Meeting Title: Henry <> Robert Interview Date: 2025-06-17 Meeting participants: Hannah Wang, Henry Zhao, Robert Tseng
WEBVTT
1 00:02:20.430 ⇒ 00:02:21.110 Hannah Wang: Hello!
2 00:02:22.550 ⇒ 00:02:23.870 Henry Zhao: Hi, Hannah! How are you?
3 00:02:23.870 ⇒ 00:02:25.019 Hannah Wang: Good! How are you?
4 00:02:25.420 ⇒ 00:02:26.030 Henry Zhao: Thanks.
5 00:02:27.740 ⇒ 00:02:33.189 Hannah Wang: Are you in New York? Are you on Eastern? I’m assuming you’re in New York, Eastern.
6 00:02:33.497 ⇒ 00:02:35.650 Henry Zhao: No, I’m in Brazil right now, actually.
7 00:02:35.760 ⇒ 00:02:37.640 Hannah Wang: You’re in Brazil, you said.
8 00:02:37.640 ⇒ 00:02:39.030 Henry Zhao: Got it for the week. Yeah.
9 00:02:39.030 ⇒ 00:02:42.430 Hannah Wang: Oh, wow! For work, or for fun, or.
10 00:02:42.430 ⇒ 00:02:46.180 Henry Zhao: No, I have an apartment here. I I used to live here a a few years. So yeah.
11 00:02:46.180 ⇒ 00:02:50.539 Hannah Wang: Oh, wow! That’s wild. So what time is it there right now?
12 00:02:50.540 ⇒ 00:02:51.920 Henry Zhao: It’s 1 Pm.
13 00:02:52.270 ⇒ 00:02:55.500 Hannah Wang: Okay, that’s not terrible. Cause, I know. Like.
14 00:02:55.500 ⇒ 00:02:56.169 Henry Zhao: Another call.
15 00:02:56.370 ⇒ 00:03:03.459 Hannah Wang: Everywhere around the world is so different. So it’s just like interesting. Time zone. Wise. Okay, cool.
16 00:03:03.840 ⇒ 00:03:04.980 Henry Zhao: On the weekend.
17 00:03:05.240 ⇒ 00:03:07.255 Hannah Wang: I’m in pacific time.
18 00:03:07.760 ⇒ 00:03:08.100 Henry Zhao: Okay.
19 00:03:08.100 ⇒ 00:03:16.980 Hannah Wang: I’m in la right now. Yes, I’m like the furthest away from like the rest of the world. So it’s like the biggest
20 00:03:17.120 ⇒ 00:03:28.530 Hannah Wang: time. I guess Hawaii is the biggest time difference, but it’s far from other people. So awesome. Okay.
21 00:03:28.530 ⇒ 00:03:30.809 Robert Tseng: Hey? Sorry. I’m sorry I’m late.
22 00:03:32.980 ⇒ 00:03:59.219 Hannah Wang: Alright. So, yeah, just quickly. I, yeah, this, I mean, Robert will like run the interview, basically as the outline I will go off camera so that well, 1st of all, are you okay? If we like, record this and then post, okay? So I’ll go offline off camera. So I’m not in the screen. But I’ll just be lurking here, making sure everything’s going well. And then, yeah, we’ll take it from there. So go ahead, Robert.
23 00:04:00.110 ⇒ 00:04:08.142 Robert Tseng: Okay. Cool, hey, Henry? It’s been a while, I know. Hope you hope to move, I mean, are you in Arizona? Are you in New York now?
24 00:04:08.620 ⇒ 00:04:10.229 Henry Zhao: Right now I’m in Brazil, actually.
25 00:04:10.630 ⇒ 00:04:16.159 Robert Tseng: Moving back to Brazil. Oh, wow, okay, cool. Yeah. I mean, I guess, like.
26 00:04:16.329 ⇒ 00:04:19.399 Robert Tseng: thanks for doing the agreeing to do this. And
27 00:04:19.930 ⇒ 00:04:28.450 Robert Tseng: kind of all the coordination with Hannah like I I sorry I was a bit late. I don’t know exactly what she said. It’s not like a traditional interview or anything like, I think
28 00:04:28.600 ⇒ 00:04:31.270 Robert Tseng: we’re kind of just doing a series where.
29 00:04:31.340 ⇒ 00:04:50.659 Robert Tseng: yeah, I just meet cool data people and just try to talk with them. Obviously, like, I kind of pick up on some things about people’s background and some topics that I think would be really cool to discuss. And so I kind of want to see it more as a conversation, and hope I didn’t overwhelm you with all the outline questions like, I don’t think we’ll get to everything. I think we’re just kinda
30 00:04:50.660 ⇒ 00:05:01.761 Robert Tseng: and I go go with the flow. But then, yeah, I mean, I think you know 45 min, Max, I think is kind of how what I would expect this kind of take. So
31 00:05:02.640 ⇒ 00:05:13.579 Robert Tseng: yeah, I mean, I mean, I’ll probably do most of the kind of prompting, but like happy to make a 2 way conversation, and don’t really need this to be like, you know. Just me asking you questions, and
32 00:05:14.260 ⇒ 00:05:18.600 Robert Tseng: you talking for 30 min, and I only talk for like 5 min. Kind of thing.
33 00:05:19.790 ⇒ 00:05:20.650 Henry Zhao: That’s right.
34 00:05:20.870 ⇒ 00:05:46.159 Robert Tseng: Cool. Alright. Well, then, let’s let’s jump into it. We don’t need to do any fancy intro. I kind of like record an intro like later on with when you’re not, when you’re not around. So I can talk, talk you up a bit more, because I think before usually too humble to really talk about themselves. In an Intro way. But yeah, I mean, just to kind of bring us back to like, yeah, I mean, what is it like almost a month ago when I met you at the rudder stack event in New York?
35 00:05:46.860 ⇒ 00:06:03.909 Robert Tseng: Yeah, I mean, I think, what really stuck out, and why we followed up was felt like, you know, you you successfully kind of migrated a mid market enterprise like size, company from segment into rudder, stack and kind of like went through all these. You know the whole
36 00:06:04.360 ⇒ 00:06:29.470 Robert Tseng: vendor, you know. This bought purchasing decision and implementation of of like a pretty massive, you know, piece of technology there. So I thought it would be really good to just kind of chat with you. More about that experience. But yeah, I guess before that, I mean, I know you’re in Brazil now, and we didn’t really get to talk too much about kind of like your background. And but yeah, I’d love to learn more about like
37 00:06:29.470 ⇒ 00:06:48.030 Robert Tseng: what it was like spending a good chunk of your career working in data in another country, and then also kind of like what? That? How that change, you know, like kind of the difference between that and working in the States. So wondering if you have any. If you want to kind of share a bit about your experience. There.
38 00:06:48.820 ⇒ 00:07:03.240 Henry Zhao: Yeah. So as a reminder, I 1st moved to Brazil with Facebook. So I was working at Facebook in the Us. And I transferred to Facebook in Brazil, so that transition was pretty easy in terms of, since it is a global company and everyone was required to speak English.
39 00:07:03.807 ⇒ 00:07:10.500 Henry Zhao: There was just some minor cultural differences. But I would say, working at Facebook in Brazil was very similar to working at Facebook in the Us.
40 00:07:10.928 ⇒ 00:07:16.779 Henry Zhao: But eventually I ended up at a Brazilian startup, and I would say, that is a lot more different.
41 00:07:17.369 ⇒ 00:07:26.959 Henry Zhao: But all of the founders were, you know, they went to Upenn they went to Harvard. They spent some time working in the Us. So culturally. I think they also had
42 00:07:27.972 ⇒ 00:07:31.519 Henry Zhao: kind of the same working style and kind of the same
43 00:07:32.399 ⇒ 00:07:42.009 Henry Zhao: decision making when when looking into software. So that’s why we end up with segment, whereas a bunch of different Brazilian solutions came up to us.
44 00:07:42.330 ⇒ 00:07:59.950 Henry Zhao: For example, long, long ago there was this Brazilian Dashboarding company that was, you know, AI. First, st they said, you know, you guys are using tableau and looker, but we are much better because we use Llm. For your team to be able to write a prompt and be able to answer data questions on their own.
45 00:08:00.590 ⇒ 00:08:01.680 Robert Tseng: How long goes on.
46 00:08:01.850 ⇒ 00:08:05.700 Henry Zhao: This was like, 5 years ago. So this is oh, wow!
47 00:08:05.700 ⇒ 00:08:18.380 Henry Zhao: You know, this is at the start of the pandemic. So I would say, maybe maybe 4 years ago. We didn’t like it, because the issue with the company at that point was, people didn’t know what were the right questions to ask.
48 00:08:19.170 ⇒ 00:08:23.260 Henry Zhao: And also the data was very messy. So even if they asked the right questions
49 00:08:24.355 ⇒ 00:08:37.760 Henry Zhao: that solution wasn’t right for us at the time it was much more of a of a need to clean up the data and make sure that people are asking the right questions. So, for example, if they ask how many of our customers came from
50 00:08:37.900 ⇒ 00:08:40.790 Henry Zhao: Facebook, or how many people came from Google?
51 00:08:41.515 ⇒ 00:08:46.533 Henry Zhao: There’s a lot of double counting. Because we set up Utms. But
52 00:08:47.010 ⇒ 00:09:13.010 Henry Zhao: whenever someone clears cookies, or whenever they access something from the browser or from their phone, it would be 2 different IP addresses, right? So accounting people, we were seeing a lot of direct traffic that wasn’t actually direct traffic. So it wasn’t. That Looker wasn’t convenient in that. People needed to be able to ask questions using an Llm model. But we needed to have cleaner data. We needed to direct people to ask the right questions
53 00:09:13.390 ⇒ 00:09:19.759 Henry Zhao: and figure out the insights that would actually drive the business forward. So that was, I think, what I learned from from that experience.
54 00:09:20.120 ⇒ 00:09:22.650 Robert Tseng: Yeah, yeah, I mean, I still think that is the
55 00:09:22.840 ⇒ 00:09:46.259 Robert Tseng: the bare bones that you need in order to do any sort of like Llm prompting on your code base. I mean, now, we’re using tools like, I’m gonna have a whole AI section later in this talk. But, like, you know, we’re using cursor on top of, like, you know, data models that we put together. And it actually works pretty well once you have all the data cleaned up. But if you don’t, then like, if it’s going to make inferences or kind of
56 00:09:46.260 ⇒ 00:09:57.620 Robert Tseng: it’s not good at like kind of layering assumptions on top of it, and like it’s kind of a black box, right? So being able to still like, own that logic and kind of have that
57 00:09:57.680 ⇒ 00:10:04.399 Robert Tseng: more deterministically set up, I think, is still like necessary for any sort of like
58 00:10:05.050 ⇒ 00:10:08.189 Robert Tseng: data activation, using like AI already.
59 00:10:08.610 ⇒ 00:10:09.005 Henry Zhao: Right
60 00:10:09.700 ⇒ 00:10:10.110 Robert Tseng: Yeah.
61 00:10:10.925 ⇒ 00:10:35.159 Robert Tseng: I mean, I’m curious from like a cultural perspective. You felt like the decision making processes, and that really different. Everyone was kind of either from the Us. Kind of brought over, or they were trained in the Us. Or educated in some way. But I imagine, like, you know the teams that you were leading or other people you’re working with may not have had that experience. So I’m curious. If you notice anything kind of worrying about like differences in how people
62 00:10:35.280 ⇒ 00:10:42.819 Robert Tseng: you know approach data or interpreted data like in another culture, I guess specifically, Brazil versus like what you saw in in the States.
63 00:10:43.820 ⇒ 00:10:57.049 Henry Zhao: Yeah, I would say the like. The leadership were a lot of like us people with us. Experience but the marketing team, for example, was mostly Brazilian. So there’s a lot of challenges in terms of communication with the marketing team
64 00:10:57.620 ⇒ 00:11:02.679 Henry Zhao: like explaining why we need Utms explaining why the data is not clean.
65 00:11:04.560 ⇒ 00:11:09.039 Henry Zhao: And then the engineering team. Also, I would say there were a lot of data bugs or
66 00:11:09.496 ⇒ 00:11:13.409 Henry Zhao: issues with the data because of maybe a lack of English skills.
67 00:11:13.890 ⇒ 00:11:20.070 Henry Zhao: Oh, so, or column might have a typo in it, and then that would break things or.
68 00:11:20.070 ⇒ 00:11:20.800 Robert Tseng: I see.
69 00:11:20.950 ⇒ 00:11:23.930 Henry Zhao: Kind of messed up when I have to write a SQL. Query. Right? So
70 00:11:24.270 ⇒ 00:11:29.910 Henry Zhao: I would say those were the only only things I noticed about working with people from a different culture.
71 00:11:30.060 ⇒ 00:11:30.460 Robert Tseng: Yeah.
72 00:11:30.726 ⇒ 00:11:42.729 Henry Zhao: And then my most recent company is a very global company. But we work with a lot of teams in India. We worked with a lot of global teams. And I would say, just the working speed is a little bit different.
73 00:11:43.390 ⇒ 00:11:51.519 Henry Zhao: The team is a lot more cost based. So we have to like open a jira for everything we want to do, and processes just end up being really slow.
74 00:11:52.020 ⇒ 00:11:56.820 Henry Zhao: But other than that. I I don’t see any other like cultural differences or challenges.
75 00:11:57.150 ⇒ 00:12:00.855 Robert Tseng: Got it. Yeah, no, that makes sense language barrier. And then also, just like.
76 00:12:01.210 ⇒ 00:12:10.559 Robert Tseng: obviously like how you probably I guess how work is executed. Everything is off tickets. Right? So I mean, we run. We run a remote global team as well. And
77 00:12:10.560 ⇒ 00:12:37.988 Robert Tseng: we feel that pain like if getting people to kind of stretch to like. More than one task is kind of difficult, because it’s not all kind of put into a ticket. Then sometimes people don’t really think beyond the box, and that’s not to throw shade at just my team. I just think this in general, even the clients we work with, and we’re just remote teams in general. There’s less of that like being able to tap someone’s shoulder and be like, Hey, like, could you actually also do this or that or and like, yeah, I think,
78 00:12:38.300 ⇒ 00:12:49.869 Robert Tseng: yeah, we kind of live in a in a world of expectation for any and any sort of engineering work is that it has to be like spelled out completely before someone kind of executes on it right.
79 00:12:49.870 ⇒ 00:12:50.580 Henry Zhao: Right.
80 00:12:50.950 ⇒ 00:12:51.590 Robert Tseng: Yeah.
81 00:12:52.211 ⇒ 00:13:15.528 Robert Tseng: Yeah, I’d love to kind of transition towards like, I mean, just like over overall. So beyond culture is just industry experience that you’ve had. Obviously, you’ve worked in like massive like B, 2 B companies and B to C companies as well. So I guess b 2 b in the sense that, like your most recent company, I think, was that. And then, maybe obviously, Facebook being, I mean having both both sides of that marketplace as well.
82 00:13:16.055 ⇒ 00:13:24.989 Robert Tseng: Yeah. I mean, I don’t really meet too many people who kind of like cross the chasm between the, you know, like kind of the 2, and I’d love to kind of spend some time talking about like
83 00:13:25.110 ⇒ 00:13:31.149 Robert Tseng: the differences that you see in kind of b 2 b versus B, 2 C. Kind of analytics work.
84 00:13:31.960 ⇒ 00:13:40.130 Henry Zhao: Yeah. So over the past 4 years this has been true for B, 2 B+B, 2 CI think my biggest challenge has been identity resolution. So
85 00:13:40.340 ⇒ 00:13:49.770 Henry Zhao: I work a lot with marketing data, right? And so our goal is to market our product or market our service to as many people that are going to buy our product or service as possible. Right
86 00:13:50.352 ⇒ 00:13:55.999 Henry Zhao: so we set up utms, we build the multi-touch attribution model with rudder, stack and twilio segment.
87 00:13:56.140 ⇒ 00:14:05.059 Henry Zhao: But then the issue is what counts as a individual. Right? So if we’re doing B to a large 500 person corporation.
88 00:14:05.170 ⇒ 00:14:20.589 Henry Zhao: who are we actually reaching? Who’s the actual decision maker? And even when we’re doing B to C right? Like I mentioned before, the challenge is when they clear cookies, or when they are on their phone versus their computer, they kind of get a new IP address. So who who are they anymore? Right?
89 00:14:20.590 ⇒ 00:14:20.910 Robert Tseng: Yeah.
90 00:14:21.530 ⇒ 00:14:37.409 Henry Zhao: So identity. Resolution, I would say, is the biggest issue. So we’ve tried to solve that with, you know, rudder, stack and segment having the anonymous id, we look at pseudo user, Id and Google analytics. And I think even transitioning from universal analytics to J, 4 was a challenge.
91 00:14:37.880 ⇒ 00:14:38.510 Henry Zhao: Yeah companies
92 00:14:38.530 ⇒ 00:14:48.709 Henry Zhao: than that. In terms of understanding how we want to set up the events. What? How does? How do we set up Google tag manager? So that it makes sense. For what we want to analyze.
93 00:14:49.561 ⇒ 00:14:54.970 Henry Zhao: Still have an ongoing challenge in the face of all the people wanting. You know, more data privacy.
94 00:14:55.575 ⇒ 00:15:06.259 Henry Zhao: And I believe that less people are are accepting cookies. Less people are allowing themselves to be tracked over over different apps. And now, with Gdpr. And all these regulations that we have to ask.
95 00:15:06.530 ⇒ 00:15:07.649 Henry Zhao: I think.
96 00:15:08.160 ⇒ 00:15:17.789 Henry Zhao: people are getting more privacy. But I think marketing is becoming also less effective because it’s less easy to target people properly to follow user journeys
97 00:15:18.285 ⇒ 00:15:24.330 Henry Zhao: and to get the full, you know, customer 360, or understanding of the users marketing
98 00:15:24.530 ⇒ 00:15:25.870 Henry Zhao: journey, I would say.
99 00:15:26.120 ⇒ 00:15:26.680 Robert Tseng: Yeah.
100 00:15:26.680 ⇒ 00:15:42.060 Robert Tseng: yeah, I mean, we’ll talk about the identity resolution piece a bit more. I mean, I mean, I echo you on the Utm piece like, I remember when I started doing like any kind of just trying to identify anonymous users. And when the first, st you know, when when some of those. You know, big
101 00:15:42.800 ⇒ 00:15:57.889 Robert Tseng: regulation changes happened. I would say that, you know, utms maybe like 1015% was like, kind of insolvent. Or like, you know people. It was clear that it was. It was kind of leading you to rabbit holes. Now, I’m seeing it’s, you know.
102 00:15:57.980 ⇒ 00:16:10.949 Robert Tseng: 2530% like. And so I think it’s more and more people are opting out or using kind of privacy, first, st browsers and stuff. And yeah, like using any sort of like cookie based tracking is becoming less and less reliable.
103 00:16:10.970 ⇒ 00:16:23.340 Robert Tseng: That said, like, I think, for a lot of growth stage companies, it’s still directionally, you know, helpful to be able to identify up to that. And so I think companies are evolving to becoming more
104 00:16:23.350 ⇒ 00:16:52.179 Robert Tseng: okay with making decisions in some degree of ambiguity. And I think that actually creates more focus in some ways, because you can clearly tell which channels are performing. You may not be able to see it for the full like kind of pie. But at least you know 60 70% of your users. You know that like directionally, that’s where you need to go. And so I think this problem definitely becomes more amplified like later on, as you’re more mature in the market. And you’re not, like, you know, like being able to make
105 00:16:52.260 ⇒ 00:16:56.289 Robert Tseng: optimizations to your channel investments, you know, is.
106 00:16:56.550 ⇒ 00:17:09.080 Robert Tseng: you know, ha! Has some multimillion dollar impact to your to your bottom line as opposed to, you know something that’s much more insignificant earlier on. So I’m curious, kind of like your perspective on, like.
107 00:17:09.310 ⇒ 00:17:22.420 Robert Tseng: yeah, like, what? What point in a organization’s maturity. Should they be worried like kind of more concerned about identity resolution, and like what’s enough to get by with, like as like a company is kind of growing.
108 00:17:23.010 ⇒ 00:17:29.589 Henry Zhao: I would say, at the very initial stage, the data should already be there, right? So they should already be logging the data, logging the events.
109 00:17:30.221 ⇒ 00:17:51.889 Henry Zhao: trying to identify users as much as possible, whether it’s by their email or just have some sort of unique identifier that you’re gonna say, this is what we count as a person like, I think most of the time it’s email, right like, if they marketing form or they purchase anything, and they provide us their email, that is, gonna be their Id for for this starts right? Or it could be phone number. Who knows
110 00:17:52.020 ⇒ 00:18:13.339 Henry Zhao: but have the data there, because if you wait till a later stage to then implement it, then you don’t have as much historical data that you’ve gathered, and then it becomes hard to. You have to wait a lot longer for the data to start getting collected, to have enough sample size to be able to analyze what you want to analyze. And also there’s gonna be errors in the way you set up your data, and I’d rather find that out in the beginning than
111 00:18:13.420 ⇒ 00:18:32.929 Henry Zhao: you know when you’re rushing. If you’re like a startup when you’re rushing to get fundraising money, or when you’re rushing to show your stakeholders that or the shareholders that the the product has has value and potential by then it might be too late. So get the data set of as early as possible. But you don’t need to, I think, worry about getting
112 00:18:33.040 ⇒ 00:18:34.819 Henry Zhao: accurate data until
113 00:18:35.310 ⇒ 00:18:41.749 Henry Zhao: you feel like you almost need to start showing value to to investors, or whoever it may be.
114 00:18:41.920 ⇒ 00:18:48.600 Robert Tseng: Yeah, yeah, no, I think that’s that’s that’s really good. I mean, I feel like not enough people take that perspective. So so that’s usually
115 00:18:49.050 ⇒ 00:18:53.150 Robert Tseng: any identifying users ends up kind of being like a
116 00:18:53.540 ⇒ 00:18:59.610 Robert Tseng: I feel like companies. Push it off as much as as much as they can, and try to. Just, you know.
117 00:18:59.710 ⇒ 00:19:15.649 Robert Tseng: they just use Google analytics, or you know something out of the box and just accept that. That’s maybe enough for where they’re at. But yeah, I think being able to put a lot more attention into the upfront setup and how it yields returns later on. I think that’s that’s a that’s a good, that’s a good perspective.
118 00:19:16.484 ⇒ 00:19:29.459 Robert Tseng: If we can ask one more question on that point, like, so yeah, obviously, you’re unifying off of email that like, you know, some of some of the clients that we work with especially in the healthcare. And I won’t say too much more than that, like.
119 00:19:29.710 ⇒ 00:19:35.400 Robert Tseng: yeah, you do a fan out in your like user kind of in your user model, where you end up having, like
120 00:19:35.400 ⇒ 00:19:59.860 Robert Tseng: 2030 plus like different anonymous ids, or whatever it is like pseudo, anonymous ids all these different things that you know. It just gets wider and wider in terms of like what you what you can in your user identity graph. And so it becomes harder and harder to stitch even the longer that longer goes on. Right? So I think that’s also something that’s like counter
121 00:19:59.860 ⇒ 00:20:19.210 Robert Tseng: intuitive. It’s like, yeah, you may set it up earlier on. But then you think that you’ve you’ve identified your users early on, but it actually gets more and more complicated. And you may have to update things like later on. So I don’t know if you’ve dealt with that kind of problem that fan out in your in the work that you’ve done, and curious, if you have any thoughts there.
122 00:20:19.750 ⇒ 00:20:22.400 Henry Zhao: No, I haven’t. So. But that’s a that’s a good. Yeah.
123 00:20:22.690 ⇒ 00:20:23.320 Robert Tseng: Yeah.
124 00:20:23.480 ⇒ 00:20:46.169 Robert Tseng: okay, well, I mean, and then on the b 2 b side, like, I think, like you said finding the right decision maker, I think that’s an interesting kind of piece as well. So yeah, what? How do? How would you recommend? Like A, b 2 b kind of company, like figuring out like, how how should they be setting up their I identity resolution at the account level and at the individual level.
125 00:20:46.940 ⇒ 00:20:55.036 Henry Zhao: Yeah. So I would say most. I don’t know if this is safe to assume. But I would say, most b 2 b companies use some sort of crm like salesforce right?
126 00:20:55.410 ⇒ 00:20:58.559 Henry Zhao: And I think salesforce is pretty good at
127 00:20:59.289 ⇒ 00:21:05.570 Henry Zhao: setting up hierarchies and having an account id, but also a contact id, but also opportunity ids
128 00:21:06.004 ⇒ 00:21:11.970 Henry Zhao: and just understanding the the graph there. But I do think there’s also opportunity for salesforce to improve
129 00:21:12.200 ⇒ 00:21:13.080 Henry Zhao: kind of
130 00:21:13.200 ⇒ 00:21:37.780 Henry Zhao: the identity resolution they have when we were looking to cdps, thinking that that might be a solution to identity resolution we salesforce is one of the cdps that we were looking at, and I think the whole team had a concern of. If you guys are good at identity resolution, why is your salesforce product. Not that good at identity resolution and got. Was that well, you kind of have to pay for our
131 00:21:38.170 ⇒ 00:21:43.830 Henry Zhao: Cdp solution to really get our good stuff, which I don’t know if that’s right with with our team.
132 00:21:44.582 ⇒ 00:21:46.759 Henry Zhao: But that’s something that I always wonder is
133 00:21:47.790 ⇒ 00:21:49.700 Henry Zhao: arsenal that I haven’t looked at yet.
134 00:21:49.980 ⇒ 00:21:54.490 Henry Zhao: That can help us with identity resolution. But for now we just kind of use
135 00:21:55.860 ⇒ 00:21:58.850 Henry Zhao: like our analysts. Knowledge of saying.
136 00:21:59.240 ⇒ 00:22:17.359 Henry Zhao: You know, one of the fields we have in salesforce is account size. So if the account is like one to 5 people in the company then makers, just whoever is reaching out to us, but if it’s 500 plus, then maybe we go by title where it’s like whether it’s a manager in the title or I fire whatever may be right. So
137 00:22:17.520 ⇒ 00:22:22.219 Henry Zhao: we were. That was kind of a long term project we were working on to figure out our own identity resolution.
138 00:22:22.670 ⇒ 00:22:34.379 Robert Tseng: Yeah, no, I mean, that’s I mean, sounds like you’re creating your own heuristics for how you should be like labeling these different accounts. And yeah, it’s it’s funny. I think we mentioned this at the event we were talking about, like how
139 00:22:34.560 ⇒ 00:23:03.159 Robert Tseng: vendors typically hold out like the best features that you want from you, unless they pay enterprise prices. And it’s often just like this one thing that you really want. And it’s like, I don’t really care about the other 80% of the things they’re giving me. I just want this one thing, and you don’t have access to it right? And so that forces you to innovate because you have to like find workarounds because nobody wants to pay that enterprise tier. But that’s I feel like that’s very much the story of many data people’s careers, and like how they get better at some really specific niche.
140 00:23:03.730 ⇒ 00:23:04.530 Robert Tseng: Yeah.
141 00:23:04.920 ⇒ 00:23:22.600 Henry Zhao: But ultimately we like, I said, we were at that company. We worked with a lot of global teams. And there was a global team that was just ended up being tasked with identity resolution. So they were just doing a year long effort of cleaning up emails, cleaning up duplicate accounts and just making sure that we have a unique identifier for the purpose that we needed for.
142 00:23:22.910 ⇒ 00:23:33.110 Robert Tseng: Yeah. Yeah. One more. One more question on this kind of B, 2 b versus B, 2 C side, so obviously on the b 2 c side like these are, you know, maybe high or high frequency kind of
143 00:23:33.110 ⇒ 00:23:56.389 Robert Tseng: products like the the decision threshold is probably lower. And that’s why you’re able to do. You know much higher volume in terms of your marketing efforts? And maybe a lot of the business. I know there’s a big shift in E-com, specifically moving towards subscription and wanting kind of saas, like predictable revenue, and also margins. But you know, generally speaking, that’s not how it is across the industry
144 00:23:56.692 ⇒ 00:24:06.689 Robert Tseng: versus, like, you know, software traditional. B, 2 B saas like very high touch. Once you’re locked in, you’re locked in for a year long, maybe multi year long. Kind of contracts.
145 00:24:07.098 ⇒ 00:24:28.589 Robert Tseng: From like a data person’s perspective. And I guess, like, in terms of how you serve analytics to your stakeholders like, have you noticed, like a big difference in the stakeholders, and how what they expect from you in terms of the quality, accuracy, velocity of like the work that you deliver. Yeah.
146 00:24:29.450 ⇒ 00:24:33.089 Henry Zhao: But quality, velocity, and accuracy. I’d say no, I would say, every company
147 00:24:33.190 ⇒ 00:24:40.789 Henry Zhao: kind of wants to focus on those things right? But I would say startups care a little bit more about speed, because they’re always trying to meet
148 00:24:41.120 ⇒ 00:24:48.819 Henry Zhao: a fundraising round, or they’re trying to make a quick decision on whether to cut a product or to continue a product or what to go in.
149 00:24:49.354 ⇒ 00:24:58.350 Henry Zhao: Additionally, I would say, companies where the the product is a quick decision turnaround, right? Like, whether it’s buying a a Cpg product or
150 00:24:58.843 ⇒ 00:25:15.466 Henry Zhao: just signing up for something really quickly, versus like a very expensive software that you that for A, b 2 b, that the company would be locked to have to have training, have to implement implement that. Obviously that looks very different. Right. So for those quick purchases, it’s very much about.
151 00:25:16.340 ⇒ 00:25:23.290 Henry Zhao: What are the impressions that the individual are seeing from our marketing? What is their product? Lifecycle. How can we upsell them?
152 00:25:23.760 ⇒ 00:25:32.360 Henry Zhao: Those types of things? Whereas the b 2 b like more heavy products, there is a lot more of an ecosystem where we need to do
153 00:25:32.879 ⇒ 00:25:38.139 Henry Zhao: such attribution. We need to see the full buyer journey. We need to see the product journey
154 00:25:39.470 ⇒ 00:26:03.370 Henry Zhao: because people take longer to make the decision right? They’re not going to just see one ad and be like, all right, I’m gonna buy this 1 million dollar software also the sales teams that get involved. Right? So at Facebook, you’re not gonna have a salesperson following up with you on a Facebook ad you saw. But for b 2 b, you’re gonna have sales teams, and then sales is gonna want credit for their sales, whereas marketing wants to get credit for bringing sales. Those leads. So
155 00:26:03.470 ⇒ 00:26:11.079 Henry Zhao: the expectation is, how do we, as a data team, figure out who actually was responsible for that revenue
156 00:26:12.290 ⇒ 00:26:17.449 Henry Zhao: while motivating each stakeholder to still do their best right.
157 00:26:17.830 ⇒ 00:26:18.460 Robert Tseng: Yeah.
158 00:26:18.630 ⇒ 00:26:22.289 Henry Zhao: So you want to motivate marketing to bring in quality leads.
159 00:26:23.370 ⇒ 00:26:38.229 Henry Zhao: But then you also want to motivate the sales teams to to make as many sales as possible, and do as many upsells as possible. So how do you attribute and give credit to the proper team while motivating them to to maximize revenue. It’s kind of like the prisoner’s dilemma. Kind of situation. But
160 00:26:38.910 ⇒ 00:26:41.990 Henry Zhao: it’s something that we are still kind of trying to figure out.
161 00:26:42.490 ⇒ 00:27:05.219 Robert Tseng: Yeah, no, that’s that’s really interesting. I mean, I’m interested. I don’t really know that much in that in that world. I haven’t worked at the scale in terms of enterprise on your side. So I’m curious. What have you learned about like what motivate? I mean? Obviously, everything’s tied to a monetary incentive I’m assuming. But like, what does that distribution look like? And especially from like a multi-touch attribution perspective, like.
162 00:27:05.280 ⇒ 00:27:29.250 Robert Tseng: I think, a lot of the folks that I talked to especially I mean, you know, there’s a lot of buzzwords being thrown around mixed media modeling multi-touch attribution. Everybody wants to identify, you know customers beforehand and be able to see what are the different touch points. I think. Like you said. You know, if you’re just buying like a hundred dollar product or something. That’s, you know, a couple $100. Maybe one or 2 ads will get you there, but obviously with a much longer sales cycle for
163 00:27:29.280 ⇒ 00:27:39.110 Robert Tseng: couple 100 billion dollar product very different. So I’m curious, like, kind of what what is like, what does that extreme? What does that side of the world look like, yeah.
164 00:27:39.290 ⇒ 00:27:45.971 Henry Zhao: Yeah. So so for the marketing side, we try to motivate them by 2 metrics, which is number of leads
165 00:27:46.758 ⇒ 00:28:02.069 Henry Zhao: by segments. So we have different segments which will act differently right? We have different segments that act differently in terms of conversion rates or amount of time from the lead generated to closed one, and then we also closed one right? So
166 00:28:02.320 ⇒ 00:28:06.820 Henry Zhao: just assuming that every salesperson is equal and equally skilled.
167 00:28:07.140 ⇒ 00:28:26.919 Henry Zhao: what does each campaign or each marketing rep bring in terms of percent close one. So if somebody brings in a million leads, but only 1% are closed one, whereas another one brings in 100,000, but 5% are closed. One, we kind of want to have that balance of you don’t want to. Just you want to bring those good leads that are easy to close.
168 00:28:27.150 ⇒ 00:28:27.510 Robert Tseng: Yeah.
169 00:28:27.510 ⇒ 00:28:34.310 Henry Zhao: And the sales team. One thing that we saw was people was people were creating new accounts from the sales team.
170 00:28:34.470 ⇒ 00:28:39.599 Henry Zhao: The sales team was creating new accounts which was leading to duplicate accounts in salesforce because they wanted their credit.
171 00:28:39.790 ⇒ 00:28:44.799 Henry Zhao: So we, when we were cleaning up duplicate accounts, we were trying to de incentivize them from doing that.
172 00:28:45.410 ⇒ 00:28:46.230 Robert Tseng: Oh!
173 00:28:46.230 ⇒ 00:28:59.890 Henry Zhao: Then, maybe, like a note like we bounced. An idea of maybe putting a note of like marketing brought this to us, but this was already one of our leads. And here’s the proof, or here’s so, you know what like those are the kind of discussions we had on. How do we
174 00:29:00.340 ⇒ 00:29:06.300 Henry Zhao: figure out if it actually was the salesperson’s lead first, st or the marketing teams? Right? Because the salespeople no.
175 00:29:06.300 ⇒ 00:29:06.880 Henry Zhao: to have.
176 00:29:07.613 ⇒ 00:29:11.420 Henry Zhao: They have people that they know in the industry, so they have their own leads.
177 00:29:12.020 ⇒ 00:29:16.499 Henry Zhao: But they always want credit for every single content account that they work on right. So.
178 00:29:16.500 ⇒ 00:29:16.880 Robert Tseng: Yeah.
179 00:29:17.119 ⇒ 00:29:20.709 Henry Zhao: These are just examples of some of the things that we saw and things we’re trying to fix.
180 00:29:21.130 ⇒ 00:29:40.770 Robert Tseng: Yeah, I mean, that’s tough. I mean, like a hypothetical scenario, I mean, like, maybe marketing is already. There’s a lot of impressions for a particular account already, but then, maybe, like one of your sales, reps like met somebody at a conference and already kind of built that relationship. And it’s kind of hard to know you can’t just use like, you know, Timestamps, because, like, yeah, maybe the marketing impressions came first, st but
181 00:29:40.770 ⇒ 00:30:03.100 Robert Tseng: and like, maybe the League got brought in at some point. But like, really, if you didn’t have that in person, touch point at a conference like maybe that would have never even been like a live lead. So like, I think it’s a lot harder than on the Cpg side, where oftentimes we’re just like looking measuring like time, like, you know, just differences in time. Right? And are you using those like time period based
182 00:30:03.100 ⇒ 00:30:12.909 Robert Tseng: calculations to assign assign credit? But there’s a lot. The waiting mechanisms seem to be much more complicated on the when you’re considering the marketing sales impact right.
183 00:30:13.440 ⇒ 00:30:42.989 Henry Zhao: Yeah. And then B, 2 C, there’s a similar issue that we had. So we talked about identity, resolution and cookies being clear and things like that. So sometimes we would have influencers right where they would have a code or discount coupon that they can share on Youtube or Instagram, whatever it may be, and the customer can put that in at the time of purchase, and we would know that they came from the influencer. Right? We don’t need to look at attribution or like that, because we have that they use the coupon code. So then the issue becomes.
184 00:30:43.130 ⇒ 00:30:50.150 Henry Zhao: if we look at the user journey, and the person had already come in before the influencer. Do we still give credit to the influencer.
185 00:30:50.540 ⇒ 00:30:51.739 Robert Tseng: Oh, yeah.
186 00:30:51.980 ⇒ 00:30:55.830 Henry Zhao: So for a smaller startup where we want to grow, the influencer base.
187 00:30:55.980 ⇒ 00:30:56.850 Robert Tseng: Yeah, it would give us.
188 00:30:56.850 ⇒ 00:30:57.630 Henry Zhao: That. Yeah.
189 00:30:57.630 ⇒ 00:30:58.200 Robert Tseng: Yeah.
190 00:30:58.400 ⇒ 00:31:18.990 Henry Zhao: Whereas if it’s a more mature cycle and the influence is already very big, I would maybe put some limit on if they’re already a loyal user, or they already have an account. I wouldn’t give the influencer credit, or maybe give them partial credit right? So I think every company has this issue of. We want to motivate people to bring us customers. But how do we give the appropriate credit where it’s due?
191 00:31:19.340 ⇒ 00:31:29.180 Robert Tseng: Yeah, yeah, no. I mean, that’s a great example. I think in a lot of the growth stage Cbg companies we work with. If they have a coupon we automatically always attribute to the coupon. It’s just like.
192 00:31:29.800 ⇒ 00:31:35.009 Robert Tseng: yeah. So it’s either way we’re not. We haven’t even really broken it out with more nuance there.
193 00:31:35.480 ⇒ 00:32:00.450 Robert Tseng: Super interesting. I feel like you know, we I want to move on to another kind of topic in our talk and think we talk a lot about kind of method and strategy, and so more interested in talking about the stack a bit more so. I’d love to hear kind of like your kind of your personal experience, your your dream stack in terms of like what really enables marketing teams, you know. I think there’s different ways to put it. Maybe it’s just like
194 00:32:00.470 ⇒ 00:32:10.489 Robert Tseng: what are the different tools. Or maybe there’s like certain like workflows that are crucial to set up. Maybe it’s a combination of both. I I open to your interpretation on how you, how you receive that.
195 00:32:11.880 ⇒ 00:32:19.700 Henry Zhao: I think it depends on the marketing team. I think it depends on the marketing team, the stakeholders and their level of comfortability with with the data.
196 00:32:20.700 ⇒ 00:32:21.250 Henry Zhao: So I’ve.
197 00:32:21.250 ⇒ 00:32:21.850 Robert Tseng: For the correct.
198 00:32:21.850 ⇒ 00:32:36.379 Henry Zhao: Where, I would say, the marketing team has very little knowledge of data. They don’t understand how the back end data works. In that case. I like having just the dashboarding tool whether it’s tableau or metabase or looker studio
199 00:32:36.886 ⇒ 00:32:41.389 Henry Zhao: where they can play around with the filters. They can request additional data.
200 00:32:41.819 ⇒ 00:32:46.150 Henry Zhao: They can basically play with the data in the way that it makes sense.
201 00:32:48.190 ⇒ 00:32:52.510 Henry Zhao: Whereas if they have a lot more experience with data. I like to give them more flexible tools.
202 00:32:53.349 ⇒ 00:32:57.240 Henry Zhao: I don’t know what that looks like, but maybe even giving them an ability to
203 00:32:57.510 ⇒ 00:33:12.920 Henry Zhao: write sequel queries themselves, or, be able to use parameters and other types of more complex filters, or be able to export into excel, and then kind of mess with the data themselves. If it’s a marketing team with not a lot of knowledge of the data, I’m more hesitant for them to export, to excel
204 00:33:12.920 ⇒ 00:33:27.729 Henry Zhao: and then do lookups because they could be double counting things. They could be misinterpreting the data. I just want to be a little more hands on with them. So I kind of make a more rigid stack, so that they have to come to me. And they need additional data. So.
205 00:33:27.730 ⇒ 00:33:28.290 Robert Tseng: Yeah.
206 00:33:28.630 ⇒ 00:33:30.610 Henry Zhao: It’s it’s kind of like that, I would say.
207 00:33:30.610 ⇒ 00:33:55.099 Robert Tseng: Yeah, no, I think that’s a great way of framing it. I think. Yeah, kind of the data person is the kind of the gatekeeper to the data right? And I think if you gatekeep too much, then people complain, or you know there’s like a Oh, like, why does it. Why is there so much friction having to go through and kind of all these? You know this in order to to make these requests and things, and people just want that elusive concept of like self. Serve right? I’m curious, like.
208 00:33:55.400 ⇒ 00:34:10.139 Robert Tseng: yeah. And what is there has been a spin, a situation you’ve been in where you’ve seen self serve and kind of the execution. The delivery of that be really work. Work well, especially for your marketing counterparts, and maybe another situation or yeah, and if not, then you can maybe talk about where it hasn’t worked.
209 00:34:11.538 ⇒ 00:34:30.969 Henry Zhao: Yeah, I would say one thing that I wish self serve would work better is the marketing team understanding what I’m capable of as a data person. Right? So here’s an example. So let’s say, they have a dashboard that they want to look at with 3 different filters. Right? So maybe it’s a month user type, and then I don’t know utm medium.
210 00:34:31.260 ⇒ 00:34:37.149 Henry Zhao: right? So I’ll put these filters for them, and it works fine for them. But then they come up with a question where it’s like.
211 00:34:37.510 ⇒ 00:34:40.460 Henry Zhao: actually, I don’t care about medium. I care about Utm source.
212 00:34:40.710 ⇒ 00:35:05.380 Henry Zhao: So instead of coming to me and asking me for that, because for me, all I do need to do is just change literally the code from utm medium utm source, and it’s done for them, but they might think that that would cost me a lot of work, or it might take a long time, so they might export it to excel, do a vlookup from medium to source, and then like, spend a lot of time, or maybe filtering medium by medium, by medium, by medium, by medium, and exporting each individual one a lot of time there.
213 00:35:05.752 ⇒ 00:35:16.439 Henry Zhao: So I think that’s the challenge that I’d like to address, which is, how do I get the marketing team to not do a lot of manual work and talk to me 1st to see if there’s an easy way I can pull. Pull that for them.
214 00:35:16.630 ⇒ 00:35:21.990 Henry Zhao: then I would understand their needs better, and also self. Serve them, have them self, serve better.
215 00:35:22.390 ⇒ 00:35:44.770 Robert Tseng: Oh, that’s so interesting. I feel like I mean, I mean, I run an agency. I feel like the I we come across the opposite problem. I wish people tried to solve their own problems more often, I think anything that’s missing is just like, Hey, like, can you? Can you add this. Can you add this? And there’s just always this like additive mentality of like, if you if I don’t know how to do something immediately, I’m not gonna try to figure out how to do it. I’m just gonna request it.
216 00:35:45.313 ⇒ 00:35:51.980 Robert Tseng: Maybe that’s just the expectation we’ve set up that. That’s how we work with them. But yeah, I think that’s that’s interesting.
217 00:35:51.980 ⇒ 00:35:52.780 Henry Zhao: But yeah.
218 00:35:53.030 ⇒ 00:35:53.669 Robert Tseng: So I like.
219 00:35:53.670 ⇒ 00:36:15.770 Henry Zhao: I’ve come into companies where the previous data person, for example, they had a dashboard, for, like Utm, one by day, utm 2 by day. Utm 3 by day. Utm, one by week. Utm 2 by day. We can, I say, make a filter for like whether it’s day week, and then you can make a filter for the Utm. Then you don’t need 128 dashboards.
220 00:36:15.770 ⇒ 00:36:16.850 Robert Tseng: Oh, my goodness!
221 00:36:17.880 ⇒ 00:36:27.059 Henry Zhao: So I think it goes both ways. But I like seeing the extreme of how much manual work they’re doing, so that I can simplify and streamline for them.
222 00:36:27.150 ⇒ 00:36:29.727 Robert Tseng: Yeah, okay, that’s a good point of view.
223 00:36:30.050 ⇒ 00:36:35.190 Henry Zhao: I’d rather that than doing manual work, and I not know about it. Yeah. And I them a lot of time.
224 00:36:35.570 ⇒ 00:36:38.030 Robert Tseng: Yeah, yeah, no, that’s that’s that’s fair.
225 00:36:39.160 ⇒ 00:36:50.282 Robert Tseng: I I’d love to talk more about kind of your the migration from segment to rotor stack. Specifically I think we talked about kind of motivations around it like just being have more flexibility.
226 00:36:50.790 ⇒ 00:37:20.309 Robert Tseng: and obviously the cost driven as well. Yeah, I guess anything else that you felt like you were able to when you were setting that when you were, you know, building that case for your stakeholders on like why they need to move from one tool to another. I mean, I imagine, at least in my shoes, a lot of the time. The person who controls the budget. I mean, they’re pretty tool agnostic. They don’t understand. It’s a lot of it is back end work that they don’t really see. And so being able to. I mean, it really has to be a cost driven decision, for that’s that’s the biggest. That’s the biggest thing.
227 00:37:20.566 ⇒ 00:37:24.920 Robert Tseng: But anything else that you felt like really helped you kind of like push for that decision.
228 00:37:26.020 ⇒ 00:37:28.619 Henry Zhao: No, I say, it’s strictly just a budget thing, right? Segment.
229 00:37:28.620 ⇒ 00:37:28.950 Robert Tseng: Yeah.
230 00:37:28.950 ⇒ 00:37:51.640 Henry Zhao: Once you get to a certain amount of events or ma use, I don’t remember which one it was. They significantly jacked up the price, and every company has budgets right? We can’t just pull money in there. So we looked at a decent solution that was that fit into our budget and the rudder stack team was very helpful in in getting us set up and understanding. You know how things work, and they were very helpful to us.
231 00:37:52.080 ⇒ 00:38:19.909 Robert Tseng: Yeah, okay, cool. And then, as far as like, you know, segment kind of comes out of the box. A lot more boundaries and kind of things in place that you set up in a ui obviously meant to be more friendly for non technical folks to also migrate their workflows into it, whereas, like rudder stack, everything ends up running through the data team and having to kind of all anything lives in the in the warehouse now. So, were there any like, yeah, was that what you anticipated? Any unexpected challenges you ran into with kind of
232 00:38:20.180 ⇒ 00:38:25.050 Robert Tseng: kind of switching the the paradigms of how how some of these data workflows work.
233 00:38:26.260 ⇒ 00:38:35.650 Henry Zhao: No, I think just working with engineering was a little bit of a challenge, because they always have so much going on right. I’m sure anyone at every company can understand how engineering is always swamped.
234 00:38:36.408 ⇒ 00:38:44.419 Henry Zhao: I would say, just working together with them and making sure that it’s get set up as quickly and as correctly as as possible.
235 00:38:44.900 ⇒ 00:39:01.140 Robert Tseng: Yeah, I’m curious, like in in, you know, obviously in the phase roll out like, kind of what kind of what were those phases for you. I mean, if there’s anything that’s in meaningful to talk about there and then, like, what what do you feel like? Let’s get what what you guys got hung up on like, you know, maybe.
236 00:39:01.270 ⇒ 00:39:02.970 Robert Tseng: Yeah, I mean, if that’s.
237 00:39:02.970 ⇒ 00:39:19.180 Henry Zhao: I would say on my end. One of the challenges that I am still facing is exploit. So every a lot of companies have, you know, a segment or rotor stack. But then they also have Google analytics for. But then they also have some sort of telemetry tool, right? Whether it’s a mix panel or an amplitude to log events
238 00:39:19.690 ⇒ 00:39:26.029 Henry Zhao: and analyze those events. I think a challenge is explaining to stakeholders and explaining to marketing why those numbers don’t match.
239 00:39:26.725 ⇒ 00:39:38.829 Henry Zhao: Explaining why, what the difference between client side and server side is, I think that is always a challenge. We try to explain it like the the purpose of these, this logging is different.
240 00:39:39.820 ⇒ 00:39:44.490 Henry Zhao: But yeah, I think it’s something that still a lot of people don’t understand.
241 00:39:44.650 ⇒ 00:39:45.560 Henry Zhao: 4 week
242 00:39:46.317 ⇒ 00:39:51.720 Henry Zhao: and and that can, and that can lead to a an erosion of trust. Right?
243 00:39:52.390 ⇒ 00:39:52.740 Henry Zhao: Yes.
244 00:39:52.740 ⇒ 00:39:59.789 Henry Zhao: Google showing me this one thing. But then segment is showing me something else, and then we changed rudder stack, and that shows yet a different number
245 00:40:00.278 ⇒ 00:40:13.970 Henry Zhao: and then our mixed panel is also showing a different number. I don’t know which number to trust, so I’m gonna trust none of them. So that’s an an a situation I think we want to avoid. And I think you should be able to be the person to understand what tool should be used for for what purpose?
246 00:40:14.439 ⇒ 00:40:18.589 Henry Zhao: And also so that it’s telling at least the same story, and not wildly different stories.
247 00:40:18.830 ⇒ 00:40:35.579 Robert Tseng: Yeah, I mean, yeah, I think I handle those objections all the time, like, I love to talk more. I mean, from my experience, like Google analytics, the inflation like it like they inflate the numbers the most. And I think my kind of explanation is more of like, okay, Google analytics. It’s, you know.
248 00:40:35.580 ⇒ 00:40:51.939 Robert Tseng: directionally, you get a good view of your traffic. I mean, I think you know, out of the box does a pretty good job of breaking out by channel. And some of these, like higher level, like when you’re trying to compare channels like Google analytics is great for that. Once you want to drill down to user specific data. You definitely need a telemetry tool
249 00:40:52.285 ⇒ 00:41:12.330 Robert Tseng: my preference is really to instrument the the events through like a segment or rudder stack rather than through amplitude and and mixed panel. Just so like you can actually get your Cdp to line up with your with with your product analytics tool? So I think that helps to eliminate the difference there.
250 00:41:12.712 ⇒ 00:41:20.679 Robert Tseng: But yeah, as far as like handling the objections between Google analytics and and segment, I think what I found is that really unlike
251 00:41:20.680 ⇒ 00:41:44.289 Robert Tseng: for the for for the part of the marketing house, that’s more qualitative, maybe, like the brand or web web team that’s responsible for bringing traffic into the website. They tend to hate the Cdp tools and mixed panel amplitude, or whatever, because, yeah, it doesn’t look as the numbers don’t look as good for them, whereas, like the lifecycle marketers really love
252 00:41:44.885 ⇒ 00:41:46.670 Robert Tseng: those tools because
253 00:41:46.710 ⇒ 00:42:01.090 Robert Tseng: they don’t really care so much about the top of the funnel. They just really want to be able to get their win back and retargeting campaigns really dialed in. And they love the audience building that they can have there. I don’t know about if there’s anything that you would add from what you’ve observed, there.
254 00:42:01.730 ⇒ 00:42:02.820 Henry Zhao: No, that makes sense.
255 00:42:02.820 ⇒ 00:42:03.420 Robert Tseng: Yeah.
256 00:42:03.780 ⇒ 00:42:29.359 Robert Tseng: Okay, cool. Yeah. I think, that’s that’s that. That’s that’s good on the Cdp side. I mean, I had. Yeah, I mean, ultimately ends up being like a budgeting kind of decision. And so I think for for listeners that kind of are evaluating kind of what vendors to choose. Yeah, I mean, ultimately, like you have to that. That’s the primary thing that you have to work with. And a stakeholder really cares about. You know, from a scaling perspective.
257 00:42:29.700 ⇒ 00:42:41.430 Robert Tseng: You know, I think it’s important to model out your growth so that you know that when you’re gonna hit certain thresholds and then things get kind of exponentially more expensive. And if you think that the company is going to grow to that point, you can
258 00:42:41.560 ⇒ 00:42:51.000 Robert Tseng: being able to re platform across these different tools. Would you say? That’s a hard effort to go from segments? Rotor stack or something like that? Can you like ladder up into it?
259 00:42:51.878 ⇒ 00:42:56.789 Henry Zhao: No, yeah. I don’t think it’s that challenging, just because I think every company provides a lot of support
260 00:42:58.140 ⇒ 00:43:06.140 Henry Zhao: to make the transition and also once you understand the concept of how to implement it. That, I think, is the initial challenge that once you overcome, that it’s not
261 00:43:06.880 ⇒ 00:43:08.390 Henry Zhao: not a big deal anymore.
262 00:43:08.700 ⇒ 00:43:09.920 Robert Tseng: Yeah, got it?
263 00:43:09.920 ⇒ 00:43:10.450 Henry Zhao: Yeah.
264 00:43:11.220 ⇒ 00:43:39.199 Robert Tseng: Cool. Well, I mean, I know we’re like, I mean, we might go a little bit over if that’s okay. But I’d love to kind of talk more about, like, yeah, your perspective on the use of AI, kind of in in yeah. For in in data and specifically marketing data. You know, I kind of send you a few questions that are kind of just more high level things about like, yeah, what does it mean for a data team to be ready for AI. But I’d love to kind of hear, you know, for your perspective, like.
265 00:43:39.900 ⇒ 00:43:51.470 Robert Tseng: Yeah, are there any of the work. What’s like a real what’s like an interesting workflow that you think will really be disrupted by the usage of AI. That, like people may not be talking about today.
266 00:43:53.383 ⇒ 00:44:00.606 Henry Zhao: Well, initially, I would just say, I think a lot of companies right now over over or misuse the term AI
267 00:44:01.060 ⇒ 00:44:15.870 Henry Zhao: upper leadership says, you know, everyone’s doing AI. We’ve got to do it as well figure out for me where it makes sense. That’s I. I don’t. I didn’t seen that in person. But that’s kind of how it feels to me. The the leadership direction comes and feels a little bit forced
268 00:44:16.868 ⇒ 00:44:20.809 Henry Zhao: like in situations that I don’t don’t think require AI.
269 00:44:21.448 ⇒ 00:44:31.579 Henry Zhao: Then a lot of people, I think working data think that they’re gonna use AI to write queries for them. And I’m so not sure that that is the best use right now.
270 00:44:33.890 ⇒ 00:44:39.970 Henry Zhao: I really do think that the opportunity for AI right now is bridging that gap between, what marketers are asking for
271 00:44:40.750 ⇒ 00:44:45.579 Henry Zhao: and what they are so like, what they ask for and what they actually are looking for. Right.
272 00:44:45.580 ⇒ 00:44:46.250 Robert Tseng: Yeah.
273 00:44:46.530 ⇒ 00:44:57.550 Henry Zhao: Because right now, humans are evaluating that. But that doesn’t really need to be evaluated by humans. Because if you analyze the work of thousands of data analysts. You’re going to see some of the same themes. You’re gonna see some of the same patterns
274 00:44:58.582 ⇒ 00:45:05.659 Henry Zhao: where AI can definitely swoop in and basically be a translator between marketers and and data people. I think
275 00:45:05.980 ⇒ 00:45:12.160 Henry Zhao: I see the most value from from a data perspective. And I don’t see it as much in terms of
276 00:45:12.260 ⇒ 00:45:14.700 Henry Zhao: like a marketer needing to ask.
277 00:45:14.820 ⇒ 00:45:19.860 Henry Zhao: Where are my users coming from as opposed to just selecting a filter or
278 00:45:20.110 ⇒ 00:45:24.970 Henry Zhao: or writing a script or asking for a specific dashboard like I I think
279 00:45:26.300 ⇒ 00:45:27.729 Henry Zhao: you know what I’m saying. Like I I think.
280 00:45:27.730 ⇒ 00:45:28.280 Robert Tseng: Yeah.
281 00:45:28.440 ⇒ 00:45:32.330 Henry Zhao: AI doesn’t need to help them ask the question. It needs to help them transfer the question.
282 00:45:32.550 ⇒ 00:45:35.510 Robert Tseng: Yeah, no, I think that’s a really good perspective. I think.
283 00:45:36.360 ⇒ 00:46:01.189 Robert Tseng: yeah, I mean, obviously, all of these, like AI kind of features that are kind of integrated into a lot of these tools have all defaulted to using like a chat base. You can ask any questions that you want to the data. I don’t really know if it’s displacing like just having clean views set up your preset filters like, you know, there, there’s like a routine that, like operators really go through where they wake up or whatever their workflow is like. They’re just going checking a few different
284 00:46:01.440 ⇒ 00:46:11.089 Robert Tseng: screenshots. And you’re and you’re able to connect the dots pretty quickly on whether or not your key core core metric is moving up or down, and how you should respond to that so being able to like.
285 00:46:12.120 ⇒ 00:46:14.590 Henry Zhao: Yeah, like the analogy I’ll give is
286 00:46:14.830 ⇒ 00:46:24.019 Henry Zhao: if I wanted to know what the weather is going to be like this week. So I know what to wear, or or yeah, where, if I’m going on vacation, you want to know the weather, what the weather is going to be like in Spain, so I know what to pack.
287 00:46:24.180 ⇒ 00:46:37.229 Henry Zhao: I don’t need AI to help me ask, what is the weather going to be like in Spain? Weather, app, click on Spain and see the weather right. But where AI Handy is, figure out that what I’m really asking is knowing what to pack
288 00:46:38.230 ⇒ 00:46:52.219 Henry Zhao: and able to suggest you should pack ABC similar to marketers. Right? Where are my users coming from? The the marketer actually wants to know, where should I be investing more marketing dollars in, or where should I maybe stop cutting spend on?
289 00:46:52.887 ⇒ 00:46:55.460 Henry Zhao: So it needs to be able to translate to insights
290 00:46:55.700 ⇒ 00:47:00.439 Henry Zhao: instead of just helping ask the question to generate a query. That’s what I’m saying.
291 00:47:00.900 ⇒ 00:47:15.259 Robert Tseng: Yeah, no, I think that’s that’s a really good analogy. And I’ll I’ll add one more thing. So I think for from from my perspective one of the biggest barriers to better marketing. Like.
292 00:47:15.260 ⇒ 00:47:33.690 Robert Tseng: like like activation for for data is just the simple fact that maybe the the lifecycle person or the other. My marketing counterpart is not fully stretching the capabilities of like what they could be using to retarget. And so we end up kind of just like circling the same set of like
293 00:47:34.627 ⇒ 00:48:02.499 Robert Tseng: like descriptive you know, attributes. For like, you know, where users coming from location like like kind of maybe different types of source different grains of of source and maybe some basic behaviors like, you know, number of orders or whatnot, but being able to stretch their perspective on like, there are a lot, you know, what is that like ideal like user model that we can kind of create for you. So that.
294 00:48:02.800 ⇒ 00:48:22.539 Robert Tseng: yeah, we we layer in some predictive attributes in their predicted Ltd, or something like that, and be able to help them to be more creative with the different segments that they’re using in their campaigns. And so I find that you know, AI definitely helps fill that gap and be able to
295 00:48:23.430 ⇒ 00:48:51.010 Robert Tseng: give them that coaching so that they can also level up like what their capabilities of, for how they can purse be more personalized, and and the retargeting of customers. So yeah, I think that’s I mean, that’s something that I’m still kind of working through with with my clients. But I think that’s probably the most exciting part of this particular workflow to me of like, how I want to be able to use AI better to be able to help kind of coach coach my counterparts to do that.
296 00:48:51.480 ⇒ 00:48:52.060 Henry Zhao: No.
297 00:48:52.480 ⇒ 00:48:53.140 Robert Tseng: Yeah.
298 00:48:53.974 ⇒ 00:49:19.750 Robert Tseng: Cool. Well, I mean, I think, with the remaining kind of time we have left, I think. I like to do just like rapid fire questions. I think I kind of already cheated and gave you a few of them up in advance. But yeah, I think these are just more kind of like one word answers, or kind of phrases, like little sound bites that we could kind of use. I aggregate this data kind of at the end of I’d love to just know, like hey, like within my
299 00:49:20.220 ⇒ 00:49:40.640 Robert Tseng: pool, or like my rolodex of people, this is a tool that we all love or like. You know, we we talk about this the same way or not. So I think that’s really the purpose of this last segment usually doesn’t end up getting kind of thrown into the interview. So yeah, I love to hear. Like, is there? Yeah, what’s like underrated tool that you that you love, that people don’t talk about.
300 00:49:42.037 ⇒ 00:49:49.799 Henry Zhao: I like looker studio a lot. So looker studio used to be different, right? Like you would have to write. Look, ml, you’d have to set up
301 00:49:50.630 ⇒ 00:50:00.349 Henry Zhao: all these different structures but now it’s pretty much blended into data studio from Google. And I like how flexible it is with all the different data sources, and how easy it is to use.
302 00:50:01.006 ⇒ 00:50:10.490 Henry Zhao: I think every dashboarding tool has some flaws, but this one, I would say, is more underrated, whereas some other tools like maybe power Bi, are a little bit overrated.
303 00:50:10.860 ⇒ 00:50:11.920 Henry Zhao: That’s my opinion.
304 00:50:12.460 ⇒ 00:50:22.509 Robert Tseng: Yeah, I mean budget and flexibility. Wise look or studio definitely gives you that. Yeah. Yeah, I mean, I guess maybe you already answered this, but like kind of overhyped kind of
305 00:50:22.880 ⇒ 00:50:37.859 Robert Tseng: not necessarily term in AI, but like we talked about overhyped usage of AI right in data. So I think your analogy kind of answered that. But is there anything else that you feel like? You’re just like pretty indignant about how AI is being thrown around in data.
306 00:50:41.630 ⇒ 00:50:54.069 Henry Zhao: no, I just think it’s a little, sometimes a little corny when we’re making when we’re evaluating a software with with the sales team, and then the software, the sales team will always say like, and oh, and plus, we use AI for.
307 00:50:54.445 ⇒ 00:50:54.820 Robert Tseng: Excellent.
308 00:50:54.820 ⇒ 00:50:57.050 Henry Zhao: That we don’t really even care about.
309 00:50:58.120 ⇒ 00:51:05.499 Henry Zhao: My advice to salespeople in the in the tech space would be, don’t feel like you have to say, AI really understand your customer and what their needs are, and if it makes.
310 00:51:05.500 ⇒ 00:51:05.930 Robert Tseng: That’s.
311 00:51:05.930 ⇒ 00:51:07.909 Henry Zhao: Mention it, but
312 00:51:08.010 ⇒ 00:51:13.300 Henry Zhao: don’t just don’t say Oh, well, we’ve checked that box, the the box that everyone is is trying to check
313 00:51:14.360 ⇒ 00:51:16.740 Henry Zhao: the box, you know, because it comes across as a little bit, Corny.
314 00:51:17.130 ⇒ 00:51:24.807 Robert Tseng: Yeah, no, that’s that’s good. Know your customer. And yeah, AI is kind of over use at this point, so don’t don’t bring it up unless you really know what you’re saying. So.
315 00:51:25.040 ⇒ 00:51:27.380 Henry Zhao: Yeah, exactly. Yeah. Also know what you’re saying.
316 00:51:27.722 ⇒ 00:51:35.937 Robert Tseng: Yeah. And then, if not in data like, what would you be doing? Like? Yeah, career, wise or just, maybe not. Career.
317 00:51:36.820 ⇒ 00:51:39.089 Henry Zhao: I think I would like to be a doctor. Maybe.
318 00:51:39.390 ⇒ 00:51:40.570 Robert Tseng: Oh, wow!
319 00:51:40.570 ⇒ 00:51:43.556 Henry Zhao: In like medical biology, those kind of things.
320 00:51:44.550 ⇒ 00:51:48.178 Robert Tseng: I’m staying in data because it’s too late for me to go back to middle school school, but
321 00:51:48.770 ⇒ 00:52:00.469 Robert Tseng: never too late. Right? So yeah. And yeah, okay. And then, is there like a book or kind of concept paradigm that like framework that really shaped?
322 00:52:01.301 ⇒ 00:52:04.749 Robert Tseng: Kind of the way that you approach data.
323 00:52:05.971 ⇒ 00:52:12.210 Henry Zhao: I would say, for like from a corporate perspective, I think radical candor is a good book to read.
324 00:52:12.669 ⇒ 00:52:20.580 Henry Zhao: Because it’s so important when dealing with communication giving feedback and just working with a lot of cross functional, different teams
325 00:52:20.820 ⇒ 00:52:25.870 Henry Zhao: from a fun perspective. There’s a book that I think is really fun for people that like data which is called data Clysm.
326 00:52:26.500 ⇒ 00:52:32.720 Henry Zhao: it’s basically that talks about the data behind a lot of sociological concepts like dating
327 00:52:33.070 ⇒ 00:52:46.060 Henry Zhao: as one example. So one thing they did. They ran a lot of data from Okcupid to say, like, what are the difference between what men look for and what are the differences between what women look for and the charts are are super interesting.
328 00:52:46.490 ⇒ 00:52:50.110 Henry Zhao: Oh, wow! I gotta read that book. I haven’t read it before. That’s so interesting. Yeah.
329 00:52:50.110 ⇒ 00:52:53.190 Henry Zhao: yeah. But it’s more above. Yeah, yeah, go ahead.
330 00:52:54.520 ⇒ 00:52:55.170 Henry Zhao: Yeah.
331 00:52:55.170 ⇒ 00:53:03.190 Robert Tseng: Yeah. And and radical candor is actually a book that I recommend as well to a lot of data folks. It’s like, it’s not really technical books that you need at a certain point, like
332 00:53:03.460 ⇒ 00:53:09.939 Robert Tseng: more than half the battle is communication. And so yeah, you guys got to be a really good communicator. If you’re gonna go far data.
333 00:53:10.300 ⇒ 00:53:10.990 Henry Zhao: Yeah.
334 00:53:11.320 ⇒ 00:53:11.890 Robert Tseng: Yeah.
335 00:53:12.190 ⇒ 00:53:28.049 Robert Tseng: Okay, well, I mean, that’s that. Thank you for kind of sit bearing with us as we kind of, I mean, yeah, we’re kind of coming up on time. So yeah, it’s really really great catching up and kind of talking through some of these topics that have been on my mind that I wanted to chat with you about
336 00:53:28.525 ⇒ 00:53:40.529 Robert Tseng: yeah, I guess you’re. I’m just off the record at this point. But like, yeah, you’re in Brazil right now, kind of like, what are your? What are your kind of plans? Now? I guess you’re gonna come back to the States at some point, look for next role.
337 00:53:40.900 ⇒ 00:53:51.340 Henry Zhao: Yes, I’m going on vacation next week. It was my preplanned Pto, anyway, so I’ll be in Europe for 2 weeks but in the meantime, just still job hunting and also looking for potentially just independent contractor roles.
338 00:53:51.450 ⇒ 00:53:54.650 Henry Zhao: because that might be something that I just wanna do, moving forward.
339 00:53:55.070 ⇒ 00:54:01.050 Robert Tseng: Okay, yeah, I mean, let’s stay in touch about that. If you were interested in I mean, yeah, we get, we get a bunch of.
340 00:54:01.410 ⇒ 00:54:14.650 Robert Tseng: We have a lot of data work and need senior people to kind of be architects or strategist about it. And if you’re interested in kind of staying loop on that like happy to kind of collaborate with you and see see if there’s something that you’d like to work on together.
341 00:54:15.220 ⇒ 00:54:16.950 Henry Zhao: Absolutely. Yes, please. Yeah. Please keep me.
342 00:54:18.640 ⇒ 00:54:20.660 Robert Tseng: And then in Europe, where? Where are you headed?
343 00:54:21.694 ⇒ 00:54:31.099 Henry Zhao: So Spain France, and then I’m going to London for a friend’s wedding. So the reason I’m going actually is at the end. I’m going to my friend’s wedding in London.
344 00:54:31.410 ⇒ 00:54:36.029 Robert Tseng: Oh, awesome. I just came back from Europe. So I was just in
345 00:54:36.490 ⇒ 00:54:45.059 Robert Tseng: like Netherlands, and then Portugal. But yeah, it’s a it’s a good time to go, so hope you enjoy your time off. I know it’s been a long run.
346 00:54:45.710 ⇒ 00:54:52.969 Robert Tseng: and yeah, I mean, I guess we’ll we’ll I’ll be in touch. I won’t bother you while you’re on vacation, but maybe in a couple of weeks, when you’re back, then we’ll
347 00:54:53.860 ⇒ 00:54:56.119 Robert Tseng: we’ll we’ll we’ll find something.
348 00:54:56.540 ⇒ 00:54:59.429 Henry Zhao: No bother me anytime. Yeah, I’m looking forward to to following up.
349 00:54:59.870 ⇒ 00:55:04.410 Robert Tseng: Okay, cool. Alright. Well, thanks for your time, Henry, and I’ll talk to you soon. Then.
350 00:55:04.410 ⇒ 00:55:05.540 Henry Zhao: Thank you, take care!
351 00:55:05.540 ⇒ 00:55:06.669 Robert Tseng: Alright! See ya.
352 00:55:06.960 ⇒ 00:55:09.909 Hannah Wang: Bye, bye, thank you. Bye.