Meeting Title: iCustomer Snowflake Marketplace Launch Planning Date: 2025-09-11 Meeting participants: Robert Tseng, Uttam Kumaran, Jiten’s AI Notetaker, Abhi Yadav, Jiten Kumar
WEBVTT
1 00:00:12.700 ⇒ 00:00:15.660 Robert Tseng: boost sales caps. I have, data…
2 00:00:16.910 ⇒ 00:00:18.000 Uttam Kumaran: Oh, nice.
3 00:00:28.940 ⇒ 00:00:30.999 Robert Tseng: I like the, the fall decor.
4 00:00:31.830 ⇒ 00:00:36.259 Uttam Kumaran: Yeah, can you tell that’s… I… I definitely did that?
5 00:00:38.130 ⇒ 00:00:42.180 Uttam Kumaran: this week has been, we should put a bunch of fall decor, and I was like, I need…
6 00:00:42.600 ⇒ 00:00:48.369 Uttam Kumaran: wait till Friday. I can help you on Friday. We can go to Costco and get more… get more stuff.
7 00:00:48.550 ⇒ 00:00:50.059 Robert Tseng: What else has she gotten there?
8 00:00:51.010 ⇒ 00:00:55.060 Uttam Kumaran: We have, like, a wreath, and then we’re gonna put some lights up, and… yeah.
9 00:00:55.300 ⇒ 00:00:59.390 Robert Tseng: Nice. You’re gonna do full-on, like, Halloween lights and everything, and do trick-or-treating?
10 00:00:59.390 ⇒ 00:01:01.189 Uttam Kumaran: That was the one thing that…
11 00:01:01.300 ⇒ 00:01:09.839 Uttam Kumaran: she had, which is boxes of decorations. I see. Lugging them around, and so yeah, now we have, like, Christmas, fall.
12 00:01:10.080 ⇒ 00:01:13.130 Uttam Kumaran: So we’ll do some stuff for fall now.
13 00:01:13.130 ⇒ 00:01:13.830 Robert Tseng: Nice.
14 00:01:14.940 ⇒ 00:01:16.490 Uttam Kumaran: That’s nice, nice change.
15 00:01:16.740 ⇒ 00:01:17.420 Robert Tseng: Yeah.
16 00:01:19.620 ⇒ 00:01:22.469 Robert Tseng: Isn’t it still hot in Austin?
17 00:01:22.900 ⇒ 00:01:24.379 Uttam Kumaran: Not as hot.
18 00:01:24.600 ⇒ 00:01:25.460 Robert Tseng: Bye.
19 00:01:25.500 ⇒ 00:01:26.750 Uttam Kumaran: 85?
20 00:01:27.360 ⇒ 00:01:29.420 Robert Tseng: Oh, wow. Yeah, that’s… that’s much better.
21 00:01:29.710 ⇒ 00:01:30.390 Uttam Kumaran: Yeah.
22 00:01:30.530 ⇒ 00:01:31.940 Uttam Kumaran: Okay, so…
23 00:01:35.140 ⇒ 00:01:36.609 Uttam Kumaran: Yeah, not bad.
24 00:01:38.220 ⇒ 00:01:44.149 Robert Tseng: I’ve been seeing this Notetaker TLDV cover a lot more. I don’t know what it is, I’ve never looked at it.
25 00:01:44.150 ⇒ 00:01:45.039 Uttam Kumaran: Oh, really?
26 00:01:54.260 ⇒ 00:01:55.470 Robert Tseng: Yeah, I’m good.
27 00:01:55.710 ⇒ 00:01:57.779 Robert Tseng: A new one every… every few months.
28 00:02:07.330 ⇒ 00:02:12.419 Robert Tseng: I also saw the Omni advice, so that was… I went in there, and I was, like, clicking around my first time in that product.
29 00:02:12.420 ⇒ 00:02:14.910 Uttam Kumaran: Yeah, so I’m gonna set up our…
30 00:02:15.380 ⇒ 00:02:21.749 Uttam Kumaran: Well, one, I’m gonna probably try to just use it for our internal stuff, and then I’m gonna set up it as a dummy thing.
31 00:02:21.950 ⇒ 00:02:22.810 Robert Tseng: Yeah.
32 00:02:22.930 ⇒ 00:02:26.029 Uttam Kumaran: And that will be our, like, enterprise demo, basically.
33 00:02:27.000 ⇒ 00:02:29.450 Robert Tseng: Oh, I see, I see. Okay.
34 00:02:30.120 ⇒ 00:02:33.209 Uttam Kumaran: Like, it’s also, like, I want to pitch Omni over…
35 00:02:33.680 ⇒ 00:02:37.820 Uttam Kumaran: Looker, I think it’s a better product, so… Yeah.
36 00:02:38.170 ⇒ 00:02:39.600 Robert Tseng: Isn’t it more expensive?
37 00:02:40.260 ⇒ 00:02:48.250 Uttam Kumaran: No, it’s cheaper, and they’re… they’re subsidizing… if they… if we recommend clients, they will help subsidize, so…
38 00:02:48.550 ⇒ 00:02:49.190 Robert Tseng: Okay.
39 00:02:49.930 ⇒ 00:02:50.620 Robert Tseng: -Oh.
40 00:02:52.060 ⇒ 00:02:56.829 Robert Tseng: Yeah, no, it looks good. I like these demos, I’ll probably use them for future calls.
41 00:02:57.100 ⇒ 00:02:57.760 Uttam Kumaran: Whoa.
42 00:02:57.910 ⇒ 00:02:58.430 Robert Tseng: Yeah.
43 00:02:58.710 ⇒ 00:02:59.840 Abhi Yadav: Hey guys!
44 00:03:00.690 ⇒ 00:03:01.729 Robert Tseng: Hey, Abby!
45 00:03:02.350 ⇒ 00:03:03.530 Abhi Yadav: Oh, you?
46 00:03:03.540 ⇒ 00:03:04.129 Uttam Kumaran: There you go.
47 00:03:04.130 ⇒ 00:03:04.899 Robert Tseng: How are you?
48 00:03:05.900 ⇒ 00:03:10.819 Abhi Yadav: Yeah, all good. Thanks, thanks for spending time with us today.
49 00:03:11.440 ⇒ 00:03:31.100 Robert Tseng: Yeah, I want… I brought my, business partner, Utam, onto the call. I think if you… I saw you added a head of… you’re a head of engineering, so if he had any, like, super technical, nitty-gritty things, I think Utah would be a better person to… to flex our expertise there. But, yeah, happy to… I just wanted to introduce both of us to you anyway.
50 00:03:31.580 ⇒ 00:03:38.970 Abhi Yadav: Cool. Nice to meet you both. Hi. Yeah. Are you guys in New York? Which part of the world?
51 00:03:39.900 ⇒ 00:03:47.370 Robert Tseng: I’m in New York, we’ve done Austin, but we’re both Bay Area boys, so we both grew up with the Bay. I think you’re in SF, right?
52 00:03:48.080 ⇒ 00:03:53.609 Abhi Yadav: Yeah, we’re, SF in Boston, kind of, crew, as well.
53 00:03:53.610 ⇒ 00:03:54.600 Robert Tseng: Okay. Cool.
54 00:03:54.600 ⇒ 00:03:55.060 Uttam Kumaran: Nice.
55 00:03:55.060 ⇒ 00:03:59.330 Abhi Yadav: So, I’m just waiting for the thing to join.
56 00:03:59.530 ⇒ 00:04:06.590 Abhi Yadav: So, so, so tell me a little bit about you guys, like, are you, like, kind of…
57 00:04:07.080 ⇒ 00:04:15.239 Abhi Yadav: You know, a boutique, agency, or, like, just give me a quick rundown, if you don’t mind.
58 00:04:15.760 ⇒ 00:04:21.750 Robert Tseng: Yeah, yeah, exactly. We are… we’re a boutique agency, we both kind of are…
59 00:04:22.480 ⇒ 00:04:31.759 Robert Tseng: previously in-house, kind of, data folks. We talked a lot more on the engineering side, and then me more on the, like, business intelligence, product analytics side.
60 00:04:31.900 ⇒ 00:04:45.780 Robert Tseng: Yeah, and then we’ve been… we’ve been working together, running this business for the past 2 years. Yeah, I guess we kind of operate as a fractional data team, and then we’ve also built out this whole AI automation kind of, like, arm to it as well, so…
61 00:04:46.910 ⇒ 00:04:54.790 Robert Tseng: Yeah, I think we view AI as, like, a extension of, like, the data problem to solve, and so we’ve been able to
62 00:04:55.110 ⇒ 00:05:14.950 Robert Tseng: be able to go really end-to-end from standing up data stacks of, you know, landing data warehousing, all the way through BI reporting, and then all the way through to be able to put insights into specific tools that are… that stakeholders work out of, whether it’s Slack, or email, or kind of surfacing kind of information wherever they need it to be done.
63 00:05:15.810 ⇒ 00:05:16.710 Abhi Yadav: I see.
64 00:05:16.780 ⇒ 00:05:17.480 Robert Tseng: Yeah.
65 00:05:17.640 ⇒ 00:05:29.760 Abhi Yadav: And, you know, from a tech stack perspective, I’m assuming, like, you’re more like a… Google, Cloud…
66 00:05:29.900 ⇒ 00:05:33.110 Abhi Yadav: Snowflake, some of this… .
67 00:05:33.110 ⇒ 00:05:39.909 Robert Tseng: Yeah, I mean, Snowflake is probably our bread and butter, but normally we have clients that are working in Azure, BigQuery, AWS, all of them.
68 00:05:41.100 ⇒ 00:05:49.920 Abhi Yadav: I see. And, just to give you our side of background, you know, we, typically…
69 00:05:50.260 ⇒ 00:05:59.849 Abhi Yadav: So, just, just my background before this company. So, I don’t know if you’ve heard the word, you know, CDP,
70 00:06:00.120 ⇒ 00:06:02.640 Abhi Yadav: But we kind of,
71 00:06:02.990 ⇒ 00:06:09.799 Abhi Yadav: you know, pioneered that category. We sort of built CDPs, and for the last 10 years,
72 00:06:09.950 ⇒ 00:06:16.430 Abhi Yadav: You know, not, like, segment, but, like, kind of a little bit different, and then…
73 00:06:16.540 ⇒ 00:06:34.310 Abhi Yadav: you know, CDPs were, like, data decision delivery, so we were kind of laser-focused on, like, identity and decisioning, like, the middle layer. And then, we sold that company, and then now, with iCustomer, we’re trying to do,
74 00:06:34.380 ⇒ 00:06:46.130 Abhi Yadav: something in between the data platform, which is, like, now we’re saying you don’t need, like, a CDP, and you can basically have, like, a homegrown…
75 00:06:46.480 ⇒ 00:06:54.510 Abhi Yadav: Snowflake as a CDP, or, you know, Databricks as a CDP, or Google BigQuery as a CDP.
76 00:06:54.910 ⇒ 00:06:58.799 Abhi Yadav: But… and that’s how, like, most of the business are now adopting.
77 00:06:59.860 ⇒ 00:07:03.929 Abhi Yadav: But there are still, like, a lot of, like,
78 00:07:04.490 ⇒ 00:07:09.420 Abhi Yadav: You know, the problem of, like, a lot of custom work to be done is too much.
79 00:07:09.580 ⇒ 00:07:18.920 Abhi Yadav: And then, it’s not the engineering problem, it’s more like, okay, why we should be doing this, or that, or this, or that, sort of thing.
80 00:07:19.230 ⇒ 00:07:29.920 Abhi Yadav: So, we were tackling the same thing and every time, and we’d love to explore possibility of working with you guys and some of our other engagement. We, we…
81 00:07:30.880 ⇒ 00:07:51.040 Abhi Yadav: We get, like, a lot of bandwidth issues at times, where we have, like, active implementation, and people saying, look, we have a CDP, can you sit on top? Because that’s what we do, and then we realize the thing is pretty broken already, and then they’re like, okay, can you create a new project, you know, within this?
82 00:07:51.200 ⇒ 00:08:01.500 Abhi Yadav: So we end up inheriting a lot of, like, past CDP work again in this company, which we don’t want to deal with, but we want to basically,
83 00:08:01.610 ⇒ 00:08:05.900 Abhi Yadav: Have somebody, like, okay, we can have, like, a partner, or some…
84 00:08:06.190 ⇒ 00:08:19.640 Abhi Yadav: consultant, kind of do the kind of building or build-out. So that’s, like, the far-along kind of thing. We always kind of exploring folks to bring in. We already have some, by the way, just so you know.
85 00:08:20.270 ⇒ 00:08:31.419 Abhi Yadav: But this particular project, which I kind of posted as… we are trying to get ourselves, like, launch in the Snowflake marketplace.
86 00:08:31.770 ⇒ 00:08:32.520 Uttam Kumaran: Nice.
87 00:08:32.880 ⇒ 00:08:43.859 Abhi Yadav: and GBQ, like, Marketplace, and all these places. So we’re trying to, like, build out a few things, as per what they have
88 00:08:43.860 ⇒ 00:08:57.699 Abhi Yadav: been asking us. We have, like, joined go-to-market plans, like, you know, AI Native, CDP, or something, which Snowflake and us, we’re gonna market together.
89 00:08:57.890 ⇒ 00:09:03.820 Abhi Yadav: So we are trying to, like, explore a little bit of, like.
90 00:09:04.830 ⇒ 00:09:09.809 Abhi Yadav: You know, we… and we have our standard approach of building multi-stage
91 00:09:10.520 ⇒ 00:09:29.470 Abhi Yadav: you know, kind of, data thing within Snowflake, and GBQ, like, right now, like, we… we have it working, like, we… we just call it as V1, right? But then, we’re always curious, like, how to, like, sort of build…
92 00:09:30.020 ⇒ 00:09:39.190 Abhi Yadav: Like, go back to, like, let’s call it 2.0 of that, and learn from our product experience, so the query can be optimized.
93 00:09:39.190 ⇒ 00:09:41.509 Jiten Kumar: Query can be optimized, billing…
94 00:09:43.090 ⇒ 00:09:43.710 Abhi Yadav: Oh.
95 00:09:43.710 ⇒ 00:09:44.250 Jiten Kumar: Oh.
96 00:09:44.250 ⇒ 00:09:44.980 Abhi Yadav: You know.
97 00:09:44.980 ⇒ 00:09:47.409 Jiten Kumar: You know, what in… when I’m…
98 00:09:48.030 ⇒ 00:09:49.619 Abhi Yadav: Can you guys hear me okay?
99 00:09:50.020 ⇒ 00:09:50.350 Robert Tseng: Yeah.
100 00:09:50.350 ⇒ 00:09:52.220 Uttam Kumaran: It’s echoing a little bit.
101 00:09:52.220 ⇒ 00:09:54.859 Robert Tseng: I could hear you twice, that’s all I can…
102 00:09:54.860 ⇒ 00:09:58.540 Jiten Kumar: Okay, sorry, sorry guys. Yeah, I’m in now.
103 00:09:59.820 ⇒ 00:10:04.130 Abhi Yadav: Okay, well, it’s nothing about you, it’s about my echo.
104 00:10:04.240 ⇒ 00:10:20.420 Abhi Yadav: Alright, so, I was just curious, like, is that something closer to this? Like, have you guys done anything? Like, tell me, like, some of the closest projects you may have done, and then we can sort of build up from there.
105 00:10:21.130 ⇒ 00:10:40.290 Uttam Kumaran: Yeah, maybe I’ll talk a little bit about the Snowflake side, and I can let Robert talk about a lot of the work we do on the CDP side. So we’ve, you know, at the company, myself and several members, we’ve done Snowflake now almost 8 or 9 years. You know, I started using Snowflake in 2018.
106 00:10:40.520 ⇒ 00:10:45.040 Uttam Kumaran: And so, you know, kind of everything that you listed around
107 00:10:45.460 ⇒ 00:10:49.070 Uttam Kumaran: Edfs, functions, snow pipe.
108 00:10:49.700 ⇒ 00:10:52.309 Robert Tseng: You know, all the way up to…
109 00:10:52.440 ⇒ 00:11:02.119 Uttam Kumaran: We have also published data sets in the marketplace. We’ve procured tools from the marketplace. So kind of really familiar with the entire Snowflake ecosystem.
110 00:11:02.450 ⇒ 00:11:19.090 Uttam Kumaran: And so, kind of understand where you guys are going into trying to package, you know, queries and functions and sort of go through their marketplace to get distribution and to get people to procure from there. So, really familiar with that. I mean, fundamentally, we understand all the core
111 00:11:19.090 ⇒ 00:11:27.839 Uttam Kumaran: sort of snowflake objects and functions. We do a lot of work in dbt. Again, a tool that we’ve been using, you know, since 2018.
112 00:11:29.050 ⇒ 00:11:40.309 Uttam Kumaran: So everything from, you know, basic models to, you know, having, like, multi-warehouse setups, macros, functions within DBT,
113 00:11:40.700 ⇒ 00:11:49.169 Uttam Kumaran: you name it, we’ve done it in dbt. So you’ve also mentioned some of the new Snowflake products, like Tasks.
114 00:11:49.350 ⇒ 00:12:09.019 Uttam Kumaran: There’s also a lot more observability and monitoring that you can do in Snowflake now versus just a few years ago. So, certainly, I agreed to try to leverage that, as much as possible. Previously, like, we had to bring in, like, a Datadog or another tool. You can actually do a lot of task scheduling, in Snowflake directly versus getting, like, a Dagster or an Airflow.
115 00:12:09.020 ⇒ 00:12:11.590 Abhi Yadav: So, definitely something we recommend as well.
116 00:12:11.590 ⇒ 00:12:17.999 Uttam Kumaran: So, a lot of experience working in Snowflake. We also do a lot of implementations within BigQuery, within Redshift.
117 00:12:18.940 ⇒ 00:12:20.490 Uttam Kumaran: For folks, as well.
118 00:12:20.520 ⇒ 00:12:33.489 Uttam Kumaran: So… and then, for a lot of our clients, we handle ETL, modeling, you know, storage, warehousing, governance, and then we handle BI4. So this seems mostly around Snowflake, around…
119 00:12:33.490 ⇒ 00:12:41.239 Uttam Kumaran: sort of data modeling, data governance, data engineering. so I… I don’t, I don’t feel,
120 00:12:41.320 ⇒ 00:12:54.469 Uttam Kumaran: you know, too nervous about anything you mentioned. I’ll let maybe Robert talk a little bit about our experience on the CDP side, and I kind of get a little bit more. I have a couple questions about, sort of, the product and your positioning. But yeah, maybe Robert, if you want to talk about some of our CDP experience.
121 00:12:54.470 ⇒ 00:13:11.260 Robert Tseng: Yeah, yeah, so, I guess, just to kind of build on that, so, yeah, we’re implementers of the CDP tools, Rudderstack, PyTouch, and Segment, but mostly, mostly we’re using Segment for clients, and so we kind of know, like, the limitations of what Segment’s good at, what it’s not good at.
122 00:13:11.330 ⇒ 00:13:21.930 Robert Tseng: What we found just across the board is that we move identity resolution into the warehouse through BBT, and so, we don’t… we don’t rely on these CBP tools to do that for us.
123 00:13:22.010 ⇒ 00:13:39.060 Robert Tseng: you know, they… Segment just builds you with a limited set of set of properties, and so it’s just… and just to have more controls over, like, what’s actually happening when we’re doing the stitching, we’ve just found to, like, bring all of that logic into… into the warehouse and orchestrate it through dbt to be the better approach.
124 00:13:39.060 ⇒ 00:13:45.180 Robert Tseng: So, yeah, we do kind of use these, you know, tools pretty bare bones, like, we just use it for
125 00:13:45.180 ⇒ 00:13:51.240 Robert Tseng: You know, if it’s… rather than setting up a bunch of custom connectors, we just kind of can use the point-and-click solution and let
126 00:13:51.290 ⇒ 00:14:14.009 Robert Tseng: non-technical teams go in and set up, you know, custom webhooks, or being able to, like, move data, you know, from point A to point B. Like, that part is fine. We usually let our clients kind of, like, work with that. Anything that’s critical data, transactions, or customer profiles that needs to be really controlled, especially for our healthcare clients that are really sensitive to, you know, PII data.
127 00:14:14.010 ⇒ 00:14:17.329 Robert Tseng: We, we, we, we take a firmer stance on,
128 00:14:17.330 ⇒ 00:14:27.819 Robert Tseng: be able to control that data in the warehouse. So, hopefully that just gives you a, you know, flavor of, like, how nuanced we are with how we use, these, these, kind of, a traditional CDP tool.
129 00:14:28.970 ⇒ 00:14:29.520 Abhi Yadav: Yeah, no.
130 00:14:29.520 ⇒ 00:14:40.629 Uttam Kumaran: what you mentioned, like, we walk into a lot of situations where sometimes we do have the decision-making power to, like, procure what we want or set it up. Sometimes we just walk into whatever the situation is, right? So, like.
131 00:14:40.770 ⇒ 00:14:50.979 Uttam Kumaran: It’s sort of give and take for us. Of course, we’re consultants, so we are there to solve the problem, but also we consider, like, short-term and long-term, which is, I think.
132 00:14:51.030 ⇒ 00:15:10.150 Uttam Kumaran: what has got us to this point in the business is we come in and we fix, like, whatever the issue is in the first 4-6 weeks, but we also help them understand, like, their vendor procurement strategy, how much they’re paying for tools, how do tools play nice with each other, how you actually build, like, an actual data organization, for, like, for long-term, you know, management.
133 00:15:10.820 ⇒ 00:15:14.610 Abhi Yadav: I see. Very helpful. So, you mentioned an…
134 00:15:14.940 ⇒ 00:15:21.990 Abhi Yadav: From our product positioning standpoint, so, we kind of…
135 00:15:22.060 ⇒ 00:15:23.340 Uttam Kumaran: you know…
136 00:15:23.360 ⇒ 00:15:29.609 Abhi Yadav: trying to follow, like, you may be familiar with tools like Hightouch and Growth Loop.
137 00:15:29.950 ⇒ 00:15:35.720 Abhi Yadav: So, we’re kind of a little bit on those lines, but not entirely.
138 00:15:35.830 ⇒ 00:15:38.660 Abhi Yadav: But…
139 00:15:38.780 ⇒ 00:15:59.329 Abhi Yadav: We’re, like, composable, we don’t, I mean, that’s the similarity with them, because, you know, you can start off with, like, within your warehouse, so we don’t have a backend, we need access, and we create project… dedicated project ID in JBK or Snowflake, depending on which warehouse you’re in.
140 00:15:59.510 ⇒ 00:16:03.970 Abhi Yadav: And then we basically bring, like, an IDE…
141 00:16:04.220 ⇒ 00:16:18.520 Abhi Yadav: resolution, if most of the customer we deal with, they already have built-in, or maybe shoddy one, maybe sophisticated one. Sometimes, most of the time, we end up redoing it.
142 00:16:18.610 ⇒ 00:16:26.809 Abhi Yadav: And the reason we redo the ID resolution part is because this is where our, like, secret sauce is.
143 00:16:26.920 ⇒ 00:16:38.860 Abhi Yadav: We also have a graph of verified people, companies, activity, device, like, built on top of a data cooperative. So…
144 00:16:38.950 ⇒ 00:16:52.730 Abhi Yadav: like, you can have a tenant-specific ID… so our ID resolution is two-stage, you know, one, the standard ID resolution, which is kind of tenant-specific, you know, A plus B is called the A sort of thing, and then…
145 00:16:53.070 ⇒ 00:17:01.260 Abhi Yadav: we enrich our verified graph IDs, so think of, like, putting Experian ID into your…
146 00:17:01.260 ⇒ 00:17:01.870 Uttam Kumaran: by the GS.
147 00:17:01.870 ⇒ 00:17:03.549 Abhi Yadav: Yeah, I am,
148 00:17:03.720 ⇒ 00:17:17.190 Abhi Yadav: And then, we bring a lot of, like, signals, fireos, and all that, so it’s, like, signal-based marketing, and some of those are kind of our most, popular use case.
149 00:17:17.420 ⇒ 00:17:19.250 Abhi Yadav: This is how we…
150 00:17:19.780 ⇒ 00:17:33.770 Abhi Yadav: talk about it, just, like, one quick slide, just to give you guys familiarity. So, whatever your internal data model is, and your docs, and… because it’s both, like, we’re also making the knowledge engineering layer.
151 00:17:33.830 ⇒ 00:17:41.720 Abhi Yadav: So… and then we have this open graph of ours, which has all the data signals mapped to entities.
152 00:17:41.830 ⇒ 00:17:52.900 Abhi Yadav: So we end up creating this live graph, which is basically nothing but, like, unified golden record plus enrichment, right? Sort of a thing.
153 00:17:53.240 ⇒ 00:18:02.700 Abhi Yadav: And then there’s some embedded decisioning, model analytics, things like that, so you end up doing a lot of use cases around paid media.
154 00:18:02.860 ⇒ 00:18:09.540 Abhi Yadav: And that’s why analytics and paid media analytics is kind of important, and then basically segmentation.
155 00:18:09.760 ⇒ 00:18:20.160 Abhi Yadav: I mean, this is… we’re super, like, and this is what we’re trying to, like, do in the V2 now, like, a warehouse native package.
156 00:18:20.290 ⇒ 00:18:29.089 Abhi Yadav: So whether it’s BigQuery, Redshift, doesn’t matter, you know, and this is the part where it’s BA Analytics, or agents, or, like…
157 00:18:29.190 ⇒ 00:18:32.589 Abhi Yadav: Activation, semantic, like, whatnot.
158 00:18:32.800 ⇒ 00:18:41.459 Abhi Yadav: So pretty, pretty standard stuff for some of you, it sounds like you’re familiar to the kind of world we are in.
159 00:18:43.550 ⇒ 00:18:45.810 Abhi Yadav: Any thought, questions so far?
160 00:18:46.020 ⇒ 00:18:48.210 Uttam Kumaran: Yeah, I mean, I always thought, like…
161 00:18:48.460 ⇒ 00:19:06.319 Uttam Kumaran: I was just… I feel like the adoption of a Snowflake marketplace for products like yours has been slow. Like, I think it’s really amazing to… for companies to be able to procure directly through Marketplace, because they already have an existing relationship through Snowflake, just like any other past marketplaces. I’m surprised that it hasn’t, like, taken off
162 00:19:06.410 ⇒ 00:19:18.700 Uttam Kumaran: I think when I first started a few years ago, the tooling was really tough, like, really tough to develop on, so I assume it’s maybe a little bit better now. Like, I was… I used to lead product at a startup where we were thinking about doing this.
163 00:19:18.720 ⇒ 00:19:36.600 Uttam Kumaran: But the lift was too much, and I was like, or it’s not… it’s not us worth building, sort of, the native connectors, but I think it’s a great way for products like yours, especially, if you don’t need a backend, to… to take advantage. And, again, if… it’ll just go straight up to their established paper and billing account, right? So the…
164 00:19:36.600 ⇒ 00:19:44.969 Abhi Yadav: Exactly, yeah. So that’s, that’s what is driving, so we’re, like, a little bit on time pressure.
165 00:19:45.150 ⇒ 00:19:57.120 Abhi Yadav: And so we have some of these, like, sales side, some connection at the CXO level, so we can, like, move things, like, faster on the thing, but…
166 00:19:57.700 ⇒ 00:20:10.610 Abhi Yadav: And they’re also willing to, like, support with resource and whatnot, sort of a thing, to your point. And we also kind of validated this when we were doing internally, that…
167 00:20:10.840 ⇒ 00:20:15.939 Abhi Yadav: It’s a lot easier, like what it was in our last startup situation.
168 00:20:16.090 ⇒ 00:20:25.169 Abhi Yadav: Like, we were actually going a little bit on, like… we were built on Snowflake, we were early even in our last company. It was a huge lift.
169 00:20:25.740 ⇒ 00:20:26.250 Uttam Kumaran: Yeah.
170 00:20:26.250 ⇒ 00:20:36.839 Abhi Yadav: To your point, like, I think we remember, like, Thin may remember, like, we had, like, 5, 6 dedicated data engineers just to do snowflake work.
171 00:20:36.980 ⇒ 00:20:43.930 Abhi Yadav: Yeah. Sort of a thing, which is huge. But now it sounds like, our estimate is, like.
172 00:20:44.290 ⇒ 00:20:53.780 Abhi Yadav: You know, within, like, 3 to 4 weeks, you can go live, and, you know, it’s basically a lot of those services can be downloadable and kind of repackaged.
173 00:20:54.040 ⇒ 00:20:54.670 Abhi Yadav: So that’s…
174 00:20:54.670 ⇒ 00:21:14.620 Uttam Kumaran: It also depends, like, what you’re doing, like, what use cases there are, but certainly, like, because some people just use it as, like, the way to sign something, and everything is off of the platform. Some people, it’s end-to-end within Snowflake, so I feel like you’re right, just to get it up there, and then everything is… you have to still talk to somebody, and there can be some manual configs, and eventually you can…
175 00:21:14.620 ⇒ 00:21:16.410 Uttam Kumaran: Self-serve a lot of it, yeah.
176 00:21:16.860 ⇒ 00:21:26.840 Abhi Yadav: So that… that’s exactly what the project is, and… and that’s why we were, like, looking for somebody more experienced. So, we wanted to… it’s like…
177 00:21:27.920 ⇒ 00:21:42.889 Abhi Yadav: It’s like a reusable situation is what we need. And, you know, we’re just about to raise a big round in, like, a couple of weeks, so we want to get it done before that.
178 00:21:43.020 ⇒ 00:21:48.659 Abhi Yadav: So we can do this announcement quickly, and, you know, we already have secured
179 00:21:49.210 ⇒ 00:22:07.539 Abhi Yadav: customers through Snowflake, GBQ, there’s, like, an active integration going on right now, so there’s already, that, and then we spent a lot of time last couple of months on integrating with a company, all for distribution first, right? Like, so we start…
180 00:22:07.720 ⇒ 00:22:22.109 Abhi Yadav: we focus a lot with, like, Railtio as a MDM company, so now, because of RailTio, we are in, like, Petco and major, like, accounts, primarily, so we… we’re trying to trade paperwork, like, contract…
181 00:22:22.350 ⇒ 00:22:38.020 Abhi Yadav: instead of getting into InfoSec approval cycle and that all stupidity, like, we’re just exchanging paperwork with Snowflake, and then with reality. And, you know, and then we’ll figure it out later, like, is what our plan is.
182 00:22:38.370 ⇒ 00:22:45.909 Abhi Yadav: So, going back to the project, so, just from a scoping, we wanted to kind of understand a little bit on, like.
183 00:22:46.020 ⇒ 00:22:49.459 Abhi Yadav: If we have to do, like, a…
184 00:22:50.660 ⇒ 00:23:01.379 Abhi Yadav: And, you know, we’re, like, super flexible to even do this, like, kind of, like, long-term and, you know, sort of basis, given, and also explore.
185 00:23:01.420 ⇒ 00:23:12.029 Abhi Yadav: If you’re interested, be on this project also for, like, when we have new customer onboarding, implementation kind of cycle, so you could kill two birds with one stone.
186 00:23:12.150 ⇒ 00:23:23.890 Abhi Yadav: But since we have never worked together, I want to, like, carve out, like, a small project, more like a pilot, in the direction of, like, this marketplace launch.
187 00:23:24.010 ⇒ 00:23:25.210 Abhi Yadav: But…
188 00:23:25.540 ⇒ 00:23:35.489 Abhi Yadav: Maybe to start with, and I’ve seen… thanks, Robert, for sharing, some of the stuff. Like, one other thing I was, like, looking at.
189 00:23:35.660 ⇒ 00:23:49.770 Abhi Yadav: what you did with the sample media and advertising reporting, like, BI stuff is, like, kind of super, kind of, like, what you call, you know, a good starting point, like, solid V1, right? No, no problem.
190 00:23:50.020 ⇒ 00:23:53.330 Abhi Yadav: But since our work is beyond
191 00:23:53.550 ⇒ 00:23:57.560 Abhi Yadav: like, the diagram I showed you, the most…
192 00:23:58.060 ⇒ 00:24:02.409 Abhi Yadav: Common use case for us is to activate audience.
193 00:24:02.910 ⇒ 00:24:03.450 Robert Tseng: Yeah.
194 00:24:03.450 ⇒ 00:24:08.710 Abhi Yadav: So, we are building audience, like, which is your entire deduplicated, unified data.
195 00:24:08.910 ⇒ 00:24:11.780 Abhi Yadav: And then you had to activate the audience.
196 00:24:12.020 ⇒ 00:24:15.050 Abhi Yadav: Right? Which is a little bit heavy lift.
197 00:24:15.150 ⇒ 00:24:23.630 Abhi Yadav: Which, as compared to, like, so, in the technical term, if you had to create multiple marts.
198 00:24:23.790 ⇒ 00:24:30.139 Abhi Yadav: You know, the Shopify mart for sales is kind of the most straightforward.
199 00:24:30.250 ⇒ 00:24:34.619 Abhi Yadav: Then for advertising and kind of analytics, it’s like…
200 00:24:34.720 ⇒ 00:24:38.059 Abhi Yadav: the other… but then this,
201 00:24:38.380 ⇒ 00:24:58.329 Abhi Yadav: you know, the time series of customer, and that behavior, and where they came in, they came not, and we have, like, Pixel on the website, and all kinds of, like, visitor intelligence we could do all that, just because we have the data. We’re trying to connect the dots there. That’s where is we feel a little bit of…
202 00:24:58.630 ⇒ 00:25:06.459 Abhi Yadav: Like, heavy lift-esque… and not, like, heavy lift in the actual weeds, but, like, architecting it in a way,
203 00:25:06.580 ⇒ 00:25:11.880 Abhi Yadav: So it’s, like, reusable for multiple clients, same package, package way.
204 00:25:12.040 ⇒ 00:25:13.330 Abhi Yadav: Does that make sense?
205 00:25:14.060 ⇒ 00:25:19.730 Uttam Kumaran: Yeah, I mean, it’s actually, Robert, it’s very similar to, like, kind of what we did at Prequel, so I…
206 00:25:20.000 ⇒ 00:25:29.340 Uttam Kumaran: Used to lead, kind of, product at this company called Prequel, before… And for Brainforge,
207 00:25:29.710 ⇒ 00:25:37.830 Uttam Kumaran: right now, they’re sort of pivoted, they’re focused on finance, but basically, we were building, like, a no-code, DBT, almost.
208 00:25:37.830 ⇒ 00:25:53.209 Uttam Kumaran: And so one of the things that we had to build was, domain-specific marts. So, we always have a mart that can support, e-com, that can support paid marketing, that can support basic finance reporting, and so…
209 00:25:53.550 ⇒ 00:26:08.020 Uttam Kumaran: But one thing that we learned is, yes, there are very well-defined KPIs within each of those. You can start to build a semantic layer that basically maps common entities into that sort of finalized mart, right? Because in a finance mart, for example, you’re looking at
210 00:26:08.020 ⇒ 00:26:19.500 Uttam Kumaran: typically transactions, and then you want typical KPIs that come out based on if you’re SaaS or if you’re a different model. So, sounds kind of similar in that what you’re doing is, like, can we…
211 00:26:19.540 ⇒ 00:26:29.169 Uttam Kumaran: depending on what the client brings, which is, like, kind of like BYO transactions Table or BYO orders table, you map it into your system.
212 00:26:29.230 ⇒ 00:26:40.880 Uttam Kumaran: there’s actually kind of two ways of kind of, like, trying to tackle this. One is you can do everything kind of like an event stream. I don’t know if you’ve considered, like, an activity schema or something, where you look at all
213 00:26:40.980 ⇒ 00:26:52.489 Uttam Kumaran: rows is kind of like an event, and an event can be a transaction, a customer visit, right? Everything can be modeled that way. An event has a timestamp, an object, and usually just, like, metadata fields.
214 00:26:52.780 ⇒ 00:26:56.700 Uttam Kumaran: You create a giant event stream, and then you can sort of parse that out as you need.
215 00:26:56.850 ⇒ 00:27:13.959 Uttam Kumaran: You could also do it like, hey, we need a transactions table with a minimum of these five columns, and if you have these columns, these are also helpful. Part of this can be more manual, part of this, ideally, handle it all, and sort of in Snowflake, like, point us to your table. Is that sort of like…
216 00:27:14.540 ⇒ 00:27:16.650 Uttam Kumaran: Kind of the direction.
217 00:27:17.200 ⇒ 00:27:27.959 Abhi Yadav: Yeah, I know, directionally, you know, it’s kind of close to what you just said, except just one thing, and you hit the nail on the head, I didn’t use the word, but…
218 00:27:28.080 ⇒ 00:27:39.070 Abhi Yadav: you know, activity… like, it’s people, companies, and activities. And by the way, our customer is, like, both consumer business, like D2C and B2B.
219 00:27:39.130 ⇒ 00:27:54.479 Abhi Yadav: Good news is our B2B side is far more mature than some of the consumer side, which is actually the harder problem to tackle. We all come from retail and e-commerce, so we’re kind of super…
220 00:27:54.480 ⇒ 00:28:00.219 Abhi Yadav: like, oh, we can handle this, because, you know, there’s no association logic issue, but in B2B,
221 00:28:00.670 ⇒ 00:28:07.880 Abhi Yadav: We’re dealing with, like, moving targets, people changing jobs, then you don’t know, like, who this person is,
222 00:28:08.100 ⇒ 00:28:18.780 Abhi Yadav: So, the activities part, what you just said, is kind of crucial, because we are… we have this agentic
223 00:28:18.940 ⇒ 00:28:27.660 Abhi Yadav: Orchestration, like, you know, erstwhile reversal, ETL, and we sometimes need to see, based on the activities.
224 00:28:27.660 ⇒ 00:28:28.130 Uttam Kumaran: Yeah.
225 00:28:28.130 ⇒ 00:28:29.829 Abhi Yadav: What to trigger, you know?
226 00:28:29.830 ⇒ 00:28:30.270 Uttam Kumaran: Exactly.
227 00:28:30.270 ⇒ 00:28:38.600 Abhi Yadav: That’s the downstream automation we have built out. But then if it’s not structured in the right way, then this whole thing suffers.
228 00:28:38.690 ⇒ 00:28:57.469 Uttam Kumaran: Yeah, so that’s… this is… and it’s typically… most people don’t do activity schema because they’ve already done dimensional modeling, right? So, even for some of our clients, like, it’s also something that, given the size of your… of your information and how many, it’s much easier to just model everything as an event.
229 00:28:57.710 ⇒ 00:29:12.210 Uttam Kumaran: toss it in somewhere, and then later sort of parse it out into the object models you need. Otherwise, it’s just kind of, like, a lot of edge cases, basically, on, like, oh, this person has this dimension. So, yeah, totally hear you.
230 00:29:12.550 ⇒ 00:29:23.869 Abhi Yadav: Yeah, and, I mean, just as simple as, like, keeping, the data model outside the database, or being it in dbt, and then…
231 00:29:24.020 ⇒ 00:29:26.820 Abhi Yadav: Kind of agnostic to, yeah.
232 00:29:27.300 ⇒ 00:29:33.810 Abhi Yadav: the table is what we kind of realized. So… so, I mean, you’re saying all the right thing, and…
233 00:29:34.050 ⇒ 00:29:38.960 Abhi Yadav: You know, just so you know, like, we, we just wanted, like, if not…
234 00:29:39.200 ⇒ 00:29:43.270 Abhi Yadav: Somebody like, you know, the… the…
235 00:29:43.900 ⇒ 00:29:52.340 Abhi Yadav: kind of reasonably understand some of these nuances. We don’t call ourselves, like, super expert, but we’re, like, a reasonable benchmark we have.
236 00:29:52.340 ⇒ 00:29:52.900 Uttam Kumaran: Sure.
237 00:29:52.900 ⇒ 00:30:06.420 Abhi Yadav: So, it’s just like our bandwidth is limited. So, we have to divide and conquer. There are a lot of projects in the flight, so we’re tackling this one, we kind of focus on the…
238 00:30:06.990 ⇒ 00:30:16.290 Abhi Yadav: you know, this agentic thing we tried doing with, like, Langrav and, like, 5,000 frameworks, now we’re now building our own, sort of thing.
239 00:30:16.290 ⇒ 00:30:16.760 Uttam Kumaran: Okay.
240 00:30:17.100 ⇒ 00:30:19.860 Uttam Kumaran: I mean, that’s… so that’s a lot big… that’s probably now…
241 00:30:19.860 ⇒ 00:30:20.490 Abhi Yadav: No.
242 00:30:20.490 ⇒ 00:30:22.919 Uttam Kumaran: or half of our business, we…
243 00:30:22.980 ⇒ 00:30:37.280 Uttam Kumaran: We do a lot of AI work, so we… the business started as purely a data company. We were using AI to build a company, and then after figuring… walking through the same process on, like, how do we even use AI, we’re now doing that for clients, so…
244 00:30:37.280 ⇒ 00:30:54.340 Uttam Kumaran: they… everything from, like, classic, like, go-to-market engineering stuff, like, Clay, more enrichment, like, browser-based for, like, automating, going to browsers and getting info, all the way down to, yeah, like, we build RAG systems using N8N, using TypeScript,
245 00:30:54.640 ⇒ 00:31:03.650 Uttam Kumaran: And, yeah, kind of, like, in the same mode. There’s not… I don’t think there’s a great answer. I don’t have as specific as an answer on, like, what’s right. It’s sort of, like, what…
246 00:31:04.000 ⇒ 00:31:08.060 Uttam Kumaran: what is working and what you guys have the talent for, like, that’s honestly what I found.
247 00:31:08.260 ⇒ 00:31:08.990 Abhi Yadav: Yeah.
248 00:31:08.990 ⇒ 00:31:15.199 Uttam Kumaran: It’s… there are some really great frameworks, but a lot of them, like, you need to be able to actually implement them in production.
249 00:31:15.380 ⇒ 00:31:32.770 Uttam Kumaran: And so we use, we’ve used them all as well. I mean, we also do a lot on the observability side for AI. So one of our big edges that we do a lot of evals, we do a lot of monitoring observability, and we measure, like, how our agents actually answer better over time.
250 00:31:32.770 ⇒ 00:31:38.760 Uttam Kumaran: And most of our use cases for customers are internal use cases. So, we have done some, like, customer-facing
251 00:31:39.100 ⇒ 00:31:55.139 Uttam Kumaran: chatbots, but, like, most of the stuff we do now are all internal automations. So, moving documents from one place to another, changing formats. A lot of it is actually mostly data engineering, like, it kind of… it’s not as, like, AI-heavy as, like, I think people think it is.
252 00:31:55.780 ⇒ 00:32:09.309 Uttam Kumaran: But, like, getting closer to, like, a deterministic output or something that you can rely on is sort of the challenge, and, like, picking the best way to consume structured, unstructured PDF images, like.
253 00:32:09.640 ⇒ 00:32:13.899 Uttam Kumaran: That’s sort of, like, the stuff that we’re building for clients as well.
254 00:32:14.460 ⇒ 00:32:18.249 Abhi Yadav: That’s awesome. And this is good information. Love to, like.
255 00:32:18.580 ⇒ 00:32:35.300 Abhi Yadav: you know, explore some of the other things. So, as a next step, and this was very good conversation, thanks, you know, Robert, for fixing it up. So, what I can do is, I can send you, like, this,
256 00:32:35.510 ⇒ 00:32:46.489 Abhi Yadav: a more detailed, like, two-pager on this immediate project, which is what I’m talking about, the Marketplace launch, seems like you have some familiarity already. And…
257 00:32:46.720 ⇒ 00:32:57.060 Abhi Yadav: It would be nice if you could suggest, like, what is your, like, proposal or suggestion, you know, sort of…
258 00:32:57.200 ⇒ 00:33:00.759 Abhi Yadav: walk, walk, run kind of an approach,
259 00:33:01.170 ⇒ 00:33:08.340 Abhi Yadav: We, we just need to make sure that we work on Like, one project…
260 00:33:08.700 ⇒ 00:33:18.129 Abhi Yadav: Quick and easy and fast, and then we can always, you know, double down, triple down on some of the other things we’re doing.
261 00:33:18.290 ⇒ 00:33:36.599 Abhi Yadav: We are also lean bandwidth, so most of our dependency is, like, you know, for folks like you, I mean, that’s usually the plan, even when we will ramp up. So they’re, like, a core team of engineers we have, and then, you know, everything were, so we can even, like.
262 00:33:36.990 ⇒ 00:33:53.030 Abhi Yadav: rely completely on, sort of, one piece of our platform. The platform is used in, like, different, what we call, like, module or hubs. So, this Snowflake and this whole data piece would be an interesting, kind of.
263 00:33:53.210 ⇒ 00:33:55.739 Abhi Yadav: Reliance, there.
264 00:33:57.040 ⇒ 00:33:57.530 Uttam Kumaran: Cool.
265 00:33:57.530 ⇒ 00:34:04.159 Abhi Yadav: Cool. Any other… Question… From you guys, Jithin, you have…
266 00:34:07.760 ⇒ 00:34:25.960 Jiten Kumar: No, I think we covered almost everything, but yeah, later on, I would be really interested to know, like, might be the pre-QL or any other use case, what was the journey, what the tech stack, and over the period, what tech stack you guys have changed, and all that, so I would be happy to hear about that.
267 00:34:26.820 ⇒ 00:34:29.340 Uttam Kumaran: Sure, yeah, Robert, we can include some of our
268 00:34:29.500 ⇒ 00:34:37.860 Uttam Kumaran: One of the things that’s fun about this business, we see so many different tools, tools that I thought a lot of companies have stopped using by now.
269 00:34:38.090 ⇒ 00:34:38.840 Jiten Kumar: So…
270 00:34:39.639 ⇒ 00:34:52.969 Uttam Kumaran: We are very opinionated about, like, what we hope clients choose, but we’re also, like, we roll with the punches. And early on in the company, we decided not to take, like, kickbacks from vendors.
271 00:34:52.969 ⇒ 00:35:05.380 Uttam Kumaran: So, we just recommend, like, usually what’s… what’s easiest for us to develop is also the best for the client, and so, yeah, I’m happy to share, sort of, like, what we did at Prequel.
272 00:35:05.380 ⇒ 00:35:13.059 Uttam Kumaran: In terms of tech stack, and then kind of, like, what we’re… what we’re seeing in the market across the entire, you know, sort of data stack, so…
273 00:35:13.170 ⇒ 00:35:24.840 Uttam Kumaran: I think, yeah, I’m excited to sort of get the two-pager. It’s a really, really interesting product. Just was gonna… was gonna poke a bit more around the website, and then kind of, like, yeah, just see if I have any further questions.
274 00:35:25.120 ⇒ 00:35:35.359 Abhi Yadav: Yeah, no, absolutely. Thank you. So, let me send this to you today, Sal, Robert, in probably a couple of hours, and then, sure.
275 00:35:35.820 ⇒ 00:35:39.970 Robert Tseng: We’ve got our emails here from this call, too, so if you… whatever you prefer, you can send it.
276 00:35:39.970 ⇒ 00:35:48.980 Abhi Yadav: Right here, and then you let us know, like, some sort of an approach note, or whatever you suggest.
277 00:35:49.410 ⇒ 00:35:50.070 Uttam Kumaran: Check.
278 00:35:50.910 ⇒ 00:35:53.149 Abhi Yadav: Awesome. Nice meeting you both.
279 00:35:53.150 ⇒ 00:35:54.330 Uttam Kumaran: Yeah, thank you.
280 00:35:54.340 ⇒ 00:35:55.540 Abhi Yadav: How are you doing?
281 00:35:55.540 ⇒ 00:35:56.400 Jiten Kumar: I’m being vegetative.