Meeting Title: Brain Forge x Addison Partnership Discussion Date: 2026-05-08 Meeting participants: Olivia Natasha, Robert Tseng
WEBVTT
1 00:03:17.630 ⇒ 00:03:18.820 Olivia Natasha: Hello!
2 00:03:19.790 ⇒ 00:03:20.730 Robert Tseng: Hey, Olivia.
3 00:03:21.140 ⇒ 00:03:22.029 Olivia Natasha: How are you doing?
4 00:03:22.570 ⇒ 00:03:23.530 Robert Tseng: Good, how are you?
5 00:03:24.950 ⇒ 00:03:25.500 Olivia Natasha: Good.
6 00:03:27.440 ⇒ 00:03:32.870 Robert Tseng: Thanks for taking… thanks for taking this call. I’m assuming… are you… are you in Indonesia right now?
7 00:03:33.350 ⇒ 00:03:34.489 Olivia Natasha: Yes, I am.
8 00:03:34.730 ⇒ 00:03:38.299 Robert Tseng: Okay, so you typically work… and what kind of hours do you work?
9 00:03:39.350 ⇒ 00:03:43.259 Olivia Natasha: It depends on the client. If my clients are American-based, I work…
10 00:03:43.760 ⇒ 00:03:51.370 Olivia Natasha: quite late. I think if it’s Eastern Time, I work up to, like, 11 p.m, maybe? Give or take?
11 00:03:51.550 ⇒ 00:03:58.579 Olivia Natasha: Okay. But I start my day later, of course. But yeah, but it’s Friday, so I usually don’t work as late in Fridays.
12 00:03:59.160 ⇒ 00:03:59.760 Robert Tseng: Okay.
13 00:04:00.010 ⇒ 00:04:05.249 Robert Tseng: I just wanted to make sure I wasn’t, like, keeping you up extra than… more than you normally are.
14 00:04:05.590 ⇒ 00:04:08.270 Olivia Natasha: It’s 7pm now, so don’t worry about it.
15 00:04:08.270 ⇒ 00:04:09.500 Robert Tseng: Okay, cool.
16 00:04:10.130 ⇒ 00:04:20.909 Robert Tseng: Yeah, well, I mean, I… yeah, I noticed… appreciate the… appreciate the time anyway, and you know, I reached out to you over a couple things that stuck out. I think,
17 00:04:21.320 ⇒ 00:04:25.049 Robert Tseng: How do you pronounce… is it Adesight? Adasite? I don’t know how you pronounce your company.
18 00:04:25.050 ⇒ 00:04:26.980 Olivia Natasha: anna’s fine.
19 00:04:27.310 ⇒ 00:04:35.589 Robert Tseng: A site, okay. Yeah, I mean, have… have seen them… have seen your firm kind of, like, you know, pretty active in the Amplitude community, so curious about…
20 00:04:35.720 ⇒ 00:04:51.159 Robert Tseng: kind of that, as you guys are testing out the AI features there, since, product analytics is not our main, work stream, but it is something that we do a little bit of as well, or… and maybe there’s an opportunity to, you know, if you guys are experts, then we can maybe partner in that way.
21 00:04:51.160 ⇒ 00:05:08.440 Robert Tseng: And then second, I’m kind of more broadly interested in kind of your experience overall there being a growth consultant, kind of how things have changed, since I run a consulting firm as well, so would be curious to learn more about, kind of, what other folks in the industry are seeing as well.
22 00:05:09.500 ⇒ 00:05:10.110 Olivia Natasha: Okay.
23 00:05:11.270 ⇒ 00:05:16.500 Robert Tseng: Yeah, so, I don’t know if you wanted to start… we’ll start with a little bit of an intro, I can…
24 00:05:16.910 ⇒ 00:05:25.460 Robert Tseng: briefly mention a little bit more about myself, and then I’ll turn it over to you. Yeah, I apologize, I’m, like, I’m, like, muting my voice a bit, because…
25 00:05:25.770 ⇒ 00:05:42.669 Robert Tseng: I’m in, like, a building, trying to look for… trying to look for a meeting room, but I’m, like, kind of just in a more public lounge right now, so trying not to disturb people that are in other spaces. But I live in New York, I’ve been running Brain Forge for 2 years. Before that was,
26 00:05:42.760 ⇒ 00:05:58.409 Robert Tseng: kind of running my own agency as well, so I merged with, a different agency, so we basically offer what we call ourselves context engineers, so, focusing on, organization… helping organizations build their company brain.
27 00:05:58.770 ⇒ 00:06:09.210 Robert Tseng: doing everything from the ground up, data warehouse builds. We’re a Snowflake partner, we’re a Google partner, so helping organizations land all their data in a data warehouse.
28 00:06:09.220 ⇒ 00:06:18.879 Robert Tseng: connecting it to, MarTech tools, CDPs, CEPs, and then, you know, product analytics as well, and then… and then we also build out the BI reporting layer.
29 00:06:18.880 ⇒ 00:06:39.409 Robert Tseng: But obviously, in the past 6 months, people are a lot more interested in being able to just hook up agents to their data directly, so I think that’s been most of our work these past 6 months, and so things continue to evolve, and so I’m always trying to keep up with the latest of experts of the tools that we already use, and also things that we should be looking out for.
30 00:06:40.410 ⇒ 00:06:51.900 Olivia Natasha: Cool. I can go next. I’m a growth consultant here in Addison. So, previously, I was a business analyst, data scientist, so I do, like, end-to-end analytics. Yeah.
31 00:06:52.140 ⇒ 00:06:59.549 Olivia Natasha: So, in my previous role, right, I do exactly what you’re doing. Basically, client has, like, 10 data stacks.
32 00:06:59.550 ⇒ 00:07:18.199 Olivia Natasha: you’re like, oh, I don’t know where my data should go, I don’t… I want them to be in, like, one place. So we sync that together in one place, we built the dashboards, we design the tracking plan, we design everything, so that, our client will have one layer, just, like, one place to look at their data and can make decisions from there, right? Yeah.
33 00:07:18.280 ⇒ 00:07:22.310 Olivia Natasha: So it’s currently pretty similar to what I do now in Out of Sight.
34 00:07:22.610 ⇒ 00:07:30.410 Olivia Natasha: But the difference is that it depends on the client’s tech stack, right? So whatever a client’s using, I try to make sure that, yes.
35 00:07:30.410 ⇒ 00:07:48.030 Olivia Natasha: let’s work… let’s find a workaround so that we can sync the data together in one place. A lot of our clients, because they are not really big, they try to find somewhere that is an easy, like, place to hook up all their data, and a lot of the clients, they land on
36 00:07:48.030 ⇒ 00:07:48.640 Olivia Natasha: you know.
37 00:07:48.640 ⇒ 00:08:01.249 Olivia Natasha: tools like Amplitude, or like Mixpanel, or any other product analytics. Most of the clients itself, they are working towards, like, oh, I want to figure out what people are doing there, right? Is my marketing working?
38 00:08:01.470 ⇒ 00:08:10.960 Olivia Natasha: how is my financial data looking? Can I combine my financials with, like, my marketing data? So, the easiest way to do this would be through
39 00:08:11.350 ⇒ 00:08:15.989 Olivia Natasha: product analytics tools like Amplitude, right? Because it has everything that has your
40 00:08:16.220 ⇒ 00:08:23.890 Olivia Natasha: it has your events there, it has a marketing data that you can hook up to, so that’s why people are doing that. And now…
41 00:08:24.080 ⇒ 00:08:31.169 Olivia Natasha: with, Amplitude having AI, it’s just much easier for a lot of people to start from, like, zero.
42 00:08:32.220 ⇒ 00:08:46.889 Robert Tseng: Okay, yeah, no, I mean, I’m familiar with some of those capabilities. I have a couple questions about that. It’s interesting that you hook up marketing, product, and financial data all in one place, and amplitude for a lot of your clients. At least, what we found was that, like, for financial data, especially since
43 00:08:47.990 ⇒ 00:09:03.000 Robert Tseng: the people who are using it, typically finance, I mean, Amplitude has pretty limited features in terms of, like, what it can offer. It’s not Excel, it’s not Sigma, it’s not, like, kind of a spreadsheet-based type of tool, so, like, even, like, tabular functions are quite limited.
44 00:09:03.060 ⇒ 00:09:14.700 Robert Tseng: I mean, obviously they do a good job with any, like, time… time series-based data, but yeah, I think that’s, like, some… that was always kind of a limiting factor for us, that, like, hey, I think finance people just want the data itself.
45 00:09:14.770 ⇒ 00:09:24.959 Robert Tseng: They do want to be able to, kind of, create their own joins, which is always something that there’s, like, there’s tension with, because, yeah, I think it’s really hard…
46 00:09:25.060 ⇒ 00:09:41.310 Robert Tseng: I mean, they would just export CSVs from multiple sources and then start to do stuff themselves. That’s, like, the typical finance way. So, like, obviously having a data team to kind of guide them through that is helpful, but I thought… always found that to be a challenging stakeholder to be, to be using Amplitude.
47 00:09:41.380 ⇒ 00:09:48.179 Robert Tseng: And then on the marketing side, like, found the integrations to be quite limited as well.
48 00:09:48.780 ⇒ 00:10:01.310 Robert Tseng: yeah, there’s, like, not that much functionality to… for reverse ETL, into amplitude, specifically for, like, regulated industries that… where, like, the out of… the connector out of the box is, like, not enough, like.
49 00:10:01.330 ⇒ 00:10:13.310 Robert Tseng: Sometimes you need to strip data out before you push it back into these ad platforms, you need to increase the frequency, or, you know, just different things like that, where it’s kind of hard to steer, like, what
50 00:10:13.450 ⇒ 00:10:27.519 Robert Tseng: kind of what Amplitude has set up with these platforms, so we end up having to, like, build custom, like, functions around… around that already. So, I mean, I know that was, like, really specific, but I’m just curious, have you run into those challenges as well, and kind of what your point of view of that?
51 00:10:27.950 ⇒ 00:10:41.789 Olivia Natasha: Yeah, so the people who approach us are… well, actually, in my previous role, the people who approach us are not really the finance team. It’s never really the finance team, because the finance team would like to look at their data separately, right?
52 00:10:42.130 ⇒ 00:10:42.520 Robert Tseng: Yeah.
53 00:10:42.520 ⇒ 00:11:02.440 Olivia Natasha: for people who are in marketing, or people who are in just the product itself, they want to know if people are actually converting, right? Because they want to optimize for some things. So, if we give them something that is similar to what finance is looking at, that is better than them having to ask finance about it, which is why… Okay.
54 00:11:02.460 ⇒ 00:11:21.239 Olivia Natasha: we connect those to, like, the product analytics. It’s not, per se, like, the actual revenue that you receive. You do have, like, some miscalculations, I would say, right? It depends on how you do your calculations to send over, to Amplitude or Mixpanel or whatever. So it really depends on the client. Most of the time.
55 00:11:21.350 ⇒ 00:11:26.770 Olivia Natasha: I would say in my previous role, right, because I was the one who was doing the ETL and everything,
56 00:11:26.770 ⇒ 00:11:27.410 Robert Tseng: Yeah.
57 00:11:27.410 ⇒ 00:11:36.010 Olivia Natasha: I would say it’s because they needed something that they can act upon, right? Most of the time, what you’re trying to do is not…
58 00:11:36.160 ⇒ 00:11:55.820 Olivia Natasha: for them, it’s not trying to get the real 100% exact same number, but trying to figure out, hey, if I want to act upon this, is this really working for me, and how do I know if it’s really working for me? Then that’s where your data is coming from. Because, like, if you actually talk to people who are in marketing, they most of the time, know GA4.
59 00:11:55.980 ⇒ 00:11:58.740 Olivia Natasha: And I think JFR is really bad.
60 00:11:59.150 ⇒ 00:12:01.400 Olivia Natasha: But that’s my personal opinion.
61 00:12:01.400 ⇒ 00:12:02.000 Robert Tseng: No, I agree.
62 00:12:02.000 ⇒ 00:12:08.160 Olivia Natasha: Like, it’s not like, great, oh my god, I love GA4, you know? I’ve never heard anyone say that.
63 00:12:08.160 ⇒ 00:12:09.000 Robert Tseng: Yeah.
64 00:12:09.000 ⇒ 00:12:27.879 Olivia Natasha: So for me, like, having them in amplitude is actually much better. It’s because people who don’t really know what’s going on, like, in product analytics and everything, they can easily configure what they need in amplitude. So that’s the reason why we try to hook them up there, so that people can get
65 00:12:27.880 ⇒ 00:12:29.770 Olivia Natasha: To where they need to go.
66 00:12:29.770 ⇒ 00:12:44.190 Olivia Natasha: with whatever they’re doing, you know, it’s not… yes, it’s not really like, oh, perfect data, where you exactly know what’s really going on. If you want to do that, obviously, you build your own custom dashboard, and I’ve done that for my previous clients, but this one here is more like, you know.
67 00:12:44.690 ⇒ 00:12:54.750 Olivia Natasha: As long as we can get to where we need to go, that’s better than nothing, and having them in, like, separate places, and, like, distrusting whatever is being displayed to you.
68 00:12:55.880 ⇒ 00:13:12.729 Robert Tseng: Yeah. Okay. No, I think, thanks for clarifying that. Yeah, I mean, you mentioned you were kind of working mostly with smaller organizations now than maybe before, which makes sense, you know, if they want something with a simpler stack, just one tool that they can kind of do all these things for. I’m curious, like, where you feel like you run into
69 00:13:13.240 ⇒ 00:13:29.030 Robert Tseng: I mean, this setup seems like it works to an extent. Maybe once you achieve a certain scale, like, I think it probably doesn’t work. Like, you do need the custom stack, or maybe you disagree. Maybe you feel like you can get pretty far with just relying on, just Amplitude or just MixPanel.
70 00:13:29.030 ⇒ 00:13:37.129 Robert Tseng: So I’m curious, kind of, like, where you feel like the trade-offs are for an organization. When do they need to start considering other… these other… the other stacks?
71 00:13:37.680 ⇒ 00:13:55.390 Olivia Natasha: So, it really depends on the industry, I would say. For example, if you’re just, like, something… if you’re just e-commerce, where you sell something, I don’t… I don’t foresee that you need a much, you know, larger stack, or, like, a much complex stack. Whatever works, if you are…
72 00:13:55.830 ⇒ 00:14:08.580 Olivia Natasha: if you know that certain events are just that way, you don’t have, like, crazy features, you don’t have any of that, I don’t believe you don’t need… I don’t believe you need to migrate to something bigger, right? Unless in a… unless whatever you’re selling
73 00:14:08.580 ⇒ 00:14:16.420 Olivia Natasha: is regulated with, like, HIPAA or, like, GDPR, you know, all of that, then that means you do have to have your own
74 00:14:16.640 ⇒ 00:14:22.920 Olivia Natasha: data warehouse and everything, but I do believe that, to some extent, if you have more features.
75 00:14:22.920 ⇒ 00:14:37.940 Olivia Natasha: and you have advanced items that you want to track, and you want to make sure that everything goes into, like, one place, or you have a really big customer, customer base, I do recommend that you follow the simple rules of thumb, is that you also have a data warehouse.
76 00:14:37.940 ⇒ 00:14:57.540 Olivia Natasha: Not because whatever you have in amplitude might not be adequate, but rather, one day when you do want to move somewhere else, and let’s say you don’t really enjoy whatever you’re seeing anymore in that data stack, you have that already somewhere in a data warehouse that you could just migrate, and you can just build yourself. So.
77 00:14:57.790 ⇒ 00:15:14.860 Olivia Natasha: I would say, when we say… I would say when we… when we look upon, okay, what is our limitation? That goes through, hey, what are we trying to achieve here? What are my people, like, needing to move forward, right? So, yeah, it depends on that, I would say.
78 00:15:15.790 ⇒ 00:15:21.490 Robert Tseng: Okay, got it. No, that’s… that’s helpful. Would you mind sharing, like, an example? I’m so…
79 00:15:21.960 ⇒ 00:15:29.299 Robert Tseng: We work with more, like, mid-sized firms, so if it’s, like, on the e-commerce CPG side, they’re doing, like, over $100 million in revenue per year.
80 00:15:29.310 ⇒ 00:15:44.800 Robert Tseng: And then, yeah, it’s, like, probably sub-billion dollar brands, so something around in that range, hundreds of billion. And then, so it sounds like you have experience with e-com, so I’d be curious, kind of, hearing more, like, an example of, like, where you feel like, if you have, like, a
81 00:15:44.800 ⇒ 00:15:51.770 Robert Tseng: Example of a larger client that you work with, or at least in that… in that… in that range, that you feel like still runs on that stack.
82 00:15:52.420 ⇒ 00:15:58.009 Olivia Natasha: No, if they’re that big, they’re not gonna run on that slacks, to be honest with you, Robert.
83 00:15:58.010 ⇒ 00:15:58.620 Robert Tseng: Okay.
84 00:15:59.080 ⇒ 00:16:08.130 Olivia Natasha: Yeah. Number one is because they have the resources, and when you are that big, usually you have, like, more things that you want to look at, right? Right.
85 00:16:08.260 ⇒ 00:16:29.500 Olivia Natasha: Not just like, oh, I want to see if my marketing is working because I’m selling product ABC, right? If you are that big, usually you do have a data stack, and you also have a data team to work on these, right? Especially right now, if you want to send your data to AI so that you can have, oh, amazing dashboards and everything, you have to polish and clean that data.
86 00:16:29.500 ⇒ 00:16:35.549 Olivia Natasha: Sometimes I notice that when people are using Product Analytics Tool, they have events that are
87 00:16:35.730 ⇒ 00:16:43.599 Olivia Natasha: you don’t understand what those are, and it’s hard to… for everyone to, like, figure out what it is, so you do… you do need someone to, like.
88 00:16:43.600 ⇒ 00:16:56.940 Olivia Natasha: clean it up, layer it out, so that you can ship it over, so that it’s good to go for other people to use. So, yeah, I’ve had clients where they were very insistent that they wanted everything in one place.
89 00:16:58.460 ⇒ 00:16:58.960 Robert Tseng: Yeah.
90 00:16:59.260 ⇒ 00:17:13.809 Olivia Natasha: So we try to make it work, right? Because we think, okay, we see where you’re going, you’re not really a big company at this point, but that’s why you want it in one place, like an amplitude or whatnot. And the result was…
91 00:17:14.319 ⇒ 00:17:33.010 Olivia Natasha: it was very hard to even… for you to be able to say that I trust in that data, because whatever you’re seeing in that platform, right, requires you to figure out how it was cleaned up, how it was displayed, before you can transfer it over to your different locations, and that’s where
92 00:17:33.150 ⇒ 00:17:36.919 Olivia Natasha: Some of these, small to mid-sized firms are failing.
93 00:17:37.030 ⇒ 00:17:50.150 Olivia Natasha: I would say. Yeah. Because they’re insistent that they can do it, but in reality, it takes a lot of maintenance and takes a lot of, you know, checking and things like that for it to actually work, and even then, some of the times that it always breaks.
94 00:17:51.270 ⇒ 00:17:51.830 Robert Tseng: Okay.
95 00:17:52.080 ⇒ 00:17:56.579 Robert Tseng: Yeah, I mean, let’s say instead of that kind of class, it’s, yeah, like, maybe…
96 00:17:56.750 ⇒ 00:18:02.420 Robert Tseng: A company that’s just setting it up for the first time, smaller scale, maybe, like, software or services, probably, like.
97 00:18:02.750 ⇒ 00:18:05.969 Robert Tseng: I don’t know, somewhere between, like, nor… I mean…
98 00:18:06.900 ⇒ 00:18:18.040 Robert Tseng: post Series A, probably, like, at least $5 million in revenue. Like, some much, much smaller organization, but probably more complex product. And, like, I think before they…
99 00:18:18.170 ⇒ 00:18:25.799 Robert Tseng: you know, up to this point, they’ve relied just on the marketing team, using maybe something like a GA4, maybe having some, like,
100 00:18:26.210 ⇒ 00:18:38.589 Robert Tseng: CRM system that they use, that they’ve set up, like HubSpot, or they’re able to basically… I mean, the application still has… is able to track some data, like, they’re able to get… their engineers are, like, blogging, like, some activity on the…
101 00:18:38.740 ⇒ 00:18:40.599 Robert Tseng: app, and it’s, like, kind of just…
102 00:18:40.990 ⇒ 00:18:53.100 Robert Tseng: connected to their CRM. They’re, like, jerry-rigging it to HubSpot and… or Salesforce or something, and, like, so, like, marketing is able to catch some early signals, and they’re also able to define some
103 00:18:53.100 ⇒ 00:19:05.429 Robert Tseng: basic lifecycle stages, but then now they’re, like, feeling like that’s not enough, and it’s too dependent on the engineering team, so they want to actually look into product analytics. I feel like that’s another, like, type of organization that we come into
104 00:19:05.480 ⇒ 00:19:14.649 Robert Tseng: come across often? Like, how would you, like, I don’t know, does that feel familiar to you? And, like, if so, like, how would you kind of steer… how would you steer that?
105 00:19:16.160 ⇒ 00:19:23.919 Olivia Natasha: I would definitely just recommend them have a… to have a product analytics tool, right? Reason is because
106 00:19:23.920 ⇒ 00:19:43.739 Olivia Natasha: regardless of what it is, the product analytics will be helpful for you, to even, like, figure out what your users are doing. Most of the time, when you are asking engineers to do this, yes, I mean, they do track some stuff, right? But it’s not really the kind of analysis that you’ll get if you are, let’s say, hiring a product analyst, right?
107 00:19:44.090 ⇒ 00:19:55.579 Olivia Natasha: Because engineers, they track the most important stuff for them, like, oh, is the application running? What software are people using? You know, like, which bug are we experiencing? But with a…
108 00:19:55.700 ⇒ 00:20:10.990 Olivia Natasha: product analysts, they might want to look at, hey, certain funnels, how are they working? How is people, like, going from point A to point B so that I can improve my product, right, with my PMs and stuff like that? So, I would still recommend most companies to still have product analytics tool.
109 00:20:10.990 ⇒ 00:20:19.830 Olivia Natasha: But if, let’s say, I have had clients whose data are super large, and it’s gonna cost them a lot of money to use stuff like Amplitude.
110 00:20:19.830 ⇒ 00:20:23.619 Olivia Natasha: Right? So, what they do is that they can use,
111 00:20:23.790 ⇒ 00:20:30.999 Olivia Natasha: other tracking technology that is being fed to their data warehouse. And then from there, they have a data engineer who cleans up
112 00:20:31.000 ⇒ 00:20:46.290 Olivia Natasha: yeah, who cleans up the data for them, and then that data runs into, like, this data warehouse where it’s easy for any product analyst to pull that data together and stitch that together. So it really depends also on, like, the company, right, I would say.
113 00:20:46.810 ⇒ 00:20:48.300 Robert Tseng: Yeah Okay.
114 00:20:48.600 ⇒ 00:21:05.539 Robert Tseng: Yeah, maybe kind of, like, shifting gears a bit, like, you know, we were talking lightly about AI, like, how do you feel like AI is, like, kind of impacting the way that you approach these engagements? How has it impacted client expectations, like, you know, and maybe, like, how do you do your implementations now?
115 00:21:07.460 ⇒ 00:21:16.850 Olivia Natasha: the most interesting thing for me would be clients think that because we have AI, everything is just dandy and beautiful, and it’s easy to go.
116 00:21:16.980 ⇒ 00:21:22.270 Olivia Natasha: But I tell people, like, hey, you know, AI learns from what you give them.
117 00:21:22.270 ⇒ 00:21:38.450 Olivia Natasha: Right? So, AI is best when you give it context, and it’s always been like that. It’s like, if you have more contacts for them, they have more things to learn about, then they’re better in terms of what they’re spitting out for you. So, sometimes I’ve had clients like, hey, AI can do this for me, why can’t I just use AI?
118 00:21:38.450 ⇒ 00:21:39.040 Olivia Natasha: But…
119 00:21:39.270 ⇒ 00:21:52.280 Olivia Natasha: I realized that when you want to use, for example, like, visual labeling, or AI has their own, like… now Amplitude is going to launch, like, automated tracking plans, right?
120 00:21:52.460 ⇒ 00:21:58.239 Olivia Natasha: But if you notice, Amplitude is not really removing their partners in terms of
121 00:21:58.380 ⇒ 00:22:13.600 Olivia Natasha: the people who help build tracking plans. Why? Because before you go ahead and do that, there still needs to be some basis of, like, layer of, like, knowledge that Amplitude needs to have before Amplitude can automate your tracking plan, right?
122 00:22:14.070 ⇒ 00:22:16.799 Olivia Natasha: So I would tell people, like, you know.
123 00:22:17.080 ⇒ 00:22:34.019 Olivia Natasha: why do you think they’re not getting rid of people who are helping you build contacts in the beginning? It’s because you have to have these contacts before you’re able to use AI for your benefit. So, that’s what I always tell people. It’s like, yes, AI is going to help you, it has helped me do my work, but
124 00:22:34.030 ⇒ 00:22:42.340 Olivia Natasha: before you go and think that AI is an oracle, it’s, you know, you have to give it some context first, let it learn what you want before you…
125 00:22:42.460 ⇒ 00:22:44.050 Olivia Natasha: Let it run free.
126 00:22:44.810 ⇒ 00:22:45.430 Robert Tseng: Yeah.
127 00:22:46.040 ⇒ 00:22:47.790 Robert Tseng: Okay, yeah, no, that makes sense.
128 00:22:49.420 ⇒ 00:22:55.599 Robert Tseng: Yeah, I mean, and then, I guess, kind of zooming out, thank you for letting me kind of poke around at your
129 00:22:56.050 ⇒ 00:23:12.590 Robert Tseng: experience, like, specifically there. Yeah, I mean, like, what do you… I mean, seems like you’ve been kind of… you’ve been in-house roles, you’ve been consulting as well, and obviously you’re in a… you’re in a more niche role now than you were your previous. Or, like, what’s… what are you looking…
130 00:23:12.590 ⇒ 00:23:20.919 Robert Tseng: for in the next, like, few years, like, how do you… how do you feel like your career trajectory is, like, what do you want to see? And, yeah, like.
131 00:23:22.390 ⇒ 00:23:22.869 Olivia Natasha: This is a…
132 00:23:22.870 ⇒ 00:23:25.519 Robert Tseng: I’ll keep it open-ended. Yeah.
133 00:23:25.520 ⇒ 00:23:39.040 Olivia Natasha: This is a very good question, because I’ve been thinking about this, actually. So, personally, I still like doing the analysis, which is why, I’ve been asking to do, like, more data-related roles. So, you know.
134 00:23:39.280 ⇒ 00:23:53.499 Olivia Natasha: going back to my basics, my education, learning, again, like, you know, helping people to stitch data together to eventually get to where they want to be. Because, for me, the most important thing is that, yes, AI can
135 00:23:53.900 ⇒ 00:24:11.639 Olivia Natasha: do a lot of the things right now. I can even do analysis, but really, what matters most is how you act upon those analysis, which is why, right now, and this is sort of like a in-the-middle role for me, which is great, I get to do the analysis, I get to work on data, I get to find
136 00:24:11.640 ⇒ 00:24:17.920 Olivia Natasha: things that are interesting, and find patterns and everything, but at the same time now, because I’m a consultant, I can help them
137 00:24:18.060 ⇒ 00:24:28.510 Olivia Natasha: figure out, hey, which area should you go? What things… what ideation should we do for you, for us to help you move forward, right? So…
138 00:24:28.530 ⇒ 00:24:40.799 Olivia Natasha: I think I… I like what I do now. I would like to do more roles like this, right? So, basically just, like, gathering all of data together, and then help clients manage
139 00:24:41.220 ⇒ 00:24:43.739 Olivia Natasha: How they proceed in the future, yeah.
140 00:24:44.450 ⇒ 00:24:45.050 Robert Tseng: Yeah.
141 00:24:45.330 ⇒ 00:24:46.620 Robert Tseng: Okay, cool.
142 00:24:46.800 ⇒ 00:24:51.099 Robert Tseng: Yeah, I mean, I feel like I… I mean, I’m happy to…
143 00:24:51.200 ⇒ 00:25:02.239 Robert Tseng: open myself up, and if you have any questions, anything you want to talk about, like, want to also give you space to kind of ask, and we can discuss anything, but I think those were the main… main questions that I had on my side, yeah.
144 00:25:02.910 ⇒ 00:25:06.200 Olivia Natasha: Okay, I do have a question. Sure.
145 00:25:06.360 ⇒ 00:25:15.080 Olivia Natasha: I know you’ve dealt with, like, bigger companies, right? Yeah. So, what would be, like, your biggest challenge, and you’re like, oh, you know what, I never thought about it that way.
146 00:25:16.560 ⇒ 00:25:21.070 Robert Tseng: Biggest challenge with our clients, or kind of, like, in what domain do you feel like you’re interested.
147 00:25:21.650 ⇒ 00:25:22.090 Robert Tseng: Yeah.
148 00:25:22.090 ⇒ 00:25:28.809 Olivia Natasha: just with your clients in general, and then once you’ve figured out a solution, you’re like, oh, I never thought about doing it that way, you know?
149 00:25:29.420 ⇒ 00:25:33.810 Robert Tseng: Oh, I see. Okay. Hmm.
150 00:25:36.830 ⇒ 00:25:48.129 Robert Tseng: Yeah, I mean, I think there’s a lot of those moments right now. I think AI is really changing the way that we’re working. So, yeah, here’s, like, I’ll give an example. Larger, like.
151 00:25:48.330 ⇒ 00:25:55.150 Robert Tseng: GLP-1 company doing, like, north of, like, $200 million a year in revenue. First, we started them.
152 00:25:55.560 ⇒ 00:26:06.979 Robert Tseng: within Mixpanel, and that was enough just to get them started. They grew a lot. They kind of… this was back when they were less than $100 million. They grew past that, they started kind of building out other departments.
153 00:26:07.160 ⇒ 00:26:18.309 Robert Tseng: We really built a data function from scratch there, so, like, trying to see… yeah, just being able to go through more and more evolutions of, like, okay, they actually want,
154 00:26:18.450 ⇒ 00:26:33.850 Robert Tseng: they’re actually building their own software now, and they, like, need to actually have a data… we have to build that software for them, we have to build a data warehouse. So yeah, I mean, just kind of giving… there’s a few milestones that they went through. I think now, the past, like, 5 months have been challenging, because,
155 00:26:34.030 ⇒ 00:26:35.989 Robert Tseng: You know, it’s like, kind of, we’ve…
156 00:26:36.360 ⇒ 00:26:46.459 Robert Tseng: I… yeah, we built… we built out the stack, but I think, like, the… sorry, the camera’s moving around. The… the stack is… it’s changing quite a bit. I think that, it’s…
157 00:26:46.630 ⇒ 00:27:03.369 Robert Tseng: the traditional, like, rigid stack, where, like, things have to be built up a certain way, and, like, you have to, like, control the… the syncing, you have to… roles access permissions, like, I think all of that has kind of, like, fallen. People just expect that they can just plug in whatever agent tool they’re using now.
158 00:27:03.410 ⇒ 00:27:06.709 Robert Tseng: And, they should be able to ask questions about
159 00:27:07.550 ⇒ 00:27:20.460 Robert Tseng: all of their data from whatever tool they’re using, whether that’s just, like, plugging and clawed, or if they’re using even, like, some sort of MCP on their, like, project management tool, whether it’s, like, using Linear or something, and they expect that, like.
160 00:27:20.460 ⇒ 00:27:30.380 Robert Tseng: their tool… whatever tool they’re using, whatever surface, or is what I call it, like, has access to all their data. And so, I think the role has shifted a lot more to
161 00:27:30.450 ⇒ 00:27:36.849 Robert Tseng: Like, data governance, and, like, being like, okay, well, we know that the client is just going to be
162 00:27:37.570 ⇒ 00:27:42.630 Robert Tseng: doing whatever, whatever they want with whatever tool. We just have to make sure that, like,
163 00:27:45.690 ⇒ 00:27:57.109 Robert Tseng: Whatever they… like, how do we set up the guardrails so that when they’re pushing changes, it’s not actually impacting the managed production kind of, like, pipeline that we have set up?
164 00:27:57.280 ⇒ 00:28:14.399 Robert Tseng: and… but then also, like, we’re able to, understand, like, what they’re asking of the system, so that we can, like, pick up ideas from how… from their actual usage, and, like, take… use that as ideas to continue building our roadmap. So, I think, like, in short, it’s, like.
165 00:28:14.620 ⇒ 00:28:16.550 Robert Tseng: Every…
166 00:28:16.630 ⇒ 00:28:30.719 Robert Tseng: everything’s become a data product. There’s event tracking across everything, because, like, everybody is using, like, agent… we can… not every tool lets us monitor, like, the questions that are being asked, but, like, there’s a lot of user intent that’s being captured across everywhere.
167 00:28:30.720 ⇒ 00:28:39.300 Robert Tseng: And people, like, are, like, kind of blurring the lines between, like, I don’t want to log into this tool for this, this tool for that, like, I want to just have one thing that I use for everything.
168 00:28:39.340 ⇒ 00:28:46.260 Robert Tseng: So I think being, like, managing that has been, like, the biggest challenge the past, like, 5 months, yeah.
169 00:28:47.440 ⇒ 00:28:55.620 Olivia Natasha: how do you manage that, you know? Because I know things are going to come up and be like, I know AI can do this, like, why can’t you do that for me, right?
170 00:28:55.870 ⇒ 00:28:59.660 Robert Tseng: Yeah, yeah. I think, like…
171 00:28:59.810 ⇒ 00:29:04.590 Robert Tseng: So, there’s, like, two approaches that we’re taking. One is,
172 00:29:06.960 ⇒ 00:29:22.959 Robert Tseng: like, we actually have to do some of the, we’re setting up embeddings, like, doing RAG, like, doing a lot of this AI engineering work, where we’re, like, actually restricting context. So, but we’re having to do that for, like, a limited set of users. So, for example.
173 00:29:22.960 ⇒ 00:29:29.889 Robert Tseng: Yeah, now the C-suite… I mean, there’s a team of, like, 400 people now, so, like, they need to be able to,
174 00:29:30.100 ⇒ 00:29:32.270 Robert Tseng: Understand, like.
175 00:29:32.270 ⇒ 00:29:51.759 Robert Tseng: if they just rely on Gemini, or they’re using Claude hooked up to Google Drive and Slack, like, the quality’s not very good. And so, knowing, like, how to, like, work… having… having folks just working on the… on the context engineering around that, so that when the executives are using their AI tool connected to this, like.
176 00:29:51.760 ⇒ 00:29:57.709 Robert Tseng: Controlled environment that we have. Like, the quality of the data that they’re trying to get out of
177 00:29:57.710 ⇒ 00:30:13.550 Robert Tseng: the, out of this system is better. So, I don’t think it’s possible to do that for every user, because, like, everybody wants different things, so this is, like, really, like, just, like, curated to just this audience. But I do think that’s, like, one part that helps.
178 00:30:13.600 ⇒ 00:30:19.009 Robert Tseng: And then I think the other approach is more, around, like.
179 00:30:19.790 ⇒ 00:30:23.550 Robert Tseng: All of the… we’re kind of redirecting,
180 00:30:25.540 ⇒ 00:30:40.950 Robert Tseng: user requests. I mean, they’re coming in the form less of, like, requests, like, we’re not taking tickets anymore. What we’re finding is, like, yeah, like I said, usage within the product, or people are trying to make changes themselves, like, everybody thinks that they can code now, so they’re, like.
181 00:30:41.130 ⇒ 00:31:00.010 Robert Tseng: we’re, like, funneling all of that activity into, like, a separate repo that’s, like, a staging environment now. And then, like, within, like, our managed pipeline, we have, like, a production repo as well. We’re, like, sending, like, AI agents into the staging repo and, like, trying to, like, curate
182 00:31:00.010 ⇒ 00:31:14.549 Robert Tseng: ideas from them, from that side, and anything that’s, like, that could be a breaking change to, like, to our system, like, that’s something that we have to, like, surface very quickly, and then we go through that more urgently.
183 00:31:14.550 ⇒ 00:31:25.599 Robert Tseng: But then there’s also, like, other ideas where it’s like, oh, that’s interesting, like, didn’t know that you, like, somebody made some sort of enrichment, to,
184 00:31:25.600 ⇒ 00:31:41.309 Robert Tseng: like, segment data in a way that, like, we didn’t have currently. It’s not a breaking change. It’s an interesting idea. It’s… there’s, like, a clear, like, data diff between what we have versus what they propose. That’s… we should just take that into our roadmap. So, I think, like, it’s kind of hard to…
185 00:31:41.590 ⇒ 00:31:47.110 Robert Tseng: I mean, I think this is still very much a work in progress, but I’m excited by it, because I think it’s very much…
186 00:31:47.750 ⇒ 00:32:03.269 Robert Tseng: how a lot of companies will operate in the future, where, yeah, I think, like, people… I mean, users are less patient, and so they’re going to just try to push things, and, like, I think the data team’s responsibility is now to,
187 00:32:03.410 ⇒ 00:32:07.659 Robert Tseng: Like, yeah, really be more of the, like,
188 00:32:07.830 ⇒ 00:32:12.130 Robert Tseng: It’s kind of gatekeeper of, like, what… what comes…
189 00:32:13.310 ⇒ 00:32:38.089 Robert Tseng: you’re just, like, taking these ideas and, like, trying to, like, actually, to harden them, like, allowing people to contribute more, rather than it being, like, reliant on just, like, a class of, like, analysts and engineers to go and, like, do all the feature development. Like, the business users have a lot of great ideas. I mean, it’s kind of chaotic when they’re using AI to go and try to implement it. So, like, there’s a lot of noise, but, like, I think our
190 00:32:38.090 ⇒ 00:32:40.570 Robert Tseng: My role now is, like, a lot more trying to
191 00:32:40.570 ⇒ 00:32:46.709 Robert Tseng: like, figure out what’s valuable from, like, all the noise that we’re getting from the clients. So,
192 00:32:47.550 ⇒ 00:32:58.429 Robert Tseng: Sorry if I wasn’t very concise, but, like, it’s a good question. I’m trying to, like, articulate it better in front of people as well. But yeah, I would say that that’s, like, kind of what we’re seeing now.
193 00:32:59.240 ⇒ 00:33:12.239 Olivia Natasha: Just, like, a clarification there, in the beginning, you mentioned that you were, you know, helping people set this up, and then you eventually, like, helped them with creating guardrails and stuff like that, so do you build the products for them, or…
194 00:33:13.280 ⇒ 00:33:13.630 Robert Tseng: Yeah.
195 00:33:13.630 ⇒ 00:33:16.849 Olivia Natasha: They connect it, and then you guide them to do it.
196 00:33:17.630 ⇒ 00:33:34.799 Robert Tseng: Yeah, I mean, in this case, we did actually build, like, their product for them. They were using an e-comm tool off the shelf, but we basically, like, kind of replaced that. I would not recommend… I would not do that again. It’s just too much… it’s too much over… overhead for my team to maintain, so… But yes, typically we’re just…
197 00:33:34.870 ⇒ 00:33:44.999 Robert Tseng: having to… I mean, there’s an expectation that whatever tool somebody is using, we can build a connector for it, or, like, we’ll just know how to get data out of it. So…
198 00:33:45.670 ⇒ 00:33:54.360 Robert Tseng: Yeah, I mean, sometimes we rely on an ETL tool. I think most of the time we do, so whether it’s, like, a 5tran or a Polytomic or something,
199 00:33:54.360 ⇒ 00:34:09.269 Robert Tseng: But sometimes we’re not able to, and so we have to actually build the custom… the custom integration, and it’s less… it’s less about creating custom API endpoints now, it’s more just about creating an MCP that can hook up to it, in which case, I think it’s a lot easier to set up an MCP
200 00:34:09.270 ⇒ 00:34:16.329 Robert Tseng: than an API, but it’s obviously less reliable, so I think there’s, like, trade-offs that you… that you make. But yeah.
201 00:34:16.520 ⇒ 00:34:18.869 Olivia Natasha: Yeah, I understand what you’re… what you mean.
202 00:34:19.290 ⇒ 00:34:23.959 Olivia Natasha: Yeah. Okay, I think that’s all the questions that I have. Very interesting. Yeah.
203 00:34:25.230 ⇒ 00:34:28.179 Robert Tseng: Yeah, no, I mean, I think it’s an interesting time, I think,
204 00:34:28.820 ⇒ 00:34:33.090 Robert Tseng: I don’t know what… how these data tools will,
205 00:34:33.659 ⇒ 00:34:38.399 Robert Tseng: coexist in the next few years, I feel like it’s gonna be more and more consolidation.
206 00:34:38.650 ⇒ 00:34:45.659 Robert Tseng: I think, you know, I know we’re a little over time, so I’ll wrap it up, but, like, the, one thing you, like, briefly mentioned was…
207 00:34:45.790 ⇒ 00:34:50.190 Robert Tseng: Yeah, even if they don’t use Amplitude or Mixpanel, like, they may rely on some other, like.
208 00:34:50.670 ⇒ 00:35:08.350 Robert Tseng: way to, like, start tracking event data. I think that already happens, like, we’re… we’re a Snowflake partner, Snowflake has Snowpipe, and Snow… it’s basically event… their event tracking layer, and I think, like, I’m sure Databricks does something similar, so it’s like, everyone is now, like, all these other
209 00:35:09.960 ⇒ 00:35:29.870 Robert Tseng: or many other tools, I think, are, like, having the same functionality in terms of, like, the level of tracking. I think Amplitude is, I think, way ahead in terms of visualization and depth of analysis, but it requires a lot of, like, human, like, kind of steering in order to get it, right? And I think people are more interested, or, like, becoming more interested in trying to, like.
210 00:35:30.440 ⇒ 00:35:47.939 Robert Tseng: like, use code and agents to try to, like, pull what they want out of the data, rather than, like, taking the time to set up all the reports and dashboards that, like, maybe previously, like, a product analytics team would do. So, I’m, like, very interested in seeing, like, kind of how
211 00:35:47.940 ⇒ 00:35:54.220 Robert Tseng: that trend, kind of, or how that continues. Maybe that’ll stop, and people will realize, like.
212 00:35:54.380 ⇒ 00:36:01.479 Robert Tseng: this is not going anywhere, it’s not really trustworthy, but, I mean, also maybe not. So, I’m curious to see what happens.
213 00:36:01.900 ⇒ 00:36:14.919 Olivia Natasha: Yeah, which slide are you leaning on? Are you, are you more traditionalist, where you’re like, okay, I still believe human touch is still important, or you’re on the, I believe AI would be able to do this themselves?
214 00:36:15.260 ⇒ 00:36:21.280 Robert Tseng: Yeah, I mean, I feel like I have to believe that AI’s gonna do it. I think,
215 00:36:21.420 ⇒ 00:36:22.909 Robert Tseng: Yeah, like, I…
216 00:36:23.860 ⇒ 00:36:38.429 Robert Tseng: Yeah, I mean, I’ve obviously prefer to build it myself. I was a product analyst at some point before. I still think it’s not good enough, but we get paid to try to make the AI work, so I think that’s… so that’s, that’s what we’re trying most of the time.
217 00:36:38.670 ⇒ 00:36:39.360 Robert Tseng: Yeah.
218 00:36:40.750 ⇒ 00:36:55.310 Robert Tseng: Cool. Well, appreciate the time. Yeah, this was… this was great hearing about your experience. Love to stay in touch, keep you up… I’ll keep you updated on what we’re… what we’re up to, and who knows, maybe there’s an opportunity to collaborate on in the future.
219 00:36:56.130 ⇒ 00:36:56.740 Robert Tseng: Right? Yeah.
220 00:36:56.740 ⇒ 00:36:57.479 Olivia Natasha: Thank you so much.
221 00:36:57.670 ⇒ 00:36:59.360 Robert Tseng: Thanks, Olivia. Take care.