too, to see how that goes. But I have a feeling that they will be very willing to learn.
359 00:36:19.810 ⇒ 00:36:24.320 Uttam Kumaran: Yeah, I mean, look, as long as they’re willing, and then they tried themselves, and then one day someone’s like.
360 00:36:24.750 ⇒ 00:36:30.979 Uttam Kumaran: I should just pick it up and learn SQL, F it. I’ve been wanting to learn it, you know? That’s great, that’s it.
361 00:36:31.200 ⇒ 00:36:34.060 Katherine Bayless: Yeah, exactly. It’s called Lingua franca for data.
362 00:36:34.060 ⇒ 00:36:35.000 Uttam Kumaran: Yeah, yeah.
363 00:36:36.460 ⇒ 00:36:39.030 Katherine Bayless: But yeah, no, that all sounds good. Okay.
364 00:36:39.350 ⇒ 00:36:42.929 Katherine Bayless: Really appreciate the, yeah, brain trust for the eTail.
365 00:36:42.930 ⇒ 00:36:47.470 Uttam Kumaran: Of course, yeah, no, this is… these are all just, like, nuances, like, that’s not a clear answer, so…
366 00:36:47.610 ⇒ 00:36:55.040 Uttam Kumaran: I’m, like, the last thing I am is married to a vendor, like, we can get this done 10 different ways, so it’s sort of just, like.
367 00:36:55.200 ⇒ 00:37:05.710 Uttam Kumaran: we’re kind of figuring out, like, how the org responds. Like, we poke it, we see how it responds, okay, we should tweak. So there’s all these kind of, like, hacks around, so… yeah, we’ll get there.
368 00:37:05.990 ⇒ 00:37:09.740 Katherine Bayless: Yeah. So, okay. Well, you know where to find me.
369 00:37:09.740 ⇒ 00:37:10.900 Uttam Kumaran: Perfect. Thank you.
370 00:37:11.150 ⇒ 00:37:11.700 Katherine Bayless: See you later.