Meeting Title: Uttam <> Robert Date: 2023-12-11 Meeting participants: Robert Tseng, Uttam Kumaran
WEBVTT
1 00:02:11.520 ⇒ 00:02:17.059 Uttam Kumaran: Hey, Tom, how’s it going? How’s the weekend?
2 00:02:17.150 ⇒ 00:02:37.620 Robert Tseng: That’s great and pretty cold here? But it’s it’s I mean cold for my standards. But I it’s it’s starting to look nice at night, and you know, like more of a Christmas fest festive vibe. So yeah, same here, it seems like more Christmas like I’m finally running the heater at my place. And I’m like.
3 00:02:37.860 ⇒ 00:02:38.710 Robert Tseng: nice
4 00:02:39.730 ⇒ 00:02:58.630 Uttam Kumaran: cool. So yeah, I guess the kind of what we talked about last week was I wanted to kind of spend a little bit of time for one of my clients, and one of the things they originally asked me to try and do was like a bit of a weather analysis on how
5 00:02:58.920 ⇒ 00:03:02.770 Uttam Kumaran: like, whether or whether events affect
6 00:03:03.220 ⇒ 00:03:08.660 Uttam Kumaran: like, have affected sales in the past, and whether they could leverage that for
7 00:03:10.260 ⇒ 00:03:27.950 Uttam Kumaran: whether they can leverage that for like this upcoming year in any way, and really find patterns. So I guess one was just like wanted to kind of discuss that with you, and then see. like kind of like, whether you have time this week to maybe look through something like that. And how we can kind of like, make that happen.
8 00:03:28.150 ⇒ 00:03:30.190 Yeah.
9 00:03:30.460 ⇒ 00:03:36.389 Robert Tseng: yeah, that sounds good. Let me just pull up some like a notepad here for me to keep track. Yeah.
10 00:03:37.670 ⇒ 00:03:43.109 Uttam Kumaran: Yeah. And then so kind of like, the general setup is like everything they have. All their data is modeled
11 00:03:43.200 ⇒ 00:03:46.560 Uttam Kumaran: and snowflake Snowflake also has.
12 00:03:46.670 ⇒ 00:03:52.299 Uttam Kumaran: like a data marketplace where you can actually get a lot of data sets on weather.
13 00:03:52.600 ⇒ 00:04:01.390 Uttam Kumaran: II don’t know from like your like flexible stuff. You guys did anything with weather or like, if you, if you have like, a pretty advanced understanding of like.
14 00:04:01.470 ⇒ 00:04:05.589 Uttam Kumaran: I don’t know if I’ve ever worked with weather data for like to play around with.
15 00:04:05.760 ⇒ 00:04:14.299 Robert Tseng: Yeah, I mean, we I mean, obviously, we looked at seasonal trends. I think seasonality is important, whether
16 00:04:15.010 ⇒ 00:04:24.050 Robert Tseng: we would never use weather as like a leading indicator of like a shift in demand. So I think that’s like an interesting analysis to do.
17 00:04:25.110 ⇒ 00:04:28.120 Robert Tseng: yeah, I guess we. I mean, if anything.
18 00:04:28.210 ⇒ 00:04:34.859 Robert Tseng: because we have, like some software that was like monitoring like delays, particularly rail. It’s like
19 00:04:34.920 ⇒ 00:04:44.610 Robert Tseng: what gets delayed the most when there are weather when there are weather issues. So we did have, like a whole initiative to
20 00:04:44.790 ⇒ 00:05:06.670 Robert Tseng: like a response team to like weather delays for for rail. Specifically. So. Yeah, I mean, that’s like deep into archives. That’s the kind of thing about like what was last like. Look back at some of those from back then, to see if there’s anything relevant to pull here, but that’s like the closest weather related thing that we did
21 00:05:06.930 ⇒ 00:05:10.170 Uttam Kumaran: at minimum zip in some cases.
22 00:05:10.320 ⇒ 00:05:17.760 Uttam Kumaran: like, of course, whole addresses, and I think even shopify gives like lat long and I assume
23 00:05:17.980 ⇒ 00:05:24.439 Robert Tseng: whether we probably have it like some sort of zip level. So so the main things that I was kind of just thinking about was
24 00:05:24.580 ⇒ 00:05:44.060 Uttam Kumaran: like I would I would I? It would be interesting to see like what people typically, they open their pools up. They take like the covers off, and they may be ordering filters and things like that. So that usually happens around like March, April in many places. Similarly, there’s a period in which we will close. So they order like cover pumps.
25 00:05:44.070 ⇒ 00:05:45.809 Robert Tseng: things like that.
26 00:05:45.940 ⇒ 00:05:49.360 Uttam Kumaran: And so one thing it’d be interesting to see is like, if places
27 00:05:49.470 ⇒ 00:05:55.299 Uttam Kumaran: like if if weather procedure gets warmer, and that has any sort of like relation to like
28 00:05:55.320 ⇒ 00:05:59.719 Uttam Kumaran: orders of a particular type of product, or in general happening in that zip
29 00:06:00.070 ⇒ 00:06:09.340 Uttam Kumaran: It’d be interesting one if there’s a sales in general, if that changes anything and like kind of like, you know, and then in particular, if there’s any products.
30 00:06:09.490 ⇒ 00:06:13.830 Uttam Kumaran: yeah, those are kind of like the main things that like.
31 00:06:14.670 ⇒ 00:06:16.330 Uttam Kumaran: I kind of thought about.
32 00:06:16.610 ⇒ 00:06:24.490 Robert Tseng: Yeah, sure, I mean, I think these are few few things we could run an easy regression on and just thinking like, we want to keep it really high level, just to see if anything.
33 00:06:24.960 ⇒ 00:06:27.580 Just like a quick way to get at this.
34 00:06:28.160 ⇒ 00:06:31.530 Robert Tseng: Yeah, I mean, like my
35 00:06:31.550 ⇒ 00:06:48.419 Robert Tseng: some. I’m sure there’s some sort of like aggregated like what monthly weather trend, and then we can kind of like, just look at correlations with sales for the different like products that you’re talking about one for like when when the season opens and when the season closes, I guess.
36 00:06:48.440 ⇒ 00:06:56.050 Robert Tseng: So. Yeah, I mean, I feel like I can. I can kind of think about like a data set that would be good to just run like an initial
37 00:06:56.230 ⇒ 00:07:01.429 Robert Tseng: like test on. And then if we see anything interesting, then we maybe like dig into it a bit more.
38 00:07:01.970 ⇒ 00:07:10.350 Uttam Kumaran: Yeah. So then, I guess, tell me about like, what’s the best setup. So I have everything in Snowflake. Would you be running a lot of that in excel or
39 00:07:10.390 ⇒ 00:07:22.869 Robert Tseng: yeah. Well, I guess what I was thinking is, yeah, I mean, I don’t think there’s I don’t. I don’t think there’s that much data we don’t really need to. I mean, we could probably do it in excel, or I’ll run it in python like python notebooks.
40 00:07:23.490 ⇒ 00:07:33.560 Robert Tseng: So yeah, I can like, tell you, like we, we could kind of think we could talk through like the different data sets that we would need to like, start to to do some like exploration on this.
41 00:07:34.080 ⇒ 00:07:40.560 Robert Tseng: some some data exploration here. And then we can kind of go from there.
42 00:07:41.060 ⇒ 00:07:45.339 Uttam Kumaran: Okay, yeah. So tell me. Tell me what you need.
43 00:07:46.030 ⇒ 00:07:47.950 Robert Tseng: Yeah. So
44 00:07:48.580 ⇒ 00:07:58.430 Robert Tseng: you mentioned so you have from their shot. You is a shopify stores where you say shopify and Amazon shopify Amazon. Okay.
45 00:07:59.890 ⇒ 00:08:15.589 Robert Tseng: yeah. So yeah, I mean, I guess it’d be good. I mean, you already mentioned how a lot of your orders kind of cluster in like a few different states. So I mean that I think that makes sense. Yeah, if we could get to like the Zip level of that data that’d be that’d be cool.
46 00:08:15.670 ⇒ 00:08:41.019 Robert Tseng: And then I think, as far as like the weather, the weather’s data set, I’m just thinking. I mean. I’d love to look at the the marketplace and stuff like to see what they have there. But if not, I’m sure, like Google, or like something has pretty, you know, some like Google trend level like data that you can easily export for like monthly averages over. Let’s just say like the past, like 3 years or something.
47 00:08:41.750 ⇒ 00:08:59.670 Robert Tseng: yeah. And then I could just take a look at like a year over year analysis over like the past 3 years, and then we can full your your over year. Comparison of like monthly, of of monthly sales and and weather pattern correlations over the past 3 years. I feel like could be a good place to start.
48 00:08:59.780 ⇒ 00:09:03.029 Uttam Kumaran: Okay, so let me let me just set up
49 00:09:03.190 ⇒ 00:09:11.190 Uttam Kumaran: within Snowflake like I’ll just have a query written that can give you all the sales data with all that dimensionality. And then
50 00:09:11.360 ⇒ 00:09:24.620 Uttam Kumaran: same thing I was just looking at. I’ll I’ll pick one of the weather things that just has like a basic one. And then we can only start there and then, if we’re like, Hey, be good to have specific weather events, there’s more dimensionality. We can go from there. I’ll just find something that’s free.
51 00:09:24.810 ⇒ 00:09:27.260 Uttam Kumaran: And then
52 00:09:27.810 ⇒ 00:09:39.869 Uttam Kumaran: tell me what’s the best way to pay and stuff. I guess, like how much you like, how much time you think for just to run like an initial test. And again, I’m hoping there’s some signal. I actually, I feel like
53 00:09:40.090 ⇒ 00:09:44.920 Uttam Kumaran: there actually might be. So it I’m kind of like looking forward to seeing like what
54 00:09:45.090 ⇒ 00:09:56.420 Uttam Kumaran: what happens. And then I’m also just looking forward to kind of seeing your process, and then seeing where else across the client, through, like, you know, in their other arenas, there’s actually opportunity to work together for sure.
55 00:09:56.550 ⇒ 00:10:05.420 Robert Tseng: I mean, I guess you, I mean, do you have a client have like hypotheses that we could kind of speak to directly? I think that would be a good way to kind of talk. Talk about signal.
56 00:10:05.920 ⇒ 00:10:11.950 Uttam Kumaran: yeah. I mean again, the I can. I’ll look back at some notes cause. We talked about it a while ago, but
57 00:10:12.420 ⇒ 00:10:15.010 Uttam Kumaran: the main thing they were saying is that
58 00:10:15.730 ⇒ 00:10:28.529 Uttam Kumaran: sometimes, if like one of the big things is, if there’s like a storm arriving, people need to cover their pools. Typically. And so in case of those types of like harsh weather conditions, there may be a spike in
59 00:10:28.540 ⇒ 00:10:33.390 Uttam Kumaran: like pool covers and things like that. Similarly, if unexpectedly.
60 00:10:33.430 ⇒ 00:10:49.880 Uttam Kumaran: there’s like the weather gets a lot warmer, people may open their pools, and they may order equipment. The way that actually gets actioned on is like, if we can kind of understand that those events are happening, or a few days before a storm. We can then run really targeted marketing to say like, Hey, there’s a storm coming in this area.
61 00:10:50.080 ⇒ 00:11:02.550 Uttam Kumaran: We can run marketing around, you know covers and things like that. Similarly, if the weather is gonna look good earlier than expected, there’s probably ways for us to run really hyper targeted marketing at like the zip or the state level.
62 00:11:02.690 ⇒ 00:11:10.460 Uttam Kumaran: So those would be. Those would be, I think, the kind of initial 2 ways where I would say, like looking back at the past.
63 00:11:10.610 ⇒ 00:11:15.170 Uttam Kumaran: Could there have been some Alpha there where, if we life.
64 00:11:15.340 ⇒ 00:11:26.549 Uttam Kumaran: if we started selling some marketing a really a little bit earlier. We could have, like at least a month or a couple of weeks out, understood that the weather is gonna change, cause we have targets and marketing. So those would be the 2 initial things that
65 00:11:26.560 ⇒ 00:11:28.120 Uttam Kumaran: could be interesting to try.
66 00:11:28.490 ⇒ 00:11:48.540 Robert Tseng: Yeah, yeah, totally. I mean, definitely, the retroactive analysis makes sense. I think. Yeah, that that would that would be. Yeah, I probably wouldn’t do anything. It’s not. It’s not truly predictive at this point. Yeah, which is fine. I think if we, as long as we could be able to say with, you know, some degree of confidence that, like knowing this pattern like this, would, we would have expected to spike or decrease in sales.
67 00:11:48.650 ⇒ 00:11:50.949 Robert Tseng: Yeah, I think that makes sense.
68 00:11:51.440 ⇒ 00:11:54.410 I think couple of things to note would be.
69 00:11:54.600 ⇒ 00:12:22.629 Robert Tseng: yeah, I mean, sometimes, like there will be. There’s a lot of conflating variables here, or confounding right? So I think about like poor weather, you know, typically later in the year, it also like coincides with like big discounts and stuff. So I think it’d be great to know, like their marketing calendar or any other like promotion sales promotions. That I could like kind of just K cut that out and try to isolate it as much as I can. I also have all of their digital marketing spend
70 00:12:23.200 ⇒ 00:12:32.049 Uttam Kumaran: by date and by platform, and that has campaigns, everything. So maybe there’s some evening out to do there, cause that, I would say is
71 00:12:32.710 ⇒ 00:12:41.510 Uttam Kumaran: like, really, it’s probably really tightly correlated to the to the revenue brought in, and I have all the granularity on the marketing side by day.
72 00:12:41.760 ⇒ 00:12:45.269 Uttam Kumaran: Just don’t just don’t have it specific by.
73 00:12:46.860 ⇒ 00:12:52.780 Uttam Kumaran: I don’t think I have specific by Geo, but at least you’ll see overall marketing spent on a daily level.
74 00:12:54.570 ⇒ 00:12:58.730 Robert Tseng: Yeah, I mean that I think that that should be that should be enough.
75 00:13:00.360 ⇒ 00:13:04.800 Robert Tseng: that. Okay.
76 00:13:05.510 ⇒ 00:13:12.699 Robert Tseng: yeah. And then. So I think those are some. I’m just trying to think through other like conflating variables that could think of other than
77 00:13:13.330 ⇒ 00:13:21.699 Robert Tseng: marketing promotions that are going on. Go ahead. Yeah. I just
78 00:13:22.060 ⇒ 00:13:25.529 Robert Tseng: also another thing, I think.
79 00:13:26.580 ⇒ 00:13:31.429 Uttam Kumaran: yeah, I’ve been. I don’t think they’ve done anything particularly new this past year.
80 00:13:31.800 ⇒ 00:13:37.230 Uttam Kumaran: no, I will. I also have a lot of the email data. It’s not like.
81 00:13:37.480 ⇒ 00:13:39.540 Uttam Kumaran: it’s not like super effective
82 00:13:39.820 ⇒ 00:13:42.860 Uttam Kumaran: and then I have like, discount data.
83 00:13:44.140 ⇒ 00:13:55.209 Robert Tseng: email data. It’s just like, what for? Cr, yeah, like, Crm, stuff.
84 00:13:59.110 ⇒ 00:14:00.820 Robert Tseng: Yeah.
85 00:14:03.180 ⇒ 00:14:10.770 Robert Tseng: well, yeah, I mean, I think that could be. I mean, that could be that tells something else. But just think, looking through like customer engagement with like your
86 00:14:11.880 ⇒ 00:14:22.700 Robert Tseng: I mean, I could I would see this as like a follow-on analysis, let’s say we assume, is some sort of correlation. And now you want to validate whether or not your targeted email or marketing campaign is, gonna have an impact.
87 00:14:22.840 ⇒ 00:14:34.690 Robert Tseng: Then you like we would kind of wanna know, like how effective your current marketing campaigns are like. II guess, particularly through email. However, you plan on pushing out that sort of proactive message.
88 00:14:35.090 ⇒ 00:14:40.710 Uttam Kumaran: And yeah, I think that that could that could be a good follow.
89 00:14:40.790 ⇒ 00:14:43.849 Robert Tseng: Yeah. In the in the case that it works, it would be great to say like
90 00:14:44.040 ⇒ 00:14:50.269 Uttam Kumaran: this is what our you know. Our current, like open rates, are given like copy, and everything’s the same. And
91 00:14:50.910 ⇒ 00:15:01.700 Uttam Kumaran: I have. You know we have all the the shop of on the Amazon side. I don’t have the ability to retarget because we don’t get the emails but on shopify side and totally.
92 00:15:02.310 ⇒ 00:15:07.309 Uttam Kumaran: we target. You know, we have a lot of that data at the email level. So
93 00:15:09.460 ⇒ 00:15:14.269 Robert Tseng: yeah, I mean, I think, well, if we have to complicate it too much, I think it’s a good start.
94 00:15:14.750 ⇒ 00:15:23.609 Robert Tseng: so maybe I’ll just confirm on my end. After this I’ll just like kind of as best as I can like. Write out the dimensions that I would need for different sets.
95 00:15:24.000 ⇒ 00:15:31.019 Robert Tseng: Yeah, if you know, if there’s anything that you feel like is valuable. More is better than less for me. So I can down
96 00:15:31.330 ⇒ 00:15:41.349 Robert Tseng: and yeah, maybe that would be that’d be good. Yeah, I think. Yeah, I’ll think a little bit after this call, I’m sure like in the next hour, so I’ll be like, Oh, there’s one more thing, so I’ll
97 00:15:41.350 ⇒ 00:16:04.469 Robert Tseng: I’ll message you when that comes up.
98 00:16:04.490 ⇒ 00:16:06.089 Robert Tseng: Okay, cool. Yeah.
99 00:16:07.840 ⇒ 00:16:09.660 Uttam Kumaran: okay, nice.
100 00:16:11.790 ⇒ 00:16:14.460 Uttam Kumaran: Cool. Yeah. Let let’s start with that. And then,
101 00:16:15.020 ⇒ 00:16:26.430 Uttam Kumaran: I do have a ton of interesting stuff I think, to talk to you about on the like supply chain side, I’m J. And then the other thing I’m really working on is like they’re a little bit sticklers on like
102 00:16:26.820 ⇒ 00:16:37.250 Uttam Kumaran: they they want like this, like daily, like vital signs dashboard that I’m trying working for them, which is pretty much like, how is the business doing every day? So I’m kind of like finishing like a first draft of that.
103 00:16:37.530 ⇒ 00:16:43.289 Uttam Kumaran: Maybe I can even send you and be like, what do you think? Like I’m not. I’m not the biggest dashboarder.
104 00:16:43.460 ⇒ 00:16:47.079 Uttam Kumaran: and I’m like, not creative at all.
105 00:16:47.090 ⇒ 00:17:13.209 Robert Tseng: so I’m kind of just put everything. Throw everything at the wall, but they’re even there. They’re not really like. They don’t have any really clear requirements. So it’s like a lose lose situation. So I think, cause you’ve seen a lot of that. Maybe you can give you some feedback on like, what’s helpful. Yeah, totally. I mean, reminds me of a client like that like he wanted like that kind of dashboard. Yeah, II just like put out a lot at first, and then shrimp it down from there.
106 00:17:13.210 ⇒ 00:17:23.180 Uttam Kumaran: Yeah, that’s exactly like what the process has been like. So okay, cool. Alright. So let me add you to slack. And then, yeah, just send me any notes you have, and then
107 00:17:23.660 ⇒ 00:17:30.219 Uttam Kumaran: we can go from there. I’ll I’ll send you. I’ll send you access to everything and kind of begin putting that stuff together.
108 00:17:30.300 ⇒ 00:17:37.179 Robert Tseng: Yeah, alright. So yeah, I’ll I’ll wait for them. Then I’ll I’ll I’ll I’ll I’ll send you my notes shortly.
109 00:17:37.270 ⇒ 00:17:41.320 Robert Tseng: Alright, man, I’ll talk to you on there.