Meeting Title: Eden Marketing Data Model Discussion Date: 2025-06-26 Meeting participants: Annie Yu, Robert Tseng
WEBVTT
1 00:01:13.420 ⇒ 00:01:14.750 Annie Yu: Hello, Robert!
2 00:01:18.290 ⇒ 00:01:19.160 Robert Tseng: Can you hear me?
3 00:01:19.875 ⇒ 00:01:21.890 Robert Tseng: Yeah, okay.
4 00:01:22.750 ⇒ 00:01:30.235 Robert Tseng: And cool. Well, I I think we’re both on the same dock. I guess.
5 00:01:31.090 ⇒ 00:01:37.439 Robert Tseng: yeah. From this time. We’re just trying to pick it out like what this would look like. And then.
6 00:01:37.820 ⇒ 00:01:43.260 Robert Tseng: yeah, so we’ll just kind of talk through both options in a bit more detail, and then I’ll break that up into tickets.
7 00:01:44.230 ⇒ 00:01:44.780 Annie Yu: Yeah.
8 00:01:45.300 ⇒ 00:01:49.770 Robert Tseng: Yeah, so we don’t have to do it in camera. Now, I can do that afterwards. But yeah, I guess
9 00:01:50.200 ⇒ 00:01:56.960 Robert Tseng: I was looking through option. One would have the lifetime and the revenue assumption of time
10 00:01:57.520 ⇒ 00:02:05.230 Robert Tseng: kind of have something like that already, or I guess maybe what would be helpful is, I’ll just pull up what we already have low, and then we can kind of talk through it as well.
11 00:02:05.490 ⇒ 00:02:06.020 Robert Tseng: Apple.
12 00:02:06.020 ⇒ 00:02:12.539 Annie Yu: Yeah. But yeah, this is the the top section of that marketing dashboard that we updated
13 00:02:13.020 ⇒ 00:02:14.500 Annie Yu: couple weeks ago.
14 00:02:14.840 ⇒ 00:02:17.000 Robert Tseng: Yeah, okay.
15 00:02:21.070 ⇒ 00:02:22.080 Robert Tseng: let me see.
16 00:02:49.580 ⇒ 00:02:50.280 Robert Tseng: Yeah.
17 00:02:57.330 ⇒ 00:02:59.510 Robert Tseng: Oh, oh.
18 00:03:02.530 ⇒ 00:03:06.194 Robert Tseng: okay, cool. So I haven’t put this all up side by side.
19 00:03:09.520 ⇒ 00:03:11.500 Robert Tseng: yeah, this is what we’re gonna do.
20 00:03:21.780 ⇒ 00:03:33.709 Robert Tseng: And then the longer term phase one is, yeah, figuring out what’s going to correlate with. Ltv.
21 00:03:36.330 ⇒ 00:03:42.729 Robert Tseng: yeah, I think this makes sense to me. I mean, obviously the product that you choose to come in on matters
22 00:03:43.320 ⇒ 00:03:48.180 Robert Tseng: the plan type as well. Getting people to do longer plans in the start.
23 00:03:48.510 ⇒ 00:03:50.349 Robert Tseng: But that was typically better.
24 00:03:52.470 ⇒ 00:03:58.262 Robert Tseng: I think something to take note of. We can look at
25 00:04:00.340 ⇒ 00:04:02.860 Robert Tseng: with people, you know, at school.
26 00:04:03.690 ⇒ 00:04:10.039 Robert Tseng: larger plans to ask anything else over here right now.
27 00:04:11.410 ⇒ 00:04:12.210 Robert Tseng: That’d be.
28 00:04:14.360 ⇒ 00:04:15.560 Robert Tseng: We work with a file.
29 00:04:19.370 ⇒ 00:04:24.150 Robert Tseng: We’ll sign up. Not really sure, I guess, is that related to community.
30 00:04:25.873 ⇒ 00:04:31.570 Annie Yu: No, I I think I just saw that we had a column.
31 00:04:32.500 ⇒ 00:04:39.569 Annie Yu: That’s a sign up. But yeah, I barely use it. And I was just like laying out some example columns.
32 00:04:40.292 ⇒ 00:04:43.920 Annie Yu: Obviously, we have more to consider. There.
33 00:04:44.740 ⇒ 00:04:51.800 Robert Tseng: Okay. Well, I mean, I think we might as well just try to think about what those are. So.
34 00:04:55.180 ⇒ 00:04:57.120 Robert Tseng: And this is non-member.
35 00:04:57.560 ⇒ 00:05:00.349 Robert Tseng: Well, I don’t want to know about this. I feel like everyone has to sign up
36 00:05:05.110 ⇒ 00:05:12.870 Robert Tseng: schedule as well. I mean, I feel like these are going to be conflating between numbers of clients in the schedule, together.
37 00:05:13.080 ⇒ 00:05:16.330 Annie Yu: And obviously, yeah, should we like demographics and pharmacy as well.
38 00:05:17.050 ⇒ 00:05:17.710 Robert Tseng: Okay?
39 00:05:18.410 ⇒ 00:05:21.630 Robert Tseng: And so yeah, we’re gonna run a what like a.
40 00:05:22.820 ⇒ 00:05:30.550 Robert Tseng: some sort of regression model to kind of see improve formation. And then, well, there’s like a
41 00:05:32.710 ⇒ 00:05:34.050 Robert Tseng: no that’s so much different.
42 00:05:34.050 ⇒ 00:05:47.129 Annie Yu: For for this face one, I was thinking. It’s more so like explore exploration, like you said, like, what’s the correlation between each field to the Ltv. And then
43 00:05:47.330 ⇒ 00:05:52.339 Annie Yu: with that information we we then do the regression model.
44 00:05:53.000 ⇒ 00:05:54.059 Robert Tseng: Yes, okay.
45 00:05:56.470 ⇒ 00:05:58.069 Robert Tseng: Alright, that makes sense to me.
46 00:06:01.660 ⇒ 00:06:06.780 Robert Tseng: Yeah. Then afterwards we’re gonna be purchase.
47 00:06:12.840 ⇒ 00:06:14.670 Robert Tseng: I think we
48 00:06:22.150 ⇒ 00:06:23.930 Robert Tseng: yeah, I mean, this looks pretty.
49 00:06:24.770 ⇒ 00:06:25.670 Robert Tseng: I can go.
50 00:06:29.630 ⇒ 00:06:37.079 Robert Tseng: Yeah. So I guess, what would you look, how? What would deployment look like in this dashboard? Or like.
51 00:06:37.560 ⇒ 00:06:40.279 Robert Tseng: yeah, like, kind of what’s what’s like? The
52 00:06:41.210 ⇒ 00:06:43.429 Robert Tseng: what’s the end result of what this is gonna look like.
53 00:06:47.180 ⇒ 00:06:51.029 Annie Yu: I I could be wrong, but the way I’m thinking it is
54 00:06:51.240 ⇒ 00:06:58.290 Annie Yu: so we would be able to tie an Ltv. To each customer. Id.
55 00:06:59.580 ⇒ 00:07:02.449 Robert Tseng: Yeah, I mean, that would be ideal. Yeah.
56 00:07:03.630 ⇒ 00:07:06.313 Annie Yu: Yeah, if it’s accurate enough.
57 00:07:09.860 ⇒ 00:07:19.449 Annie Yu: Yeah, that’s what that’s thinking. So so the end result would probably not look like, I guess if we want to have similar
58 00:07:19.790 ⇒ 00:07:25.840 Annie Yu: visuals to the to the left side. We we would then have kind of predictive value
59 00:07:26.090 ⇒ 00:07:29.380 Annie Yu: building out those empty cells.
60 00:07:31.070 ⇒ 00:07:31.600 Robert Tseng: Yeah.
61 00:07:40.420 ⇒ 00:07:41.180 Robert Tseng: okay.
62 00:07:41.810 ⇒ 00:07:52.149 Robert Tseng: so yeah, so that would be like the act like the aggregated kind of predictive fields here. So you can kind of see what that would look like. You know our certain cohorts on property from the same example.
63 00:07:52.630 ⇒ 00:07:56.769 Robert Tseng: This one is lower at 6 months than it was, you know previous months.
64 00:07:59.050 ⇒ 00:07:59.930 Robert Tseng: Oh, gosh!
65 00:08:00.660 ⇒ 00:08:03.118 Robert Tseng: So I think where this fits in.
66 00:08:04.250 ⇒ 00:08:08.970 Robert Tseng: we’re doing like the Cdp evaluation right now. One of the things that we’re testing in the pilot.
67 00:08:09.540 ⇒ 00:08:16.512 Robert Tseng: I we’re gonna I’m gonna bring in a couple of competed traits into the customer as well. And
68 00:08:17.710 ⇒ 00:08:22.410 Robert Tseng: I don’t know how much data. I’m gonna sink into each of these tools. But I may
69 00:08:24.020 ⇒ 00:08:28.846 Robert Tseng: like, my point is like, I want, maybe I’ll try this. And then, like segment has its own, like
70 00:08:30.820 ⇒ 00:08:33.895 Robert Tseng: version of a predicted Ltv thing as well.
71 00:08:34.280 ⇒ 00:08:35.899 Annie Yu: Oh! There is already one.
72 00:08:36.440 ⇒ 00:08:39.049 Robert Tseng: No, I mean, I mean, yeah, there is. But like, I don’t think
73 00:08:39.470 ⇒ 00:08:44.860 Robert Tseng: it’s kind of a black box. It’s just like a button that a marketer would press. That’s like I could show you what it looks like.
74 00:08:46.780 ⇒ 00:08:50.579 Robert Tseng: So I still think it’s better for us to understand what we’re doing.
75 00:08:52.724 ⇒ 00:08:58.820 Robert Tseng: So look at this.
76 00:09:00.450 ⇒ 00:09:04.650 Robert Tseng: Oh, newer, yeah, there’s something here about like
77 00:09:09.420 ⇒ 00:09:12.369 Annie Yu: This is a system that we pay for.
78 00:09:12.630 ⇒ 00:09:18.649 Robert Tseng: Yeah, we already paid for this. Well, we’re I mean, I’m evaluating whether or not to stay on the system or not. But
79 00:09:19.250 ⇒ 00:09:26.730 Robert Tseng: yeah, I think it pretty much just uses you know. The order completed, events, totals, and then like time. And then it has, like.
80 00:09:27.990 ⇒ 00:09:35.820 Robert Tseng: I’m guessing a a lot of our historical data of of orders per customer already flowing through this. And so let’s default using to calculate this.
81 00:09:38.760 ⇒ 00:09:39.340 Robert Tseng: Yeah.
82 00:09:39.970 ⇒ 00:09:47.889 Robert Tseng: I mean, I don’t necessarily think it’s I don’t exactly know what what model they’re running under underneath it. So I think that’s kind of where it’s kind of a black box.
83 00:09:48.060 ⇒ 00:09:59.300 Robert Tseng: but I mean I would, Jet. I would expect that, you know I would use this as like a benchmark for, like whatever you’re doing like, I do think we’re over reporting. It’s I’ve looked at this number before 700.
84 00:09:59.430 ⇒ 00:10:03.873 Robert Tseng: I don’t actually think it’s 1,200, or whatever the recovery takes, it is so.
85 00:10:05.270 ⇒ 00:10:16.330 Robert Tseng: But anyway, I haven’t. We haven’t validated it because we haven’t run our own analysis to kind of figure that out. But yeah, every Cdp tool that we’re evaluating kind of has something like this. They have like
86 00:10:17.760 ⇒ 00:10:19.310 Robert Tseng: ways to.
87 00:10:19.940 ⇒ 00:10:23.350 Annie Yu: Usually like 12 months, or how long.
88 00:10:24.530 ⇒ 00:10:28.590 Robert Tseng: No, it’s not even 12 months. They’re just doing remark.
89 00:10:31.180 ⇒ 00:10:31.820 Annie Yu: Okay.
90 00:10:31.970 ⇒ 00:10:40.550 Annie Yu: yeah, I think 3 months, probably easier than 12 months. Just because you, you can have less data to to inform the model.
91 00:10:40.900 ⇒ 00:10:41.510 Robert Tseng: Yeah.
92 00:10:45.120 ⇒ 00:10:52.269 Robert Tseng: yeah, I haven’t looked too much in here, like how I would do this. But they have different types of things cross cell position.
93 00:10:53.840 ⇒ 00:10:55.880 Robert Tseng: Fern and Ltv prediction.
94 00:10:58.010 ⇒ 00:11:01.180 Robert Tseng: I guess there’s like documentation and all of that, or whatever. But
95 00:11:02.930 ⇒ 00:11:04.960 Robert Tseng: yeah, I don’t know. I would.
96 00:11:06.800 ⇒ 00:11:12.830 Robert Tseng: I guess you’re gonna be out next week. So I don’t know if you’ll really have time to review all this, but hopefully.
97 00:11:15.690 ⇒ 00:11:22.019 Annie Yu: Okay. So the I guess before I get started, I should review this.
98 00:11:22.310 ⇒ 00:11:24.870 Annie Yu: Is that what you were recommending.
99 00:11:27.030 ⇒ 00:11:34.879 Robert Tseng: Yeah, I think you should kind of get familiar with this. Let me know, like what’s actually going on here, if you think it’s sufficient. And then.
100 00:11:37.590 ⇒ 00:11:39.229 Robert Tseng: yeah, I mean, I think
101 00:11:39.980 ⇒ 00:11:46.169 Robert Tseng: that could be. I mean, it doesn’t. Yeah. I just want to know what the what the restraints are.
102 00:11:48.710 ⇒ 00:11:53.441 Robert Tseng: Whether or not you trust it, whether you feel like you would rather run your own analysis.
103 00:11:55.990 ⇒ 00:11:56.600 Robert Tseng: Yeah.
104 00:12:00.250 ⇒ 00:12:06.060 Annie Yu: Okay. And this is so. Will you be able to share that link to me?
105 00:12:07.060 ⇒ 00:12:11.330 Robert Tseng: Oh, segment. Yeah, I think you should have it. It’s all kind of in one pass.
106 00:12:13.060 ⇒ 00:12:13.700 Annie Yu: Okay.
107 00:12:15.530 ⇒ 00:12:20.090 Robert Tseng: Yeah, but you just go into unify. Go to trade. So you can kind of stuff to look at some of these
108 00:12:21.010 ⇒ 00:12:22.829 Robert Tseng: to make change of their recommendations.
109 00:12:24.180 ⇒ 00:12:26.775 Annie Yu: Yeah, yeah, yeah. I’m I’m curious how they
110 00:12:27.350 ⇒ 00:12:31.840 Annie Yu: like, what kind of features they they put it in for their model.
111 00:12:32.400 ⇒ 00:12:32.970 Robert Tseng: Yeah.
112 00:12:36.610 ⇒ 00:12:41.190 Annie Yu: And they do have all the data that we have. Is that correct?
113 00:12:43.010 ⇒ 00:12:48.910 Robert Tseng: no, I mean, they only have. They have all the order data like you can see, like, the only thing that they’re using is the order data.
114 00:12:50.025 ⇒ 00:12:50.780 Annie Yu: got it?
115 00:12:50.780 ⇒ 00:12:55.920 Robert Tseng: Yeah, which is, I don’t know why you’re really
116 00:12:56.490 ⇒ 00:13:01.070 Robert Tseng: nate like, I think, for, like the most basic Ltv prediction, it’s just like a
117 00:13:01.390 ⇒ 00:13:05.124 Robert Tseng: buy till you die model, which is just using
118 00:13:05.810 ⇒ 00:13:13.957 Robert Tseng: recency frequency and like amount or whatever I think, those those 3 things kind of get you
119 00:13:15.060 ⇒ 00:13:18.580 Robert Tseng: like a basic predictable. L, 3 model. Anyways.
120 00:13:19.490 ⇒ 00:13:21.490 Annie Yu: Yeah, yeah. And if
121 00:13:22.030 ⇒ 00:13:28.749 Annie Yu: okay, if that’s the case, we and we will decide if we are okay with that.
122 00:13:29.720 ⇒ 00:13:32.720 Annie Yu: or or if we should develop our own right.
123 00:13:33.140 ⇒ 00:13:36.850 Robert Tseng: Yeah, yeah, I kind of want your evaluation on this. And then.
124 00:13:37.120 ⇒ 00:13:46.359 Robert Tseng: yeah, if you’re like, okay, this is good for only the 3 month prediction be great, we can use it. And then, if you’re great, if you can kind of help, I mean, you could deploy that too. And
125 00:13:46.829 ⇒ 00:13:50.060 Robert Tseng: I mean, the point is like we already pay for these tools. So like kind of like.
126 00:13:50.060 ⇒ 00:13:50.400 Annie Yu: Yeah.
127 00:13:50.400 ⇒ 00:14:01.930 Robert Tseng: Well, I’d like to get more out of them. And if not, and you really don’t like them, and they feel like doesn’t make sense, and you’d rather run your own custom model. Then we could do that. But I feel like.
128 00:14:02.620 ⇒ 00:14:03.899 Robert Tseng: yeah, okay, to.
129 00:14:04.330 ⇒ 00:14:07.280 Annie Yu: Yeah, yeah, that makes sense. Yeah, yeah.
130 00:14:08.010 ⇒ 00:14:15.759 Annie Yu: And so if if let’s say, like, if this is good, we will have a way to bring those predictive values into our data models.
131 00:14:16.410 ⇒ 00:14:20.610 Robert Tseng: Yeah, so this is kind of like the customer data model. The idea is every customer.
132 00:14:21.275 ⇒ 00:14:24.184 Robert Tseng: We’ll have all of these different properties.
133 00:14:26.410 ⇒ 00:14:32.667 Robert Tseng: I mean, we don’t really use these like all of them right now. But yeah, I mean, I had.
134 00:14:33.460 ⇒ 00:14:37.210 Robert Tseng: I think you’re right, like, I think the deliverable is like bringing that
135 00:14:37.410 ⇒ 00:14:44.010 Robert Tseng: calculated field, but bringing it into a customer profile that’s models
136 00:14:44.458 ⇒ 00:14:47.801 Robert Tseng: cause. Then you can use it for any kind of segmentation. And
137 00:14:48.770 ⇒ 00:14:53.419 Robert Tseng: yeah, it’s not okay. Yeah, like that. That will. I think that should that should work for.
138 00:14:54.330 ⇒ 00:14:55.460 Annie Yu: Okay. Okay?
139 00:14:55.650 ⇒ 00:15:01.810 Annie Yu: Oh, that’s that’s kind of cool. Okay, I did not know we had something like this. So it’s cool.
140 00:15:02.310 ⇒ 00:15:08.110 Robert Tseng: But that’s yeah. No, I think I would would like you to explore it, and I mean.
141 00:15:12.180 ⇒ 00:15:20.360 Annie Yu: Okay. So for my I guess the the the next kind of deliverable would be my
142 00:15:20.540 ⇒ 00:15:23.930 Annie Yu: might. Take away from from this right and compare.
143 00:15:23.930 ⇒ 00:15:32.129 Robert Tseng: Yeah, I mean, ideally, if it’s just kind of like more digging into this, if you can get me an answer before you head out. That’d be great. Because then.
144 00:15:32.390 ⇒ 00:15:38.649 Robert Tseng: and we’re evaluating segment again for renewal. We have 2 other tools that we’re looking at. We have ours privacy.
145 00:15:39.690 ⇒ 00:15:41.419 Robert Tseng: or what else got that wrong?
146 00:15:45.270 ⇒ 00:15:47.540 Robert Tseng: And we’ll evaluate a better stack.
147 00:15:49.700 ⇒ 00:15:53.210 Robert Tseng: So I mean, they both kind of have.
148 00:15:53.440 ⇒ 00:15:54.989 Robert Tseng: Okay, so that one’s okay.
149 00:15:56.080 ⇒ 00:16:00.420 Annie Yu: So this one also had similar protective.
150 00:16:01.030 ⇒ 00:16:18.579 Robert Tseng: Yeah. So I’m moving these into demo like, we’re gonna demo these, I’m I’m basically gonna hit the tools against each other. And and there’s a couple of use cases that we’re testing. One is for audience building, which is, or like kind of customer profile switching. So
151 00:16:18.730 ⇒ 00:16:20.179 Robert Tseng: kind of this work of?
152 00:16:20.320 ⇒ 00:16:29.519 Robert Tseng: Is it easy to get the setup customer profile like rates, and then push them into the warehouse, where we can
153 00:16:29.640 ⇒ 00:16:36.479 Robert Tseng: use some of the calculated fields that they already have, and then also enrich it with something that we have
154 00:16:36.590 ⇒ 00:16:41.509 Robert Tseng: so like. If I were to. Maybe it’s just like they make sure that all of these descriptive
155 00:16:41.959 ⇒ 00:16:49.079 Robert Tseng: properties they can be pushed into. We don’t have a customer data model, and that’s like, has that has all these things yet.
156 00:16:50.430 ⇒ 00:17:07.662 Robert Tseng: we’ll build one field that’s predicted. Ltv. In within the app within the tool. Push that into the warehouse as well. And then maybe there’s like one more trait that we calculate that has to use our warehouse data. And we have to do it. You know, from our models. Maybe it’s like
157 00:17:09.650 ⇒ 00:17:11.349 Robert Tseng: number of
158 00:17:12.670 ⇒ 00:17:22.719 Robert Tseng: cross product purchases or like, I don’t know we have. I have to kind of think about like, what’s what’s the the second example of one. So I want to pick 2 traits that we’re calculating.
159 00:17:22.960 ⇒ 00:17:26.530 Robert Tseng: Push it into the model and see which tool does it better.
160 00:17:27.094 ⇒ 00:17:28.830 Robert Tseng: And then the other.
161 00:17:29.080 ⇒ 00:17:32.190 Robert Tseng: The other test is
162 00:17:33.340 ⇒ 00:17:40.139 Robert Tseng: whether or not it’s easy to build. I guess they call them audiences. Every tool kind of has their own. But these are just like
163 00:17:41.780 ⇒ 00:17:46.820 Robert Tseng: you’ve got base figures that you’re using, or you can profile properties.
164 00:17:47.850 ⇒ 00:17:50.259 Robert Tseng: all product based things on that.
165 00:17:50.710 ⇒ 00:17:54.160 Annie Yu: Is it kind of like persona? Not really.
166 00:17:55.550 ⇒ 00:18:03.240 Robert Tseng: Like this. It would be an example. Someone had set this up before. It’s just using.
167 00:18:04.880 ⇒ 00:18:06.349 Robert Tseng: What is it using here?
168 00:18:09.260 ⇒ 00:18:10.350 Annie Yu: Oh, okay.
169 00:18:10.610 ⇒ 00:18:14.869 Robert Tseng: Yeah, there’s some random trait here that’s just like
170 00:18:15.450 ⇒ 00:18:20.749 Robert Tseng: purchase intent, and he’s assigned some value. I don’t know how it’s calculated, but somehow
171 00:18:21.080 ⇒ 00:18:30.330 Robert Tseng: this is, these are the cohort of customers that we believe are 60% likely to make another purchase in the next X number of days.
172 00:18:30.490 ⇒ 00:18:35.449 Robert Tseng: No one uses this right now, someone just built this. And it’s kind of like a side project. It’s not in the production.
173 00:18:36.730 ⇒ 00:18:40.800 Robert Tseng: But yeah, I think this is an example of an audience where we’re using
174 00:18:42.600 ⇒ 00:18:48.430 Robert Tseng: because the only type of triggers that the marketing team is using right now is like, 1st
175 00:18:48.620 ⇒ 00:19:00.319 Robert Tseng: send an email to customers who purchase a product in the last 7 days. It’s just like event based figures that are very simple and just tied to like a single transaction or something.
176 00:19:00.700 ⇒ 00:19:01.080 Annie Yu: Yeah.
177 00:19:01.080 ⇒ 00:19:07.149 Robert Tseng: But we’re not doing any sort of predictive audiences or like more complex audience building.
178 00:19:07.350 ⇒ 00:19:09.060 Robert Tseng: I think you can go.
179 00:19:09.490 ⇒ 00:19:12.330 Robert Tseng: Yeah. So I, you know, just thinking about like what
180 00:19:14.510 ⇒ 00:19:22.329 Robert Tseng: I mean. Ideally, I think there’s it’s it’s kind of an exploration exercise to is kind of figuring out doing the research figure out like, Okay, well, what are the
181 00:19:22.560 ⇒ 00:19:31.910 Robert Tseng: valid, what are valuable audiences that we can give to a marketing team and be like, yeah, you should be targeting these people? Because
182 00:19:32.240 ⇒ 00:19:38.570 Robert Tseng: of whatever reason. So, yeah, I think, like in the experiment, I need to kind of pick.
183 00:19:38.770 ⇒ 00:19:42.497 Robert Tseng: Okay, like one or 2 audiences that we can build
184 00:19:43.660 ⇒ 00:19:50.530 Robert Tseng: one using only the tool itself, like only using segment and then using another one only like using
185 00:19:50.700 ⇒ 00:19:54.249 Robert Tseng: a mix of the data that we’ll have the data warehouse.
186 00:19:56.730 ⇒ 00:19:59.400 Robert Tseng: Yeah. So and then
187 00:20:01.300 ⇒ 00:20:05.939 Robert Tseng: that that there’s so that’s that’s that’s what I’m evaluating. Over the next the next 2 weeks.
188 00:20:06.710 ⇒ 00:20:07.930 Annie Yu: Does that make sense?
189 00:20:08.620 ⇒ 00:20:19.690 Annie Yu: Yeah, I think what I can do today is review that predictive? Ltv, or is this segment.
190 00:20:20.070 ⇒ 00:20:20.939 Robert Tseng: Yes, that’s assign one.
191 00:20:20.940 ⇒ 00:20:23.800 Annie Yu: Try to figure out yeah how how they
192 00:20:24.710 ⇒ 00:20:27.580 Annie Yu: come up with it, and if it makes sense to us.
193 00:20:30.000 ⇒ 00:20:30.640 Robert Tseng: Okay.
194 00:20:37.900 ⇒ 00:20:39.480 Robert Tseng: Yeah. Would that be helpful?
195 00:20:40.050 ⇒ 00:20:41.901 Robert Tseng: Yes, yes, that would be helpful.
196 00:20:43.270 ⇒ 00:20:44.159 Annie Yu: But yeah, but.
197 00:20:44.160 ⇒ 00:20:46.269 Robert Tseng: Probably do some stuff.
198 00:20:46.270 ⇒ 00:20:54.590 Annie Yu: Hmm! But if if it’s not like clearly documented and I can’t find it, I’ll let you know. But yeah.
199 00:20:56.570 ⇒ 00:20:58.299 Robert Tseng: Okay, that sounds good.
200 00:21:03.510 ⇒ 00:21:04.270 Annie Yu: Okay.
201 00:21:04.800 ⇒ 00:21:13.110 Annie Yu: So the other tool that you were thinking is is a rudders. What’s that rudder stack.
202 00:21:13.500 ⇒ 00:21:15.529 Robert Tseng: Yeah. So
203 00:21:21.020 ⇒ 00:21:27.239 Robert Tseng: so I don’t think you need to evaluate. So they, I think they roughly do the same thing at segment.
204 00:21:29.170 ⇒ 00:21:33.299 Robert Tseng: so, yeah, I think as long as you understand how the segment one works, I think we can figure out the other.
205 00:21:33.900 ⇒ 00:21:35.210 Annie Yu: Okay. Okay.
206 00:21:35.210 ⇒ 00:21:36.030 Robert Tseng: Gaps.
207 00:21:46.410 ⇒ 00:21:48.410 Robert Tseng: Okay, yeah. So I know, like, kind of
208 00:21:49.140 ⇒ 00:22:00.869 Robert Tseng: you have the stock. It’s like, it’s good. I understand your approach to the custom model. But just to summarize, like, I’m trying to tie this project into something that we’re actively doing. Which is this full evaluation?
209 00:22:02.190 ⇒ 00:22:03.360 Robert Tseng: And so
210 00:22:03.750 ⇒ 00:22:09.960 Robert Tseng: yeah, there’s 2 pilots that we’re running for like 2 use cases that we’re testing for. I think.
211 00:22:11.340 ⇒ 00:22:26.339 Robert Tseng: what I had asked to do on the predicted Ltv side. Kind of ties in very well into one of those use cases. So just having you investigate that and then help me to. I mean, I think it’d be great if you could help me to find that experiment as well. I kind of gave you my idea, which is just like.
212 00:22:26.440 ⇒ 00:22:27.840 Robert Tseng: okay, we’re gonna
213 00:22:28.010 ⇒ 00:22:34.280 Robert Tseng: build a customer data model in the data warehouse. We’re gonna use the Cdp tools we’re gonna push traits into
214 00:22:34.420 ⇒ 00:22:36.239 Robert Tseng: like into the into the warehouse.
215 00:22:36.460 ⇒ 00:22:50.514 Robert Tseng: One is going to be built entirely with the tool itself. So maybe that predictive. Lpv, there’s another trade that’s valuable that only we can build because it’s using custom data that we have in the data warehouse. Then just Cdp tools was never seen. So
216 00:22:52.190 ⇒ 00:22:57.369 Annie Yu: Wait to clarify. So you said, if we do want to build.
217 00:22:58.300 ⇒ 00:23:03.240 Annie Yu: if we do want to consider the custom, the customized fields.
218 00:23:03.850 ⇒ 00:23:05.509 Annie Yu: it has to be our own.
219 00:23:05.670 ⇒ 00:23:07.300 Annie Yu: Is that it? Or.
220 00:23:08.070 ⇒ 00:23:20.356 Robert Tseng: Well, it’s it’s like these tools. They offer some sort of compute computation in platform. I don’t know exactly how it works, like I kinda showed you the predicts the Lpv one. That’s okay. There is something there.
221 00:23:21.730 ⇒ 00:23:27.220 Robert Tseng: I I think I I can see. I see the input it’s it’s just they’re just using events.
222 00:23:27.470 ⇒ 00:23:29.500 Robert Tseng: They’re just using order completed data.
223 00:23:29.640 ⇒ 00:23:40.359 Robert Tseng: But then there’s also other computations that we can make that are that only we can make because it uses data and the data warehouse. That segment will never sit that segment doesn’t.
224 00:23:40.520 ⇒ 00:23:50.900 Robert Tseng: doesn’t have all have access to all. The data only has access to fast data. We do have other stuff kind of flowing into our warehouse. And so
225 00:23:51.040 ⇒ 00:23:54.559 Robert Tseng: and just through the models that we’ve built that we’ve done.
226 00:23:55.350 ⇒ 00:23:58.950 Robert Tseng: I’ll yeah, like, there’s there. There could be other
227 00:24:00.630 ⇒ 00:24:04.930 Robert Tseng: other valuable traits that we can, that we can create
228 00:24:07.230 ⇒ 00:24:08.950 Annie Yu: Within their system, right.
229 00:24:09.660 ⇒ 00:24:13.819 Robert Tseng: No, I mean, I I want the profile to be
230 00:24:14.260 ⇒ 00:24:18.480 Robert Tseng: data model. The customer profile model should live in the warehouse.
231 00:24:20.420 ⇒ 00:24:25.409 Robert Tseng: I think the Cdp tool kind of just gives you the basics. But like,
232 00:24:27.920 ⇒ 00:24:33.210 Robert Tseng: yeah, which we are using the bait like we’re using like the bare bones we’re just using. Like
233 00:24:33.370 ⇒ 00:24:38.689 Robert Tseng: I mean, I don’t know. You can literally just go into bigquery and and look at what we have like. I.
234 00:24:39.840 ⇒ 00:24:40.910 Annie Yu: Yes.
235 00:25:01.080 ⇒ 00:25:02.859 Robert Tseng: I don’t know if that’s familiar.
236 00:25:03.610 ⇒ 00:25:06.299 Robert Tseng: And that’s fine.
237 00:25:16.170 ⇒ 00:25:24.329 Robert Tseng: Okay, so yeah, we have all this Pii data. We have utm stuff
238 00:25:25.020 ⇒ 00:25:53.949 Robert Tseng: from ids here. I mean, I think a lot of this is being pulled into from segment. But it would just have basic demographic data in here. This customer data model is not useful. I mean, it’s like, it’s like a directory. It’s like a phone book like it doesn’t have any. It kind of tells us basic things about where the customer came from. But there’s nothing about how much the customer has spent with us, what products they’re using, whatever like. And yeah, we have, we can do these joins and like, figure that out.
239 00:25:55.860 ⇒ 00:26:02.439 Robert Tseng: But you know, if I’m giving something to marketing, I want to give them a single model that they can use to like run.
240 00:26:02.820 ⇒ 00:26:17.640 Robert Tseng: you know, experiments off of. Maybe it won’t actually be this one, because we’ll we’ll create like a specific like marketing, you know, customer data model, and only show them stuff that’s relevant to them. But yeah, it’s like, what are all the things?
241 00:26:17.820 ⇒ 00:26:24.089 Robert Tseng: What are all the different cuts of by customer that a marketer would want to see in order to help
242 00:26:24.400 ⇒ 00:26:29.070 Robert Tseng: personalize both campaigns that we target our customers right? So
243 00:26:29.570 ⇒ 00:26:38.229 Robert Tseng: last order date that matters number of orders, matters lifetime value matters predicted output lifetime value matters as well.
244 00:26:39.940 ⇒ 00:26:46.620 Robert Tseng: yeah, it may be you’re running it. It could be like a million different things right? Like the the.
245 00:26:48.410 ⇒ 00:27:08.439 Robert Tseng: And if we could do even a sequence of of things, it’s like customers who purchase something 6 months ago, but then haven’t done anything the past 3 months because they cancel their plan. We want to do a very specific retargeting campaign on them, asking them to reactivate their membership by offering a 50% discount.
246 00:27:10.080 ⇒ 00:27:11.919 Robert Tseng: Yeah, like, that’s what this
247 00:27:12.100 ⇒ 00:27:31.460 Robert Tseng: data unlocked for the marketer. Because right now they have no idea of like how they could do any of that. So they’re just, you know, targeting customers off of their sign update. I think there is last order data, because we have a direct. I mean, I I think I see bask integration in customer I/O,
248 00:27:31.560 ⇒ 00:27:52.420 Robert Tseng: but like, that’s that’s it. It’s the email performance is really low. Of all the emails. And Eden is is hitting. They have a 7% open rate. And they’re like, barely getting anybody to like purchase off of it. So I just think that this is very low hanging fruit for us to like. Give the marketers some better data on
249 00:27:52.760 ⇒ 00:27:58.200 Robert Tseng: like who they should be going after in their campaigns. So
250 00:27:58.350 ⇒ 00:28:05.410 Robert Tseng: that’s that’s like the big picture of like what this, how this project ties into like.
251 00:28:05.660 ⇒ 00:28:08.179 Robert Tseng: I guess Eden’s marketing workflow.
252 00:28:09.606 ⇒ 00:28:13.650 Annie Yu: Got it. So the Ltv is, we just like part of it.
253 00:28:14.690 ⇒ 00:28:15.510 Robert Tseng: Yeah.
254 00:28:21.030 ⇒ 00:28:21.580 Annie Yu: Okay.
255 00:28:30.830 ⇒ 00:28:32.319 Robert Tseng: Cool, any other questions.
256 00:28:36.370 ⇒ 00:28:45.610 Annie Yu: so just 1 1 more time. So I’m gonna review that of TV prediction from segment today. And then.
257 00:28:45.610 ⇒ 00:28:46.180 Robert Tseng: Okay.
258 00:28:46.957 ⇒ 00:28:50.069 Annie Yu: I think I’ll leave the decision
259 00:28:50.570 ⇒ 00:28:53.130 Annie Yu: up to you. Is that okay? Like I will.
260 00:28:53.130 ⇒ 00:28:53.850 Robert Tseng: Yeah, of course.
261 00:28:53.850 ⇒ 00:29:00.730 Annie Yu: Document like, what’s there and how how they work. And then, yeah, we can talk about like.
262 00:29:00.860 ⇒ 00:29:03.230 Annie Yu: what’s the best next step.
263 00:29:03.900 ⇒ 00:29:04.450 Robert Tseng: Yeah.
264 00:29:05.150 ⇒ 00:29:09.310 Annie Yu: Okay, yeah, that sounds good.
265 00:29:09.780 ⇒ 00:29:14.130 Robert Tseng: Oh, okay, well, that’s it. They don’t have to take the full time.
266 00:29:14.920 ⇒ 00:29:21.990 Robert Tseng: yeah, I don’t know if I’ll join. Stand up later. But I’m gonna try to update some tickets. And then I might have you guys just wondering about them.
267 00:29:22.640 ⇒ 00:29:25.589 Annie Yu: Yeah. Yeah. Sounds. Good. Thanks.
268 00:29:25.590 ⇒ 00:29:27.000 Robert Tseng: Thanks, Danny. Bye.