Meeting Title: Eden Case Studies Date: 2025-11-13 Meeting participants: Hannah Wang, Awaish Kumar, Henry Zhao
WEBVTT
1 00:01:20.990 ⇒ 00:01:22.130 Hannah Wang: Hello?
2 00:01:29.470 ⇒ 00:01:30.430 Awaish Kumar: Hello.
3 00:01:31.620 ⇒ 00:01:32.750 Awaish Kumar: How are you?
4 00:01:33.330 ⇒ 00:01:33.880 Henry Zhao: I know.
5 00:01:34.680 ⇒ 00:01:36.029 Hannah Wang: Ned, how are you guys?
6 00:01:36.240 ⇒ 00:01:37.090 Henry Zhao: Good, thanks.
7 00:01:37.650 ⇒ 00:01:39.660 Awaish Kumar: Algaritz, how about you?
8 00:01:41.400 ⇒ 00:01:46.920 Hannah Wang: Henry, did you move yet? Or where are you right now in the world?
9 00:01:47.710 ⇒ 00:01:50.739 Henry Zhao: Right, I’m in Brazil, but I did move recently, yeah.
10 00:01:51.080 ⇒ 00:01:57.299 Hannah Wang: to… you said… mmm… the West… the West Coast? No, Arizona?
11 00:01:57.300 ⇒ 00:01:58.080 Henry Zhao: Phoenix.
12 00:01:58.270 ⇒ 00:01:59.610 Hannah Wang: Phoenix, okay.
13 00:01:59.610 ⇒ 00:02:01.899 Henry Zhao: I guess technically West Coast.
14 00:02:02.340 ⇒ 00:02:06.120 Hannah Wang: Yeah, right, yeah, okay. Yeah, welcome to Pacific Time.
15 00:02:06.120 ⇒ 00:02:06.620 Henry Zhao: too.
16 00:02:07.840 ⇒ 00:02:08.500 Henry Zhao: But…
17 00:02:08.500 ⇒ 00:02:15.340 Hannah Wang: Alright, so today I’m thinking we can crank out two case studies. I guess…
18 00:02:16.150 ⇒ 00:02:26.169 Hannah Wang: yeah, I’m not really sure what, like, Awash touched versus what you, Henry, touched, so feel free to, like, chime in, whenever you
19 00:02:26.500 ⇒ 00:02:28.270 Hannah Wang: Feel like it’s relevant, but…
20 00:02:28.630 ⇒ 00:02:37.129 Hannah Wang: The two case studies we’re thinking of is… or I’m thinking of, is the segment Customer I.O. Eden.
21 00:02:37.250 ⇒ 00:02:49.059 Hannah Wang: work that we did, that enabled cost savings and market… like, enabled the marketer. I… I’m just reading out what, I guess, Robert kind of jotted down, so…
22 00:02:49.060 ⇒ 00:02:55.449 Henry Zhao: That’s the first part, is what Awash worked on. So, Awash, if you want to talk about customer-enriched profiles, the work there, and then I.
23 00:02:55.450 ⇒ 00:02:56.689 Awaish Kumar: I don’t know, like, I…
24 00:02:56.690 ⇒ 00:02:57.410 Henry Zhao: excitement.
25 00:02:57.410 ⇒ 00:03:09.409 Awaish Kumar: Yeah, I would love to, like, separate it… like, you can start, Hannah, asking questions, all the, like, business… business-related questions, like…
26 00:03:09.730 ⇒ 00:03:22.160 Awaish Kumar: Henry will, like, help you, like, help us with answering them, and then when it comes to the technical part of it, like, sometimes, like, how I, for example, connected
27 00:03:22.340 ⇒ 00:03:28.919 Awaish Kumar: a segment, CIO, and that script, which is coming in for unifying the
28 00:03:29.080 ⇒ 00:03:34.409 Awaish Kumar: profiles, so yeah, I can chime in on the technical part of it.
29 00:03:34.410 ⇒ 00:03:54.299 Hannah Wang: Sure. Yeah, let’s… let’s do that. I’ve done case studies with both of you, so it’s gonna be, the same set of questions. And then also, just later, when we move on to the next case study, I’m thinking creating… of creating another, Zoom meeting, just so that they’re segmented and separated for me to…
30 00:03:54.300 ⇒ 00:04:01.330 Hannah Wang: So, let’s start with the Eden stuff. I guess they’re both Eden, but anyway.
31 00:04:01.560 ⇒ 00:04:07.319 Hannah Wang: This is… okay, so… pulling up the questions,
32 00:04:08.020 ⇒ 00:04:15.660 Hannah Wang: Alright, so when did we start the project, and how long did it take, and who were the team members?
33 00:04:16.610 ⇒ 00:04:20.570 Henry Zhao: Oh, wait, do you know when we started it? I don’t know, because it was started before I joined.
34 00:04:21.649 ⇒ 00:04:26.549 Awaish Kumar: So on the C… like, like, we… Like, we started it…
35 00:04:26.799 ⇒ 00:04:29.169 Awaish Kumar: I think in the, like,
36 00:04:29.319 ⇒ 00:04:35.179 Awaish Kumar: maybe in… somewhere in August, but the thing is that it’s… it was not, like,
37 00:04:35.469 ⇒ 00:04:42.789 Awaish Kumar: Like, we are going all in on this project. So, like, the actual time we spend on this is…
38 00:04:43.449 ⇒ 00:04:47.949 Awaish Kumar: It really must be, like, around, I think, 4 weeks.
39 00:04:48.310 ⇒ 00:04:49.540 Hannah Wang: 4 weeks, okay.
40 00:04:49.760 ⇒ 00:04:50.470 Awaish Kumar: Yeah.
41 00:04:51.990 ⇒ 00:04:55.279 Hannah Wang: And then who did what? And who was the PM? I’m assuming…
42 00:04:55.280 ⇒ 00:04:57.160 Henry Zhao: I would say Robert was a PM,
43 00:04:57.700 ⇒ 00:05:00.609 Henry Zhao: And I was the, like, liaison with marketing.
44 00:05:01.630 ⇒ 00:05:02.290 Hannah Wang: Okay.
45 00:05:03.310 ⇒ 00:05:18.660 Hannah Wang: Cool, so I just want to understand the context of everything that was happening before we came in to help them implement all this stuff. So, yeah, what was the context behind them wanting this type of help, and what… yeah, go ahead.
46 00:05:18.660 ⇒ 00:05:32.510 Henry Zhao: Yeah, so yeah, their email marketing platform was CustomerIO, but in there, it was very messy, so there was, like, a bajillion customer attributes, which is what, you know, the marketers build campaigns off of, right? Like.
47 00:05:32.510 ⇒ 00:05:44.380 Henry Zhao: Is it a first-time purchaser? Is it a churner? Is it a, you know, person with address? Those kind of attributes. There was, like, so many, like, there was hundreds of them with, like, A-B testing, cohorts, and…
48 00:05:44.500 ⇒ 00:05:50.470 Henry Zhao: There were, like, old attributes that weren’t being used anymore, they weren’t documented, so, like…
49 00:05:50.470 ⇒ 00:06:10.149 Henry Zhao: nobody knew what anything was, and at the same time, they were, like, moving on from their CTO, I think, who used to run these campaigns, and hired a new email marketer. And so, if you can imagine somebody coming in and seeing, like, these hundreds and thousands of attributes… not hundreds and thousands, but, like, tons of attributes, like, are not gonna know where to start, or if they’re even reliable.
50 00:06:10.150 ⇒ 00:06:13.719 Henry Zhao: So, the business context was we needed to clean that up and streamline it.
51 00:06:14.380 ⇒ 00:06:22.870 Hannah Wang: And could you explain to me… I asked… I ask people this all the time, but what is attribution? I, like, keep… I guess I keep forgetting what it is.
52 00:06:22.870 ⇒ 00:06:24.650 Henry Zhao: This has nothing to do with attribution right now.
53 00:06:24.650 ⇒ 00:06:27.250 Hannah Wang: Or what was… what was that phrase you used?
54 00:06:27.250 ⇒ 00:06:28.440 Henry Zhao: attributes, yeah.
55 00:06:28.440 ⇒ 00:06:29.320 Hannah Wang: So what’s…
56 00:06:29.320 ⇒ 00:06:40.110 Awaish Kumar: Attributes, yeah, that’s, like, characteristics of a person, like, if I have an email, a phone number, or I’m a first-time purchaser, like, these are all attributes.
57 00:06:40.590 ⇒ 00:06:44.330 Hannah Wang: Okay, so attributes is different than attribution.
58 00:06:44.330 ⇒ 00:06:48.770 Henry Zhao: Right? Oh, okay, the words sound too similar.
59 00:06:48.770 ⇒ 00:06:52.099 Hannah Wang: And so what is attribution, exactly?
60 00:06:52.100 ⇒ 00:07:12.000 Henry Zhao: Attribution is, like, figuring out… first of all, I’m gonna present on this tomorrow in the retro, but attribution is, like, figuring out who gets credit for bringing on new customers. So, if you think about, like, when you become a customer somewhere, you probably have seen ads on Facebook, on TV, you probably then go to a store and there’s a salesperson, and then you look at the price.
61 00:07:12.330 ⇒ 00:07:16.109 Henry Zhao: Who actually gets the credit for converting that sale, so you know where to invest.
62 00:07:16.820 ⇒ 00:07:24.050 Hannah Wang: Mmm, got it. Okay, and were… were there any previous efforts to mitigate
63 00:07:24.500 ⇒ 00:07:27.589 Hannah Wang: this issue that we are facing, like, I’m assuming…
64 00:07:27.590 ⇒ 00:07:28.290 Henry Zhao: No off.
65 00:07:28.450 ⇒ 00:07:29.740 Hannah Wang: Sorry? Not that you know it.
66 00:07:29.740 ⇒ 00:07:30.440 Henry Zhao: Norm.
67 00:07:30.440 ⇒ 00:07:35.020 Hannah Wang: Who, like, built the previous attributes, and why were there so many of them?
68 00:07:35.020 ⇒ 00:07:37.489 Henry Zhao: he was the CTO, and he just didn’t care, like, he just…
69 00:07:37.600 ⇒ 00:07:41.820 Henry Zhao: I was busy, so he just threw things together, and as long as it worked, it worked.
70 00:07:41.980 ⇒ 00:07:43.450 Hannah Wang: Got it. Okay.
71 00:07:43.600 ⇒ 00:07:44.620 Hannah Wang: Sad.
72 00:07:44.870 ⇒ 00:08:03.280 Hannah Wang: Alright, so, moving on to, I guess, the challenge, now that we have more context, like, what… I know you kind of already mentioned this in the context section, but yeah, just explicitly mention what specific problems, like, the business was facing day-to-day because of these poor attributes stuff.
73 00:08:03.800 ⇒ 00:08:07.749 Henry Zhao: So not easily able to create segments, and
74 00:08:08.020 ⇒ 00:08:09.989 Henry Zhao: target the right people, I guess?
75 00:08:11.480 ⇒ 00:08:13.899 Hannah Wang: So what do segments mean?
76 00:08:14.440 ⇒ 00:08:16.450 Henry Zhao: Like, just a group of people, I would say.
77 00:08:16.670 ⇒ 00:08:17.660 Awaish Kumar: Okay.
78 00:08:17.700 ⇒ 00:08:23.489 Henry Zhao: like, this segment would be new… people that signed up for the email list a week ago, right? That could be a segment.
79 00:08:24.260 ⇒ 00:08:30.100 Henry Zhao: But how would I do that? I don’t… if the… if it’s really messy, I wouldn’t know how to even… even do that, you know?
80 00:08:31.290 ⇒ 00:08:32.010 Hannah Wang: Got it.
81 00:08:32.870 ⇒ 00:08:34.530 Hannah Wang: Okay.
82 00:08:34.530 ⇒ 00:08:48.660 Awaish Kumar: It will cost us, like, it will cost us, in terms of dollars, if we don’t have a proper segment, we are targeting wrong people with our campaigns, and we are burning money on the people who are… might not convert.
83 00:08:48.660 ⇒ 00:08:50.549 Hannah Wang: Got it, I see.
84 00:08:50.550 ⇒ 00:08:56.450 Henry Zhao: And we’re spamming them too, right? If we’re sending bad emails to the wrong people, we’re also spamming these poor people.
85 00:08:57.430 ⇒ 00:09:08.450 Hannah Wang: That makes me think about how many campaigns I’m a part of, and how many times I unsubscribe. Like, yeah, I never really thought that there was, like… obviously there’s strategy behind campaigns, but I never really…
86 00:09:08.540 ⇒ 00:09:22.729 Hannah Wang: thought deeply about it until I started talking to people about the work that they do to solve these, so… fascinating. I tend to just unsubscribe to everything, unfortunately, but… Yeah. Okay.
87 00:09:23.260 ⇒ 00:09:43.120 Hannah Wang: And… okay, yeah, kind of touched on everything. Alright, so, moving on to the solution part, and like I mentioned every time, feel free to share your screen if that helps you explain it, but yeah, from end to end, just describe to me the technical nitty-gritty of how we solved this attribute.
88 00:09:43.350 ⇒ 00:09:45.080 Hannah Wang: issue.
89 00:09:45.080 ⇒ 00:09:45.900 Awaish Kumar: Basically, we…
90 00:09:45.900 ⇒ 00:09:46.810 Henry Zhao: That’ll be a wish now.
91 00:09:46.810 ⇒ 00:10:00.749 Awaish Kumar: Yeah, so basically what we did is we basically got the data, from multiple places. One is segment. What segment does is basically, joins all the
92 00:10:01.260 ⇒ 00:10:09.800 Awaish Kumar: Data, which is coming from different, different sources into one database, which it calls, like,
93 00:10:09.960 ⇒ 00:10:16.759 Awaish Kumar: unify. That means it basically segment itself unifies some of the data for the…
94 00:10:17.330 ⇒ 00:10:26.370 Awaish Kumar: for the people who are coming in from different sources. Then we, like, and we… then we clean it up, like, that… basically…
95 00:10:26.520 ⇒ 00:10:37.069 Awaish Kumar: In that process, Segment just, like, collects all the attributes it thinks are there, but it might be null, they are, like,
96 00:10:37.750 ⇒ 00:10:46.699 Awaish Kumar: like, garbage values, like, which doesn’t make sense, or not relevant, or things like that. So what I did is, I did an exercise
97 00:10:47.680 ⇒ 00:10:53.210 Awaish Kumar: Of figuring out what are the actual meaningful attributes.
98 00:10:53.210 ⇒ 00:10:53.560 Hannah Wang: Mmm.
99 00:10:53.560 ⇒ 00:11:01.960 Awaish Kumar: which we can… like, the subset of meaningful attributes which I can get from here. So, for that.
100 00:11:02.020 ⇒ 00:11:15.740 Awaish Kumar: I just have to go in and understand the distribution of each of the columns. So I have a script written in Python, which runs on Daxter, to create a table called user profiles. That basically figures out what are the
101 00:11:15.740 ⇒ 00:11:26.300 Awaish Kumar: meaningful… which one are the… which ones are the meaningful attributes? So, that means, like, we don’t have… we have minimal amount of nulls where we are getting the…
102 00:11:26.340 ⇒ 00:11:37.590 Awaish Kumar: recent values, or updated values, or things like that. And that basically helps me to, from thousands of attributes, like short, like the…
103 00:11:38.350 ⇒ 00:11:44.660 Awaish Kumar: I get, like, a subset of columns, which are maybe around… still around… Oh…
104 00:11:44.880 ⇒ 00:11:47.379 Awaish Kumar: 50, 100, or something like that.
105 00:11:48.460 ⇒ 00:11:49.630 Awaish Kumar: So, from…
106 00:11:49.850 ⇒ 00:12:01.230 Awaish Kumar: maybe, 500, 800 columns. I’m now down to a few columns, which are really meaningful. Then we enrich that table with our
107 00:12:01.550 ⇒ 00:12:16.319 Awaish Kumar: internal data, which is coming from our warehouse. So, for the people who already made a purchase with us, we do know some of the data for them, like email, like, everybody has to sign up.
108 00:12:16.320 ⇒ 00:12:27.099 Awaish Kumar: to become… to purchase anything from Eden. So basically, we know their names, email, phone numbers, and things like that. And sometimes we do have addresses, and…
109 00:12:27.330 ⇒ 00:12:39.869 Awaish Kumar: like, and other attributes, like, which are calculated attributes. So we enrich it with… we enrich what we got from segment with our internal data.
110 00:12:39.870 ⇒ 00:12:53.279 Awaish Kumar: Which we know is correct data, and then we basically calculate few more attributes on top of it, which comes from our internal data as well, like lifetime value of a customer, which is, like.
111 00:12:53.820 ⇒ 00:13:11.899 Awaish Kumar: maybe a customer made 100 orders, so we know, like, how much he spent with Edit, right? So similarly, how long he has been with Eden, and things like that. These are called, like, calculated attributes, so he calculated them and created another table, which basically includes all of it together in one table.
112 00:13:11.960 ⇒ 00:13:13.500 Awaish Kumar: And then…
113 00:13:13.560 ⇒ 00:13:26.489 Awaish Kumar: Again, I think, segment is used, basically, for reverse ETL as well, to read the data from that table, and then we send the attributes which are needed to be sent to the customer I.O.
114 00:13:26.490 ⇒ 00:13:37.740 Awaish Kumar: And when those clean, calculated, and meaningful attributes are sent to the CIO, that can help us, the teams in this, which are working on CIO platform, to basically
115 00:13:37.840 ⇒ 00:13:42.489 Awaish Kumar: Create some meaningful segments from those attributes, like…
116 00:13:42.540 ⇒ 00:14:00.779 Awaish Kumar: So now you have… you can have a… you have a column, which we know is working, which says first-time customer or a returning customer, and basically using that, they will create some group of people, we call it segment, and then send the emails only to those people in their campaigns.
117 00:14:00.780 ⇒ 00:14:06.300 Hannah Wang: Got it. And was… was this for, like, because I know Eden has a bunch of, like.
118 00:14:06.550 ⇒ 00:14:09.979 Hannah Wang: products and services, so was this for their, like.
119 00:14:09.980 ⇒ 00:14:20.270 Awaish Kumar: This is for Paul, like, this exercise is basically up to the, like, the creating that table and putting it into the CIO. We… we just…
120 00:14:20.500 ⇒ 00:14:26.350 Awaish Kumar: Put it for all the… Customers, which basically…
121 00:14:26.560 ⇒ 00:14:39.659 Awaish Kumar: are the customers of the agent, right? Doesn’t matter what product they purchase. But when they… when the campaigns are created, they can be, like, dependent on product level.
122 00:14:39.890 ⇒ 00:14:53.399 Awaish Kumar: filters, like, okay, only those people who came first time and purchased this product, we are only targeting those. So, yeah, that is, like, at a capel level. These things happen at a campaign level.
123 00:14:54.290 ⇒ 00:14:55.020 Hannah Wang: Got it
124 00:14:55.760 ⇒ 00:15:05.089 Hannah Wang: Okay, and the tools that I’ve tried to… so, basically, the tools that were used, what I heard was Segment, Customer I.O, DAX.
125 00:15:05.090 ⇒ 00:15:15.489 Awaish Kumar: Yeah, I can say Python as a language, Texture is a tool, and then Segment is another tool, then we have CIO.
126 00:15:15.490 ⇒ 00:15:15.900 Hannah Wang: Okay.
127 00:15:16.000 ⇒ 00:15:18.570 Awaish Kumar: And then BBT, also.
128 00:15:18.570 ⇒ 00:15:19.720 Hannah Wang: ET, okay.
129 00:15:20.040 ⇒ 00:15:22.039 Hannah Wang: And then…
130 00:15:22.720 ⇒ 00:15:32.210 Hannah Wang: Was… sorry if this is, like, a dumb question, but was there any dashboard that was created? Or… N-no.
131 00:15:32.210 ⇒ 00:15:34.639 Awaish Kumar: Not in this process, yeah, nope.
132 00:15:35.210 ⇒ 00:15:35.860 Hannah Wang: Okay.
133 00:15:36.180 ⇒ 00:15:40.759 Hannah Wang: Alright, so, moving on to the results, I guess…
134 00:15:41.190 ⇒ 00:15:51.740 Hannah Wang: Yeah, if there’s any metrics that you can give me, that’d be helpful, but if not, just, like, what feedback did we get? Like, what was the impact of
135 00:15:51.990 ⇒ 00:15:54.140 Hannah Wang: Of this project.
136 00:15:54.140 ⇒ 00:15:57.910 Awaish Kumar: Yeah, that… Henry, if you can… And something.
137 00:15:58.210 ⇒ 00:16:07.180 Henry Zhao: Yeah, so we reduced, the amount… let me see how many we ended up having. We ended up going from 400 attributes to… let’s see how many we ended up having.
138 00:16:07.790 ⇒ 00:16:08.820 Henry Zhao: Customer…
139 00:16:12.310 ⇒ 00:16:13.909 Henry Zhao: Where would it be?
140 00:16:17.360 ⇒ 00:16:19.370 Henry Zhao: Research customer I.O. attributes.
141 00:16:22.610 ⇒ 00:16:31.549 Henry Zhao: Just put X for now, so that I can continue talking. Basically, we had 400 attributes, we brought it down to, like, 45 attributes that we… oh, I found it, I found it, I found it, I found it
142 00:16:32.310 ⇒ 00:16:45.580 Henry Zhao: to 40 attributes that we want to put into Customer I.O. We also added some additional helpful attributes, like if they have a Zendesk ticket open or not, that way we don’t ask people for satisfaction
143 00:16:45.580 ⇒ 00:16:57.459 Henry Zhao: reviews if they’re still, like, trying to resolve an issue. We also use Customer I.O. to let us know when they need a reminder to renew their product, so that obviously
144 00:16:57.650 ⇒ 00:16:59.479 Henry Zhao: Brings in more renewals.
145 00:16:59.560 ⇒ 00:17:01.980 Hannah Wang: I haven’t pulled data on how much that actually.
146 00:17:01.980 ⇒ 00:17:06.540 Henry Zhao: Affected, but you can add a task for me to look at that if you need that data.
147 00:17:06.540 ⇒ 00:17:07.089 Hannah Wang: Okay.
148 00:17:07.459 ⇒ 00:17:12.049 Henry Zhao: Or I guess I can just easily pull how many people got those reminders.
149 00:17:12.869 ⇒ 00:17:26.379 Henry Zhao: And just associate how much renewal revenue we have from there. So that was one thing, and then we were able to train Judd, the new email marketer, on how to use these attributes and create cleaner, campaign flows.
150 00:17:27.959 ⇒ 00:17:28.879 Hannah Wang: Awesome.
151 00:17:29.119 ⇒ 00:17:32.199 Hannah Wang: Okay, great. I think…
152 00:17:32.359 ⇒ 00:17:46.329 Hannah Wang: I think that was good. We can hop on to another call. I’ll make a Zoom call and message you guys, so I’m just gonna… let’s hop off this meeting, and then… or… actually, for the next one, which is…
153 00:17:46.559 ⇒ 00:17:50.669 Hannah Wang: the Catalyst work that we did for Eden does…
154 00:17:51.029 ⇒ 00:17:53.939 Hannah Wang: Awash need to be there, or…
155 00:17:53.999 ⇒ 00:17:55.519 Henry Zhao: You can hop off.
156 00:17:57.079 ⇒ 00:17:58.389 Hannah Wang: Sorry, what was that?
157 00:17:58.390 ⇒ 00:17:59.100 Henry Zhao: Yes.
158 00:17:59.100 ⇒ 00:18:02.170 Hannah Wang: Yes? Okay. Okay, cool. I’ll make a new Zoom. See you guys there.
159 00:18:02.490 ⇒ 00:18:02.970 Henry Zhao: Yeah.