Meeting Title: Eden Project Data Integration Sync Date: 2025-11-18 Meeting participants: Awaish Kumar, Henry Zhao


WEBVTT

1 00:01:46.280 00:01:47.789 Henry Zhao: Hi, Awash, how are you doing?

2 00:01:48.940 00:01:51.060 Awaish Kumar: Hello, I’m good, how about you?

3 00:01:51.460 00:01:54.449 Henry Zhao: Good, thank you. Thanks for helping us run that, interview.

4 00:01:55.880 00:01:56.840 Awaish Kumar: By no worries.

5 00:01:57.670 00:01:59.620 Henry Zhao: Alright, let’s see, where’s the stuff?

6 00:02:00.100 00:02:12.580 Henry Zhao: Hey, so I added you to this call because I need your help for some things with Eden, and feel free to send these to Demilade if you don’t want to do it, or if you think he’s better for it. I just don’t know who to assign it to, and I think Demolade’s out today, right?

7 00:02:15.740 00:02:32.490 Henry Zhao: So the first thing is, as you know, the product ROASDash right now is getting ad spends from different platforms, right? So we have the Klaviyo spend… not Klaviyo, MetaSpend, Catalyst Spend, blah blah blah, all that stuff, but a lot of these come from Northbeam, right? So, Zuran is about to cancel Northbeam.

8 00:02:33.130 00:02:44.459 Henry Zhao: And we need to know if we can get these ad spends outside of Northbeam. So, like, probably from the platforms as APIs, and how long this would take, so he knows when we are able to cancel Northbeam and start setting up Wicked Reports.

9 00:02:45.120 00:02:48.550 Awaish Kumar: Yeah, we have to do in parallel, we can’t cancel it.

10 00:02:50.720 00:02:52.360 Henry Zhao: Yeah, when you do it parallel.

11 00:02:52.360 00:03:04.850 Awaish Kumar: You need to know, like, until and unless, Wikidreport is set up, like, we can’t, like, turn it off, because the reason is, it will take me

12 00:03:04.960 00:03:21.110 Awaish Kumar: a few weeks to build these things from scratch on somewhere else. So, in those few weeks, 4 weeks, you can actually have Wicked Report already on, so there’s no point in basically creating all those connections separately.

13 00:03:21.110 00:03:31.320 Awaish Kumar: It’s not just creating connection. Creating connection will just take a few, maybe an hour, or two, or whatever, but then it needs to be modeled, it needs to come out to the…

14 00:03:31.810 00:03:47.309 Awaish Kumar: from the raw, it goes to modeling. Modeling, how it works, we need to verify the numbers, validate it, and all of these, blah, blah, blah. And we have to do it again when we have fixed reports. It’s just, like, we’ll just increase more…

15 00:03:47.600 00:03:48.450 Awaish Kumar: Product.

16 00:03:49.370 00:03:58.809 Henry Zhao: Okay, if they don’t want to add another month of Northbeam, because that’ll be, like, $5,000, could we turn it off and just get the data manually during the month while you’re working?

17 00:03:59.370 00:04:08.260 Awaish Kumar: But it will be the same for us, like, I… I’m… I… like, we don’t have any connectors, right? We are going to use polyatomic then. Polyatomic will also

18 00:04:08.590 00:04:14.960 Awaish Kumar: tag them, right? For each factor. But not for Facebook, for Google, for TikTok, for each of them.

19 00:04:15.070 00:04:29.660 Awaish Kumar: the polyatomic will charge them, and then the extra hours we are going to put into modeling it all, and then… then we know that it needs to be turned off. I’m not sure why we want to put so much effort in…

20 00:04:29.780 00:04:32.470 Awaish Kumar: Something which we know we are… we are just going to…

21 00:04:33.540 00:04:37.009 Awaish Kumar: Like, just going to get a bit, yeah.

22 00:04:37.520 00:04:42.590 Henry Zhao: So you’re saying we just implement this in Wicked Reports, if that’s possible, and then turn it off after it’s implemented in Wicked Reports?

23 00:04:42.590 00:04:49.470 Awaish Kumar: Yeah, yeah, that’s what I’m, like, saying. Like, at the… at the end, we wanna get this, all of this data from WickedReport, right?

24 00:04:49.890 00:04:50.530 Henry Zhao: Yeah.

25 00:04:50.780 00:04:52.089 Awaish Kumar: That’s what I’m saying.

26 00:04:52.090 00:04:52.630 Henry Zhao: Yes.

27 00:04:52.740 00:04:54.559 Awaish Kumar: And we asked Geron about that, yeah.

28 00:04:55.020 00:05:11.380 Awaish Kumar: So when we, know that in 4 weeks, we have a tool ready, from when we are going to get this data, so I’m not sure why we can put, like, 40 or 50, 60 hours of efforts in something which is going to get

29 00:05:11.790 00:05:13.540 Awaish Kumar: Turned off in 4 weeks.

30 00:05:15.090 00:05:17.829 Henry Zhao: Okay, so let me ask him, will we be ready at Sven from Wicked Reports?

31 00:05:17.970 00:05:21.189 Henry Zhao: Same way we do for North Beam.

32 00:05:22.040 00:05:23.859 Henry Zhao: from… from North Beam.

33 00:05:24.960 00:05:26.000 Henry Zhao: Indeed.

34 00:05:26.110 00:05:34.280 Henry Zhao: our dashboards… If so… Yeah, let’s… let me just start asking that. Okay.

35 00:05:34.920 00:05:37.330 Henry Zhao: So, hold off on this, let’s,

36 00:05:40.560 00:05:42.099 Henry Zhao: BigQuery data modeling.

37 00:05:43.650 00:05:46.889 Awaish Kumar: Yeah, I can work on this this week.

38 00:05:47.360 00:05:48.419 Awaish Kumar: Alright. Okay.

39 00:05:49.920 00:05:50.740 Awaish Kumar: Hmm…

40 00:05:51.520 00:06:01.180 Henry Zhao: The next thing is, if you can reach out to Stuart, he wants to split the spend by product for Samoralin a different way. So, right now, in Product ROAS NL2B…

41 00:06:01.180 00:06:04.509 Awaish Kumar: We have a product mapping sheet.

42 00:06:04.760 00:06:05.390 Awaish Kumar: Right.

43 00:06:05.390 00:06:07.439 Henry Zhao: He wants to do it a different way, so…

44 00:06:07.440 00:06:08.820 Awaish Kumar: That’s what I’m saying.

45 00:06:09.230 00:06:18.060 Awaish Kumar: We have a sheet, like, ask him for that mapping that he wants to set up, and we can put that… put that in this sheet.

46 00:06:18.260 00:06:22.569 Awaish Kumar: And then, we can adopt it. So that’s the process.

47 00:06:22.910 00:06:26.459 Henry Zhao: Can I have you talk to him about this? Then you can just deal with him directly?

48 00:06:28.080 00:06:33.220 Henry Zhao: that I don’t really know… Yeah, he just wants this spend broken out by ad group.

49 00:06:33.580 00:06:35.500 Henry Zhao: Not ad group… right now, it’s ad group.

50 00:06:35.600 00:06:41.589 Henry Zhao: But he wants to break it down a different way. Right now, I think it’s broken down in the DBT by, order volume, right?

51 00:06:42.260 00:06:43.380 Henry Zhao: proportionally.

52 00:06:43.380 00:06:50.010 Awaish Kumar: No, no, we have… like, right now, for Sermaline, like, for each product, we get the…

53 00:06:50.240 00:07:08.689 Awaish Kumar: ad spend, right? We have the… the… each of these have either campaign name or ad name, where it says the product name, basically. Okay. So you identify if it is thermaline ODT or Cermoline injection. So, based on that name, we classify into one of those.

54 00:07:10.190 00:07:13.070 Henry Zhao: Let me check with him, and then if it’s so good, I can just close this, right?

55 00:07:13.420 00:07:14.400 Awaish Kumar: Yeah.

56 00:07:14.400 00:07:15.689 Henry Zhao: Okay, got it. Thank you.

57 00:07:18.150 00:07:20.209 Henry Zhao: Was that it? Let me check something. Alright.

58 00:07:20.660 00:07:21.690 Henry Zhao: That might have been it.

59 00:07:25.790 00:07:28.439 Henry Zhao: I think that’s it. On this one, he asked us…

60 00:07:28.710 00:07:31.889 Henry Zhao: Can we check if this conversion is legitimate? Alright, let’s take a…

61 00:07:31.890 00:07:37.940 Awaish Kumar: I think that’s very easy, yeah, you can just… Login into the…

62 00:07:41.700 00:07:42.180 Henry Zhao: back here, right?

63 00:07:43.960 00:07:47.749 Awaish Kumar: Yeah, you can check it here, but you can also check in PASC.

64 00:07:48.970 00:07:50.200 Henry Zhao: Oh, in Basque? Okay.

65 00:07:50.410 00:07:54.269 Awaish Kumar: We can search for this transaction ID, and maybe it will pop up.

66 00:07:54.980 00:08:01.500 Awaish Kumar: the order. Otherwise, you can obviously look at fact transaction. If it shows up there, that means it’s valid.

67 00:08:02.340 00:08:03.220 Awaish Kumar: Sounds like…

68 00:08:03.490 00:08:04.669 Henry Zhao: It’s not showing up there.

69 00:08:05.140 00:08:06.239 Henry Zhao: So it’s not valid?

70 00:08:06.650 00:08:07.310 Henry Zhao: Oh.

71 00:08:07.600 00:08:08.610 Awaish Kumar: Oh, yeah.

72 00:08:16.430 00:08:17.719 Henry Zhao: Nope, not showing up.

73 00:08:17.970 00:08:24.469 Awaish Kumar: Yeah, that could be the reason. I already shared one of this scenario. Yeah, can you go back to Victory?

74 00:08:25.360 00:08:29.769 Awaish Kumar: And, say, fact transaction with all orders.

75 00:08:30.170 00:08:30.880 Henry Zhao: Yeah.

76 00:08:35.409 00:08:41.369 Awaish Kumar: Yeah, so this is… the status was maybe moved, canceled, or something. That’s why.

77 00:08:41.370 00:08:42.840 Henry Zhao: Okay, canceled. Alright.

78 00:08:43.429 00:08:43.999 Awaish Kumar: Okay.

79 00:08:45.170 00:08:46.720 Henry Zhao: Let’s say this was a canceled order, right?

80 00:08:46.990 00:08:47.660 Awaish Kumar: Yeah.

81 00:08:48.270 00:08:58.460 Henry Zhao: Yeah, but we do want these orders in the attribution table, because we want to later analyze, like, is there a higher percentage of canceled orders from a specific UTM? So that’s why I use the raw tables.

82 00:09:00.680 00:09:05.580 Awaish Kumar: Yeah, but we, like, we don’t need, raw tables even for that, for…

83 00:09:05.710 00:09:11.370 Awaish Kumar: affect transactions, you have the UTM, parameters.

84 00:09:11.590 00:09:12.879 Awaish Kumar: In the factology.

85 00:09:15.010 00:09:16.509 Henry Zhao: Yeah, yeah, but this is from Basque.

86 00:09:16.770 00:09:17.780 Henry Zhao: So…

87 00:09:18.340 00:09:18.860 Awaish Kumar: So what?

88 00:09:18.860 00:09:20.089 Henry Zhao: What I’m doing is the edge layer data.

89 00:09:20.090 00:09:32.119 Awaish Kumar: Yeah, that’s what I wanted to say. You can join your Agile data with these fact transaction tables. You don’t need to go into the order completed or order updated tables.

90 00:09:32.350 00:09:39.750 Awaish Kumar: Because, like this, like, cases like this, the order is canceled, like, in the fact transaction, there’s no entry for this data.

91 00:09:39.990 00:09:43.309 Awaish Kumar: And you don’t have to take care of these edge cases

92 00:09:44.230 00:09:47.770 Awaish Kumar: If you just use fake transaction. Otherwise, you have to…

93 00:09:47.770 00:09:49.820 Henry Zhao: I need those edge cases, though.

94 00:09:50.100 00:10:08.549 Awaish Kumar: Yeah, yeah, but it, like, depends what you need, but, like, we have fact knowledge with all orders, if you need all the orders. So we have worked on, like, there are duplicate things, in fact, order updated, in order completed also. There are duplicate entries of the same order, so… and it just,

95 00:10:08.750 00:10:12.280 Awaish Kumar: Like, you are now… you are going to access the full table.

96 00:10:12.450 00:10:18.669 Awaish Kumar: In the order updated, which is raw table, and that’s, like, costly to access in BigQuery as well. So, yeah.

97 00:10:18.670 00:10:26.529 Henry Zhao: But the downside of that is that if I use the dbt tables, if one of those breaks, or one of those is wrong, it’s gonna have bad downstream effects.

98 00:10:27.710 00:10:28.619 Awaish Kumar: Sorry?

99 00:10:30.280 00:10:36.380 Henry Zhao: If I use one of the, like, model tables, if something breaks or something ends up being wrong, then it’s gonna break these tables also.

100 00:10:36.480 00:10:39.569 Henry Zhao: I feel it’s just safer to use the raw tables.

101 00:10:39.570 00:10:46.020 Awaish Kumar: Yeah, like, in my experience, that’s not the best case scenario, but…

102 00:10:46.150 00:10:50.610 Awaish Kumar: If you say so, because I have been using that, like,

103 00:10:50.820 00:10:55.989 Awaish Kumar: like, there are a lot of things. If you read the code, like, if you read

104 00:10:56.100 00:11:01.149 Awaish Kumar: affect transaction, if you read Production Summary, there are a lot of things going

105 00:11:01.380 00:11:14.730 Awaish Kumar: into these SQL we wrote that I can’t even remember all these edge cases if I have to do it again. So, that’s what I’m saying, that if you’re going directly into RAW, you miss a few things.

106 00:11:19.450 00:11:25.050 Henry Zhao: I don’t think so, though, because what we’re doing is completely separate.

107 00:11:26.980 00:11:29.779 Awaish Kumar: Yeah, okay, yeah, that’s okay.

108 00:11:30.180 00:11:45.179 Henry Zhao: We’re literally just connecting Zoran’s Webflow data just based on the transaction ID. Like, that’s basically it. So, I don’t want to filter by anything or change anything. So I’m not… I don’t want any of the fact transactions logic. If I do want it, I do a join with this table to fact transactions, and that actually will be coming up next.

109 00:11:46.460 00:11:50.139 Henry Zhao: Because what I need to do is verify transactions UTMs to the edge layer UTMs.

110 00:11:51.630 00:11:55.969 Awaish Kumar: But that’s… that’s the point. You can correctly draw an edge clear data.

111 00:11:55.970 00:11:59.349 Henry Zhao: No, but then I can’t… then I’m comparing Apple… then I’m comparing…

112 00:11:59.550 00:12:07.540 Henry Zhao: like, something that’s already done to something that’s already done. Like, I need to compare something that’s newly created to something that’s already done to look at the differences.

113 00:12:10.550 00:12:14.380 Awaish Kumar: Like, the newly created data is edge layer data, right?

114 00:12:14.380 00:12:33.579 Henry Zhao: Exactly, right. So I’m comparing it just to… connecting it just to the BASC order completed. I’m looking at the whole history based on anonymous ID. Fact transaction is different. It uses BASC data, it has all these filters and all these manipulations. Later, we need to compare these two tables and say, out of the transaction IDs for November, now that we’ve implemented both, what are the differences?

115 00:12:34.650 00:12:45.380 Henry Zhao: And then either something might be wrong with fan transactions, or we might be getting additional data from EdgeLayer. That’s what Eden wants to know. It’s like, did we improve our UTM tracking by implementing EdgeLayer versus relying on the BASC data?

116 00:12:45.640 00:13:04.009 Awaish Kumar: Because right now, that… the bath data is not good enough for them. The UTMs don’t look right. They’re not able to… Yeah, that you can even get it, yeah. You… that’s not the… like, order completed and the fact000, basically the same data. It’s just more normalized, you can say, but it’s the same source, basically.

117 00:13:04.740 00:13:05.320 Henry Zhao: Yeah.

118 00:13:10.170 00:13:16.049 Henry Zhao: Okay, so… so that’s my explanation. Those are the only things I need help with. I will get back to you on the Stuart thing,

119 00:13:16.440 00:13:19.480 Henry Zhao: But, let me know if you have any… anything else you want to talk about.

120 00:13:20.820 00:13:23.759 Awaish Kumar: Yeah, no, no, I don’t have anything else right now.

121 00:13:23.760 00:13:25.470 Henry Zhao: I’ll be out Thursday and Friday, so…

122 00:13:25.940 00:13:26.600 Awaish Kumar: Okay.

123 00:13:26.790 00:13:28.369 Henry Zhao: Alright, we’ll be more things coming your way.

124 00:13:29.830 00:13:38.279 Awaish Kumar: I may be working on, yeah, I don’t know, these new requests coming from Rayon. He wants us to now, again, update the logic for Catalyst.

125 00:13:41.780 00:13:45.330 Awaish Kumar: Then this ticket, which you showed for…

126 00:13:45.870 00:13:51.389 Awaish Kumar: Gabby, so yeah, I will try to… Prioritize those this week.

127 00:13:51.390 00:13:51.950 Henry Zhao: Yup.

128 00:13:52.730 00:13:55.259 Henry Zhao: Okay, sounds good. Alright, thanks, Alish.