Meeting Title: PP2G | SyncUp Date: 2025-03-19 Meeting participants: Luke Daque, Amber Lin, Yoon, Payas Parab
WEBVTT
1 00:00:12.210 ⇒ 00:00:13.490 Amber Lin: Hi! Hiyaz
2 00:00:17.190 ⇒ 00:00:18.549 Payas Parab: Hey? How are you guys
3 00:00:19.490 ⇒ 00:00:25.759 Amber Lin: Good. We actually just came from another meeting. I saw you posted something in the
4 00:00:26.130 ⇒ 00:00:27.040 Payas Parab: Yeah.
5 00:00:27.040 ⇒ 00:00:29.409 Amber Lin: Miss, can you go over that a little bit? I think
6 00:00:29.410 ⇒ 00:00:31.159 Payas Parab: Yeah, no. Problem.
7 00:00:32.049 ⇒ 00:00:34.369 Payas Parab: Yeah, is, we is.
8 00:00:34.510 ⇒ 00:00:37.639 Luke Daque: Is that Kyle? Or is that you, bo
9 00:00:37.640 ⇒ 00:00:38.990 Yoon: Yeah, that’s just me. Both
10 00:00:38.990 ⇒ 00:00:41.279 Payas Parab: It just comes. It just says you. So it’s like
11 00:00:41.280 ⇒ 00:00:41.930 Yoon: Okay.
12 00:00:41.930 ⇒ 00:00:43.959 Payas Parab: Your your letter pops up as a C.
13 00:00:44.780 ⇒ 00:00:53.280 Yoon: Yeah, that’s that’s weird. I just had to log into another account in Zoom. That’s why
14 00:00:54.028 ⇒ 00:01:02.240 Payas Parab: I want. I don’t wanna derail amber anything you you had like, anything you wanted to discuss with anyone else. I don’t want to derail that before we can get to that. If if that’s helpful.
15 00:01:02.240 ⇒ 00:01:07.780 Amber Lin: This is the main focus, because I am getting my wisdom through withdrawn. So I have to run real quick. So
16 00:01:07.780 ⇒ 00:01:09.299 Payas Parab: Oh, no worries.
17 00:01:09.300 ⇒ 00:01:13.930 Amber Lin: Here this actually, I’ll pass. I’ll let you. I’ll let you run it. It’s just
18 00:01:13.930 ⇒ 00:01:24.820 Payas Parab: Sure. But did you have? Did you still have outstanding questions on anything? Let me because yesterday I didn’t have a chance to sync. I was pretty heads down. Did you still have any outstanding questions on anything
19 00:01:26.040 ⇒ 00:01:31.159 Yoon: Yeah, I mean, I’ve been working on this on the shopify data. I guess.
20 00:01:31.950 ⇒ 00:01:34.239 Yoon: Yeah, the shopify data that I was working.
21 00:01:34.760 ⇒ 00:01:39.749 Yoon: not sure how I can do it with the Asia connection data, because that’s
22 00:01:39.910 ⇒ 00:01:44.119 Yoon: the data is kinda out outdated. We only have it up until
23 00:01:44.120 ⇒ 00:01:45.500 Payas Parab: Yeah, yeah.
24 00:01:45.500 ⇒ 00:01:47.000 Yoon: Or I guess, yeah.
25 00:01:47.280 ⇒ 00:01:59.839 Yoon: I mean, there’s there’s sort of like we could theoretically do a linear forecast. But I’m down to see how the shopify and that that. And I assume you’re just like pulling from all orders that shopify Amazon, like all the digital platforms. Right?
26 00:02:00.170 ⇒ 00:02:02.850 Yoon: Yeah. From from the data model that
27 00:02:03.680 ⇒ 00:02:08.049 Payas Parab: Yep, excellent. Okay for the Asia connection. Then I think there’s
28 00:02:10.370 ⇒ 00:02:24.829 Payas Parab: like, maybe we should just focus on the the digital digital platforms. For now, like, let’s just start there because it looks like you’ve already made like pretty awesome progress there. Maybe we can like bring that over to the finish line ish, and then kind of go from there. Does that make sense, do you think, or do you think it’s
29 00:02:25.010 ⇒ 00:02:27.170 Yoon: Yeah, yeah, sure. I mean, that’s
30 00:02:28.030 ⇒ 00:02:32.820 Yoon: the the data on on the shopify side is is cleaner and
31 00:02:32.820 ⇒ 00:02:33.520 Payas Parab: Yeah.
32 00:02:33.520 ⇒ 00:02:34.859 Yoon: Your taxes too
33 00:02:38.230 ⇒ 00:02:39.790 Yoon: Can work on that 1st
34 00:02:40.200 ⇒ 00:02:45.299 Payas Parab: Sure. Let’s do that then, and then I can pull up my quick notes that I posted in here.
35 00:02:46.165 ⇒ 00:02:48.229 Payas Parab: Just trying to make sure we review
36 00:02:48.530 ⇒ 00:02:54.790 Payas Parab: essentially like just considerations for you. I I like, do. The profit model is sick. I have you used that before
37 00:02:55.420 ⇒ 00:02:59.280 Yoon: Profit model. Yeah, yeah, I’ve experienced that
38 00:02:59.280 ⇒ 00:03:11.289 Payas Parab: I’ve never used that. I used to just do like old school boomer, core raw dog in the dummy variables for the seasonality. So I didn’t even know that was an option to just have it baked in automatically. That’s awesome.
39 00:03:11.290 ⇒ 00:03:14.919 Yoon: Yeah, profit is really cool. It’s really easy to use as well. So
40 00:03:14.920 ⇒ 00:03:31.369 Payas Parab: That’s awesome. I think the the. So I basically just like there’s a couple of considerations that are just like worthwhile. And it’s just to give second eyes. If you disagree with any of it like, please let me know like I’m just the quantity so demand you’re saying is like demand is based on quantity. Right? And that’s what we’re looking at.
41 00:03:31.790 ⇒ 00:03:32.340 Yoon: Yeah.
42 00:03:32.340 ⇒ 00:03:33.969 Payas Parab: Broken out by product class
43 00:03:35.200 ⇒ 00:03:55.570 Payas Parab: The only thing that I think might be helpful is. And and this is just like more about like structuring. Because I’ve done these data projects where I’m just raw dogging it in a Jupyter notebook. And then I can’t like go back and like, pull data out or like redo stuff without basically rewriting the code. I don’t know. Have you pushed it to Github, or where are you? Where are you saving this and tracking this
44 00:03:56.290 ⇒ 00:04:00.330 Yoon: The my, my jupyter notebook
45 00:04:00.330 ⇒ 00:04:03.559 Payas Parab: Your Jupyter notebook. Yeah, are, are you like saving that anywhere? Are you like
46 00:04:03.560 ⇒ 00:04:06.059 Yoon: It’s it’s just locally I’m I’m just saving local
47 00:04:06.060 ⇒ 00:04:14.643 Payas Parab: No worries. It might be helpful just to post that into Github, so like other people, have access to it and kind of have it, if that’s helpful.
48 00:04:15.380 ⇒ 00:04:26.740 Payas Parab: I think, though, the one thing that I so the the 1st 1st note I have here is that we have the flexibility to turn the quantity into a revenue model. I think that is somewhat important. I I like, was trying to figure out
49 00:04:27.390 ⇒ 00:04:53.170 Payas Parab: like Dan may want sort of like. There’s a demand forecast. But can we convert that demand forecast into like some financial port forecast? And and that’s like not a trivial thing to do so. I think it’s just making sure that the quantity, whatever quantity we have going in. We also have affiliated price data saved somewhere, and we can pull that out, because ultimately, by product class my fears. There’s 5. I believe there’s like 5 product classes, right? 4 or 5
50 00:04:53.970 ⇒ 00:04:56.580 Yoon: No, I think there was more than 5, but
51 00:04:56.580 ⇒ 00:05:01.860 Payas Parab: Oh, there’s more than 5. So my, my big, my, my big thing, is we
52 00:05:02.210 ⇒ 00:05:28.490 Payas Parab: we Dan. Dan may get overwhelmed by like 10 to 15 charts, right of like individual forecasts. And I think the individual forecast will be helpful, right? I’m imagining, like a dropdown where you can like click a a product line and like, see the forecast right? What might be useful is like, we basically would need to make sure. We can also create like a revenue forecast from it. So wherever you’re saving code, I would also pull in addition to quantity like average cost, and
53 00:05:29.000 ⇒ 00:05:33.460 Payas Parab: I would just pull something so we can check the distribution of average cost to make sure it’s not like
54 00:05:34.080 ⇒ 00:05:54.050 Payas Parab: again, like if there’s a in pool pumps right? If there’s like a super low cost one and a super high cost one. We’ll know that average cost list. Basically, can we fold that up easily into like, consolidate the different product classes into a revenue forecast. I wanna make sure we just like upfront build that. So we don’t have to like, go read rejigger stuff and redo that? Does that make sense
55 00:05:54.050 ⇒ 00:05:59.910 Yoon: Do you mean you want to see, like a histogram of each of product classes of
56 00:05:59.910 ⇒ 00:06:06.750 Payas Parab: Yeah. Yeah. So so what I’m saying is so in the quantity, like when when you’re feeding in the model, right? You’re pulling the quantity right?
57 00:06:07.020 ⇒ 00:06:10.910 Yoon: Yeah, that that’s the the quantity is basically the target variable
58 00:06:10.910 ⇒ 00:06:18.510 Payas Parab: The quantity is a target variable. And I and I agree with that. My my thing is for that same exact quantity that we’re forecasting. Right?
59 00:06:19.250 ⇒ 00:06:35.440 Payas Parab: Let’s make sure we save down the average cost or like, and some distribution points of the average cost, so that you could easily take the forecast. Right? Like, I’m just basically thinking, like, I’m imagining your forecast array. You can multiply that by this average cost. And then you could add them up right
60 00:06:37.170 ⇒ 00:06:42.780 Payas Parab: like. What I’m saying is, how would you fold the quantity into a high, level metric? That’s meaningful.
61 00:06:43.650 ⇒ 00:06:51.060 Yoon: So so do you mean like adding, adding the average cost as another variable to fit in the model? You said
62 00:06:51.060 ⇒ 00:06:55.885 Payas Parab: Not not feed into the model, not feed into the model, to specifically to
63 00:06:57.797 ⇒ 00:07:04.750 Payas Parab: specifically to like just having it right next. So when the array spits out right, also generating a revenue forecast.
64 00:07:04.860 ⇒ 00:07:06.250 Payas Parab: byproduct class
65 00:07:10.820 ⇒ 00:07:11.790 Yoon: Also, so
66 00:07:11.790 ⇒ 00:07:17.599 Payas Parab: So you don’t. You don’t change the modeling. The modeling is still the same. I think you you target your target. Variable remains quantity
67 00:07:18.160 ⇒ 00:07:34.909 Payas Parab: But the later portion, I think, like in the as you’re pulling out that data right? You’re like, here’s the quantity forecast. Here’s what I plotted as the quantity for each product class. We should also have a way to pull out, based on average cost the revenue forecast for that product clause, class
68 00:07:36.440 ⇒ 00:07:40.770 Yoon: So so that I’m sorry that that will be another forecasting model that
69 00:07:40.770 ⇒ 00:07:44.299 Payas Parab: It wouldn’t. It wouldn’t be a forecasting model. It would just be like a multiplication, right like
70 00:07:45.230 ⇒ 00:07:46.240 Payas Parab: like, if you have, if you.
71 00:07:46.850 ⇒ 00:07:51.380 Payas Parab: That’s like quantity for the next 6 months. Right? Let’s say, okay, okay.
72 00:07:51.380 ⇒ 00:07:55.320 Payas Parab: If I multiply that by the average cost I get the forecasted revenue for the next 6 months
73 00:07:55.320 ⇒ 00:07:56.650 Yoon: Oh, yeah. Yeah. Yeah. Yeah.
74 00:07:56.650 ⇒ 00:08:18.540 Payas Parab: I just wanna make sure that like as part of the process like, what I don’t want is like your model spits out a bunch of stuff, and then we have to like, go back and get that array later and then multiply it by average. We just create more work down the road. I think one of the outputs should be revenue, because and the the rationale is that for Dan, we can sort of give him a much higher level view and then dive into the granular stuff. Does that make sense
75 00:08:18.540 ⇒ 00:08:25.720 Yoon: Yeah, yeah, yeah. Sure. Then, then, do you? Do we want just an average overall, or like by month?
76 00:08:26.370 ⇒ 00:08:30.479 Payas Parab: I do. I just do by product class. And I and I would do like trailing.
77 00:08:30.890 ⇒ 00:08:36.170 Payas Parab: We. We have to figure out that’s there’s not not an easy answer to that right? Because, like, if there’s a lot of variation.
78 00:08:36.289 ⇒ 00:08:38.241 Payas Parab: we have to check on that right
79 00:08:38.950 ⇒ 00:08:41.319 Yoon: Us by months, or by either
80 00:08:41.320 ⇒ 00:08:50.020 Payas Parab: Yeah, I I wouldn’t even be like, I wouldn’t feel like if we just did like a trailing 3 month average cost that. That’s the end of the world like that’s fine
81 00:08:50.020 ⇒ 00:08:55.680 Amber Lin: Hi, sorry! Can you actually just open the doc and make edits in there?
82 00:08:55.680 ⇒ 00:08:56.090 Payas Parab: Yeah, yeah.
83 00:08:56.090 ⇒ 00:09:12.570 Amber Lin: This is great. I just have trouble mapping where this belongs to, and honestly, the comments most of the comments in the bottom. I don’t know what it means, because I did not personally spit it out. Based on my understanding. I work with Chat Gpt, so
84 00:09:12.570 ⇒ 00:09:12.890 Payas Parab: Of course.
85 00:09:12.890 ⇒ 00:09:16.249 Amber Lin: It’s not right. It’s you’re going to be right. There’s actually
86 00:09:16.250 ⇒ 00:09:23.659 Payas Parab: Like really, really good ideas in there. I I’m not trying to like, I just want to like, make sure, like, it’s just helpful, like we think through these things ahead of time. So then.
87 00:09:23.660 ⇒ 00:09:24.240 Amber Lin: Per line
88 00:09:24.240 ⇒ 00:09:28.739 Payas Parab: Bo’s not building something, and then we’re like, Oh, wait, Bo, let’s rejigger this into something else.
89 00:09:28.740 ⇒ 00:09:41.049 Amber Lin: Totally so. Can you open the notion, Doc? And let’s just write. Let’s just write the plan, and then we can write the next steps in there as well, so that this is a document that we can refer back
90 00:09:41.050 ⇒ 00:09:41.370 Payas Parab: Actually
91 00:09:41.370 ⇒ 00:09:46.079 Amber Lin: And we can send to Dan as well. So this is something I want to send to Dan
92 00:09:46.660 ⇒ 00:09:49.469 Payas Parab: Got it. Has Utam said we should send this to Dan
93 00:09:49.470 ⇒ 00:09:50.290 Amber Lin: Yes.
94 00:09:50.290 ⇒ 00:09:52.839 Payas Parab: Sweet. Okay? Awesome. Just making sure.
95 00:09:58.590 ⇒ 00:10:03.540 Payas Parab: Alright. So, yeah. So so the short term forecast, the target variable
96 00:10:03.540 ⇒ 00:10:05.300 Amber Lin: Oh, I think we’re still on your slack
97 00:10:05.770 ⇒ 00:10:06.240 Payas Parab: Sorry
98 00:10:13.430 ⇒ 00:10:14.940 Payas Parab: in the notion. Now.
99 00:10:15.710 ⇒ 00:10:16.520 Payas Parab: Awesome
100 00:10:16.840 ⇒ 00:10:23.700 Payas Parab: also. Thank you. Amber. That’s like someone needs to tell me to do that. Otherwise we could just talk in circles without any concrete stuff I really appreciate.
101 00:10:29.979 ⇒ 00:10:34.129 Payas Parab: So Bo, we’ll just add. So one of the outputs of the short term forecasting right
102 00:10:40.890 ⇒ 00:10:42.470 Payas Parab: Revenue, forecast.
103 00:10:42.930 ⇒ 00:10:45.949 Yoon: Would you like that, like in a table chart
104 00:10:46.420 ⇒ 00:10:47.200 Payas Parab: Yeah.
105 00:10:47.200 ⇒ 00:10:47.890 Yoon: Okay.
106 00:10:48.070 ⇒ 00:10:53.405 Payas Parab: Like so. And I can even like just to give you like, basically, what I’m imagining is a stacked bar chart
107 00:10:53.950 ⇒ 00:10:55.700 Payas Parab: of the revenue forecast.
108 00:10:56.090 ⇒ 00:10:56.860 Yoon: Okay.
109 00:10:57.250 ⇒ 00:10:59.199 Payas Parab: With confidence. Interval bands.
110 00:11:03.370 ⇒ 00:11:08.549 Yoon: But like like the one we see on the, on, the, on, the graph above, like the the
111 00:11:08.550 ⇒ 00:11:13.950 Payas Parab: Yeah. Yeah. So like, imagine there were like, cause there’s there’s a bunch of product classes right?
112 00:11:13.950 ⇒ 00:11:14.600 Yoon: Yeah.
113 00:11:14.600 ⇒ 00:11:42.329 Payas Parab: The question is, is like, I. I agree the forecasting needs to be done at the product class level right? Like, if we put in demand for a bunch of stuff. It’s gonna be all over the place also, like an added note, too, is, there’s like certain things that sell before pool season, certain things that sell after pool season. Right? So at the product level, we’re going to do this forecast. But what we’re gonna then have is 15 charts, right, or however many like whatever. How many, however many product categories, there are right
114 00:11:42.330 ⇒ 00:11:46.429 Yoon: I I think we should only do it for the top 3 or 4
115 00:11:46.430 ⇒ 00:11:47.060 Payas Parab: Sure.
116 00:11:47.060 ⇒ 00:11:52.800 Yoon: The other ones doesn’t really have much data. I believe
117 00:11:54.060 ⇒ 00:12:00.150 Payas Parab: That’s fair. Then then we can only do the top 3 and 4, I think that’s where like my head was coming from of like there’s only a few, but like
118 00:12:00.150 ⇒ 00:12:00.570 Yoon: Yeah.
119 00:12:00.570 ⇒ 00:12:04.669 Payas Parab: It’s heat pumps, cleaners, brushes, right? Something like that.
120 00:12:04.670 ⇒ 00:12:05.530 Yoon: Yeah, yeah.
121 00:12:05.780 ⇒ 00:12:12.110 Payas Parab: So okay, basically like additional output. Right? So add
122 00:12:14.000 ⇒ 00:12:18.279 Payas Parab: develop a average cost. Trailing metric, right?
123 00:12:22.250 ⇒ 00:12:26.239 Payas Parab: Create a single consolidated financial forecast
124 00:12:30.950 ⇒ 00:12:32.740 Payas Parab: for each product class.
125 00:12:39.850 ⇒ 00:12:44.140 Payas Parab: multiply the revenue or the quantity forecast
126 00:12:47.290 ⇒ 00:12:48.920 Payas Parab: by the average cost.
127 00:12:52.110 ⇒ 00:13:00.010 Payas Parab: and then create a joint table of revenue forecast by product
128 00:13:01.880 ⇒ 00:13:05.090 Payas Parab: I will also just say, check the distributions
129 00:13:06.370 ⇒ 00:13:13.570 Payas Parab: and time series, or the overtime pattern of this
130 00:13:15.580 ⇒ 00:13:19.892 Payas Parab: make sure average cost is a decently reliable right? It doesn’t have to be perfect
131 00:13:26.070 ⇒ 00:13:34.950 Payas Parab: because demand. The reason I think about this right is like demand. There’s 2 ways to look at it. Right is like the actual quantity by product, but also like for a bigger picture like Dan Level, like
132 00:13:35.840 ⇒ 00:13:41.069 Payas Parab: like, what’s the overall revenue opportunity? Right? That’s also on his mind. That’s where I’m coming from. Does that make sense
133 00:13:42.130 ⇒ 00:13:43.050 Yoon: Yeah.
134 00:13:43.660 ⇒ 00:13:52.300 Yoon: What? What I don’t understand will be the the bar chart, the confidence interval bands. I’m not sure how, and
135 00:13:52.300 ⇒ 00:13:54.780 Payas Parab: Let me let me just Google you you still seeing my screen.
136 00:13:54.780 ⇒ 00:13:55.470 Yoon: Yeah.
137 00:14:09.870 ⇒ 00:14:13.659 Payas Parab: Something like this is what I was imagining. It wouldn’t be this 100% chart
138 00:14:16.900 ⇒ 00:14:18.140 Yoon: Oh!
139 00:14:18.140 ⇒ 00:14:25.340 Payas Parab: Right. So you add the confidence interval. There, there’s a way to do this in sk, learn. I can help you like figure that out. But it’s something like this.
140 00:14:26.750 ⇒ 00:14:29.579 Payas Parab: I can tag that in the notion. So we have it
141 00:14:32.250 ⇒ 00:14:33.730 Yoon: Confidence and approval.
142 00:14:34.250 ⇒ 00:14:37.616 Yoon: I mean the the profit forecast only has the
143 00:14:38.980 ⇒ 00:14:42.750 Yoon: the y upper hat and the y lower hat.
144 00:14:43.310 ⇒ 00:14:48.239 Yoon: I guess I’m not sure if that’s the same thing with the confidence interval.
145 00:14:49.560 ⇒ 00:14:50.040 Yoon: But
146 00:14:50.520 ⇒ 00:14:52.400 Payas Parab: Say that again? What what is it called?
147 00:14:52.720 ⇒ 00:14:53.319 Yoon: The
148 00:14:54.720 ⇒ 00:15:02.024 Yoon: on, on the on the profit profit forecast. There is a Y hat hopper which is the
149 00:15:03.010 ⇒ 00:15:04.780 Payas Parab: That that would be the confidence interval. Yeah.
150 00:15:04.780 ⇒ 00:15:05.240 Yoon: No, no.
151 00:15:05.240 ⇒ 00:15:19.171 Payas Parab: What I had is your predicted. It’s your predicted like your predicted target variable. And then the upper and lower bound tends to be like confidence interval like that. This this output like is like pretty fucking solid. For if the and this just comes out of the box
152 00:15:19.450 ⇒ 00:15:19.870 Yoon: Yeah.
153 00:15:20.290 ⇒ 00:15:31.939 Payas Parab: That’s awesome. That’s fucking awesome. Alright. So cause this looks awesome, right? This looks awesome at the product class level. But what I’m worried about is, we present 6 of these charts, and Dan goes, okay, cool. Right?
154 00:15:32.800 ⇒ 00:15:37.289 Payas Parab: I want to fold it up into something that we can then give him a bigger picture
155 00:15:37.840 ⇒ 00:15:38.970 Payas Parab: that make sense
156 00:15:39.190 ⇒ 00:15:40.450 Yoon: Yeah, there, you go.
157 00:15:40.450 ⇒ 00:15:45.759 Payas Parab: And I just wanna make sure we also track the confidence interval. That was my second. Let me
158 00:15:46.280 ⇒ 00:15:55.329 Payas Parab: like model model validation. We do need to do some sort of. I I know that the Meta one is pretty good. I did do some research that it looks pretty pretty fucking, solid
159 00:15:55.530 ⇒ 00:16:01.460 Yoon: Yeah, there is a way to do some model validation with profit. I’ll have to
160 00:16:01.920 ⇒ 00:16:03.709 Yoon: take a look at it a little bit more
161 00:16:03.710 ⇒ 00:16:04.410 Payas Parab: Yeah.
162 00:16:04.800 ⇒ 00:16:10.250 Yoon: Then for the stack bar chart. Do we want just one
163 00:16:10.540 ⇒ 00:16:14.670 Yoon: bar chart for each product class, or just an overall
164 00:16:15.750 ⇒ 00:16:16.560 Payas Parab: Say that again
165 00:16:16.750 ⇒ 00:16:21.049 Yoon: Do, do we do? We just want to stack barture for each of the product classes
166 00:16:21.050 ⇒ 00:16:22.869 Payas Parab: So the stacked one would have all of them
167 00:16:23.440 ⇒ 00:16:24.990 Yoon: All of them added together right
168 00:16:24.990 ⇒ 00:16:33.719 Payas Parab: Yeah, yeah, that’s why it would be like stack. It would be stacked on revenue. But if you stack quantity, that’s not meaningful. Because, like brushes cost way less than heat pumps right
169 00:16:33.720 ⇒ 00:16:34.810 Yoon: Okay, okay, got it.
170 00:16:34.810 ⇒ 00:16:37.770 Yoon: If you’re just like, here’s your overall demand.
171 00:16:38.530 ⇒ 00:16:48.510 Payas Parab: That would be one additional thing. And then the second thing in the optimization phase, right? I might also try adding unique customers as the target variable
172 00:16:55.050 ⇒ 00:16:58.780 Payas Parab: Do you know how you fine tune these hyper parameters generally
173 00:16:59.497 ⇒ 00:17:01.700 Yoon: With the profit model we can
174 00:17:01.980 ⇒ 00:17:04.119 Payas Parab: So seasonality.
175 00:17:05.550 ⇒ 00:17:13.100 Yoon: That’s we can. We can either do additive or multi, like multiplicative. I think
176 00:17:13.260 ⇒ 00:17:14.230 Payas Parab: Okay.
177 00:17:14.930 ⇒ 00:17:21.049 Yoon: That’s the the some of the parameters holidays. I don’t. I’m not sure I can. I can add the holidays as well
178 00:17:21.560 ⇒ 00:17:27.389 Yoon: we can. We can either do custom holidays, or just use us holidays
179 00:17:27.912 ⇒ 00:17:32.569 Payas Parab: Though in the, in the model itself, right now, by default, does it count us holidays, or something like that?
180 00:17:32.570 ⇒ 00:17:35.489 Yoon: No, no, not by default. I’ll have to add it
181 00:17:35.490 ⇒ 00:17:38.160 Payas Parab: And the seasonality factor isn’t built in either
182 00:17:38.490 ⇒ 00:17:40.659 Yoon: Seasonality is built in like
183 00:17:40.660 ⇒ 00:17:41.430 Payas Parab: Great. Okay.
184 00:17:41.430 ⇒ 00:17:42.110 Yoon: I agree.
185 00:17:42.110 ⇒ 00:17:48.389 Payas Parab: So the hyper parameter tuning is the how aggressive you handle seasonality
186 00:17:48.710 ⇒ 00:17:50.150 Yoon: Yeah, yeah, yeah, that would be
187 00:17:50.150 ⇒ 00:18:03.149 Payas Parab: Cool. I I mean, honestly, like seasonality. I wouldn’t spend too much time like, if if there’s like natural, if the natural like time series seasonality is already captured. It’s probably fine, because it’s like a unique pattern.
188 00:18:04.130 ⇒ 00:18:06.739 Payas Parab: The holidays I wouldn’t.
189 00:18:07.310 ⇒ 00:18:17.519 Payas Parab: I wouldn’t stress too hard on. I think they do do some holiday promotions and sales. I think I think if we’re optimizing, I would I would actually do testing a different
190 00:18:18.230 ⇒ 00:18:18.760 Yoon: Yeah.
191 00:18:18.760 ⇒ 00:18:20.699 Payas Parab: I would do testing, yeah.
192 00:18:20.820 ⇒ 00:18:23.389 Yoon: There are ways to kind of improve it.
193 00:18:23.610 ⇒ 00:18:30.060 Yoon: We can. We can basically add additional regressors which charges additional
194 00:18:30.170 ⇒ 00:18:36.839 Yoon: like variables that we can add, like, for example, weather could be another variable that we can add to the model
195 00:18:36.840 ⇒ 00:18:42.740 Payas Parab: Yup, I would I would. I’ve done some analysis in that weather space before. I wouldn’t unlock that. I wouldn’t unbox
196 00:18:42.740 ⇒ 00:18:44.479 Yoon: That box, he added.
197 00:18:44.480 ⇒ 00:18:48.459 Payas Parab: It’s it’s a monster, Ryan. I don’t know if you remember. Like Dude, it’s it’s a
198 00:18:49.625 ⇒ 00:19:00.359 Payas Parab: it’s a it’s a beast that we may not want to unlock yet. So maybe I think I think what might be nice is trying the unique customers, because I think that could be. I worry. There’s not enough
199 00:19:01.880 ⇒ 00:19:07.150 Payas Parab: customers in general for it to be a meaningful forecast. Right? If you’re like average error band is like
200 00:19:07.330 ⇒ 00:19:24.990 Payas Parab: point 5 to you know what I mean like. If the if the number is generally smaller, the scale is smaller, the variance is like not meaningful at all. Right. If you can forecast one extra customer. That doesn’t mean a ton, but it’s I think it’s worth a shot. I I was looking at some so I would do that. I wouldn’t spend too much time on this
201 00:19:25.490 ⇒ 00:19:26.300 Yoon: All right.
202 00:19:26.794 ⇒ 00:19:30.750 Payas Parab: The weather economic indicators. I think, like this.
203 00:19:31.360 ⇒ 00:19:34.213 Payas Parab: I think we hold off on that to start.
204 00:19:34.530 ⇒ 00:19:36.700 Payas Parab: Yeah, we can. We can do the next time
205 00:19:40.980 ⇒ 00:19:47.719 Payas Parab: I think what I’d be more interested in in like in order to flip something to Dan is like, can we consolidate the data? I think that’s a more
206 00:19:47.860 ⇒ 00:19:52.799 Payas Parab: useful short term thing that we can crank out do you agree or
207 00:19:53.490 ⇒ 00:19:55.721 Payas Parab: i’m just i’m just spitballing, you so
208 00:19:56.910 ⇒ 00:20:14.550 Payas Parab: so I think that would be like if I were to like. Look at part one of this right, the short term plan. It’s like, let’s build that target. Let’s consolidate into a revenue forecast and make sure, based on confidence intervals. It’s somewhat meaningful. Then we can also try and optimize, maybe the unique customers. That was my next.
209 00:20:15.180 ⇒ 00:20:18.410 Payas Parab: I think. And we’re
210 00:20:18.450 ⇒ 00:20:26.130 Yoon: When you say, target variable, unique customer, that that would just basically be another training, another model
211 00:20:26.130 ⇒ 00:20:27.279 Payas Parab: Yes, another model. Yep.
212 00:20:27.280 ⇒ 00:20:28.580 Yoon: Okay. Okay.
213 00:20:29.230 ⇒ 00:20:31.340 Payas Parab: Yep, let me be very specific.
214 00:20:37.740 ⇒ 00:20:39.480 Payas Parab: Unique customers.
215 00:20:41.397 ⇒ 00:20:46.829 Payas Parab: You kind of get a sense of how to do that right, I think in the order table. There’s a customer email, right? So it’d be something like distinct
216 00:20:47.490 ⇒ 00:20:47.860 Yoon: Yeah.
217 00:20:47.860 ⇒ 00:20:50.628 Payas Parab: Email, something like that. Yeah, sweet.
218 00:20:52.260 ⇒ 00:20:59.160 Payas Parab: alright, yeah, I wouldn’t spend adding additional variables and stuff. I wouldn’t spend too much time on this at the moment. It’s just my my view
219 00:20:59.400 ⇒ 00:21:00.280 Yoon: Yeah, sure.
220 00:21:01.703 ⇒ 00:21:12.410 Payas Parab: Okay? So then, I think Amber raised a good point about evaluation. But I do want to tackle, since we only have a couple of minutes here. I want to tackle the kind of deployment some of this like like longer plot mopping stuff.
221 00:21:12.580 ⇒ 00:21:15.930 Payas Parab: I was looking at the granularity stuff. I think
222 00:21:16.070 ⇒ 00:21:35.790 Payas Parab: I checked, and you can check my queries like they’re linked in my message. I think the the cardinality would be too low like like they’re just too many skew. Sorry. The cardinality of the product skew is way too high to meaningful put to meaningfully put together a forecast like there’s just too many sorry like there’s just too many
223 00:21:36.430 ⇒ 00:21:40.659 Payas Parab: at the skew level. So I think product class might just make more sense
224 00:21:40.660 ⇒ 00:21:41.370 Yoon: Yeah.
225 00:21:41.710 ⇒ 00:21:50.969 Payas Parab: I I you! You can give it a rip, but what I think might be better potential. One is breaking it out by
226 00:21:51.617 ⇒ 00:21:55.080 Payas Parab: break it out by state and region.
227 00:21:55.860 ⇒ 00:21:59.759 Payas Parab: I think this helps their marketing team as well as their shipping team.
228 00:22:00.170 ⇒ 00:22:03.090 Yoon: So I think like this might be like
229 00:22:04.180 ⇒ 00:22:12.950 Payas Parab: I’m gonna put this as like p. 1. But P. 0 might be top 5 States
230 00:22:14.380 ⇒ 00:22:25.050 Payas Parab: and the remaining into regions. And there’s a query in there. I’m gonna comment it. So I don’t. I don’t wanna put all the Mumbo Jumbo if we’re sending this to Dan. So I’m just gonna comment the
231 00:22:25.730 ⇒ 00:22:29.259 Payas Parab: the snowflake query, here is that good amber? Is that okay?
232 00:22:30.250 ⇒ 00:22:32.230 Payas Parab: Sorry if you’re on your way to getting your wisdom teeth
233 00:22:32.230 ⇒ 00:22:48.269 Amber Lin: No, just just edit the doc as you see fit. Because you guys are the technical people. And you know how it’s gonna be done. Mostly, my role is just to make sure. Okay, the client sees this. And this is actually useful, because there’s a lot of tech things that we can do. That’s really fancy really cool. But the client
234 00:22:50.190 ⇒ 00:22:51.110 Amber Lin: not care
235 00:22:51.110 ⇒ 00:22:55.190 Payas Parab: So if I just add them as comments like the technical stuff, that’s okay, right? If we so that it doesn’t
236 00:22:55.190 ⇒ 00:22:56.730 Payas Parab: totally wouldn’t matter.
237 00:22:56.730 ⇒ 00:22:57.869 Payas Parab: Also just edit it.
238 00:22:57.870 ⇒ 00:22:58.500 Payas Parab: Client. Yeah.
239 00:22:58.500 ⇒ 00:23:06.210 Amber Lin: No, no, you can just also edit it. This is just a premium document. This is a base for you guys. Discuss. Just edit it in the Doc. It’ll be a lot helpful
240 00:23:06.210 ⇒ 00:23:08.839 Payas Parab: Thank you for organizing this amber. This is awesome.
241 00:23:09.870 ⇒ 00:23:10.420 Amber Lin: You know so
242 00:23:10.420 ⇒ 00:23:15.719 Payas Parab: This can be so. What I’m also thinking about is like the warehouse use cases. Bo
243 00:23:16.790 ⇒ 00:23:20.540 Payas Parab: cause, he was saying. That was something that would help the forecast. Right is like how much
244 00:23:20.740 ⇒ 00:23:29.029 Payas Parab: they keep an inventory and like where they ship from, and like identifying better shipping for what region I think this could be really helpful for them.
245 00:23:29.750 ⇒ 00:23:30.489 Yoon: Yeah, so
246 00:23:30.490 ⇒ 00:23:34.790 Payas Parab: I’d I’d prioritize that if based on our previous call with Dan, that’s the sense you got
247 00:23:35.840 ⇒ 00:23:39.910 Yoon: Yeah, but I but I think we’ll have to somewhat get
248 00:23:40.180 ⇒ 00:23:43.749 Yoon: like a feedback from Dan first, st if if you want
249 00:23:44.410 ⇒ 00:23:50.019 Yoon: like. If he wants, if he, if he keep, wants us to to do the forecasting demand
250 00:23:50.200 ⇒ 00:23:53.009 Yoon: on on the Asia connection data as well
251 00:23:53.380 ⇒ 00:23:57.700 Payas Parab: Yeah, agreed.
252 00:23:58.850 ⇒ 00:24:01.239 Payas Parab: Yeah. The Asia connection. I think
253 00:24:02.570 ⇒ 00:24:08.590 Payas Parab: my my view is that we table that for a little bit. But you’re right. We should ask Dan. So like we should ask Dan.
254 00:24:09.380 ⇒ 00:24:14.170 Payas Parab: But this is how I would approach the for the increasing this specific thing. Yeah.
255 00:24:14.520 ⇒ 00:24:19.880 Amber Lin: Can you just write a quick action item at the top of the talk to ask Dan about the Asia connection thing
256 00:24:20.130 ⇒ 00:24:20.810 Payas Parab: Yeah.
257 00:24:21.590 ⇒ 00:24:22.480 Amber Lin: Appreciate it.
258 00:24:23.290 ⇒ 00:24:31.339 Payas Parab: How about improving digital channel model versus Asia connection forecast model?
259 00:24:34.190 ⇒ 00:24:38.790 Payas Parab: Thank you. Okay, awesome. So
260 00:24:39.240 ⇒ 00:24:43.289 Payas Parab: that was the next thing was just like in that one.
261 00:24:44.308 ⇒ 00:24:46.452 Payas Parab: The other thing is this deployment. I think
262 00:24:48.090 ⇒ 00:25:03.300 Payas Parab: I like the idea of using real for visualization. I’m wondering if, for the purposes of forecasting, since the code is in Python and and Ryan, this may be a question for you. Is streamlet a better way, like a streamlet app that we deploy in Snowflake. So then, both
263 00:25:05.080 ⇒ 00:25:12.339 Payas Parab: python code is directly connected. Or is there a better way using one of these things that I I just don’t know how they work is a better way.
264 00:25:12.540 ⇒ 00:25:21.609 Luke Daque: Yeah, I’m not very familiar as well. So, but yeah, definitely, a streamlit app would be great, especially for, like python related visualization. So maybe
265 00:25:22.640 ⇒ 00:25:23.769 Luke Daque: try that out
266 00:25:23.770 ⇒ 00:25:34.169 Payas Parab: I’m trying to prevent the having the having you. You and de have to like build a pipeline from snowflake to python back into a visualization environment. I feel like that would just.
267 00:25:34.630 ⇒ 00:25:42.300 Payas Parab: I think, if we can deploy straight to streamlet, and I, bo, we can. If you’ve never have you worked with streamlet before? Beau
268 00:25:42.300 ⇒ 00:25:43.000 Yoon: Yeah.
269 00:25:43.260 ⇒ 00:25:46.309 Payas Parab: Awesome. Yeah, I think that that would make sense. Right? So
270 00:25:47.630 ⇒ 00:25:49.159 Luke Daque: Yeah, that makes sense
271 00:25:49.160 ⇒ 00:25:52.569 Amber Lin: Awesome. Just write it down. I don’t even know what the flask and the Api
272 00:25:52.995 ⇒ 00:25:59.070 Payas Parab: It’s okay. It’s okay. It’s okay. No, no forecasting. Move
273 00:25:59.360 ⇒ 00:26:06.460 Payas Parab: lows code with direct connection to streamlet into
274 00:26:09.560 ⇒ 00:26:18.249 Payas Parab: Oh, yeah. With direction connection to snowflake dB into streamlit build visualization
275 00:26:19.120 ⇒ 00:26:25.820 Payas Parab: with product level dash, state level, drop down.
276 00:26:28.210 ⇒ 00:26:32.669 Payas Parab: So I’m I’m gonna just be a little bit more here. So it’ll be a 2 page stream lit. App
277 00:26:35.650 ⇒ 00:26:40.468 Payas Parab: build visualization. So we’ll do revenue forecasting by product as one tab
278 00:26:53.010 ⇒ 00:27:11.939 Amber Lin: I just want to chime in here. I know we talked about. I know past you mentioned also to have the financial forecasting. What do you estimate is the effort or task we need to do, because I imagine we already have the data forecasted for the revenue converting it to a financial metric doesn’t really involve any forecasting. It’s mostly just calculations. Right?
279 00:27:11.940 ⇒ 00:27:13.610 Amber Lin: It’s a calculation. Yeah, okay, sounds good.
280 00:27:13.610 ⇒ 00:27:13.960 Amber Lin: Good.
281 00:27:13.960 ⇒ 00:27:19.870 Payas Parab: The the rash. The reason for bringing it up now, though, is, we need to make sure that whatever our forecast is, we align it, and
282 00:27:19.870 ⇒ 00:27:20.290 Amber Lin: There’s
283 00:27:20.290 ⇒ 00:27:25.180 Payas Parab: Sort of like like an upstream, like if we do like a rolling 3 month average of average.
284 00:27:25.870 ⇒ 00:27:30.670 Payas Parab: If we have a forecast 2 months from now. We wanna make sure there’s like a dynamic way to pull that. Not like a
285 00:27:30.670 ⇒ 00:27:31.150 Amber Lin: Hmm.
286 00:27:31.150 ⇒ 00:27:33.350 Payas Parab: That. That’s that’s sort of like where my head was at. But you’re right.
287 00:27:33.350 ⇒ 00:27:34.600 Amber Lin: Yeah. Totally. Why not?
288 00:27:34.600 ⇒ 00:27:35.419 Payas Parab: It’s not that
289 00:27:35.420 ⇒ 00:27:50.000 Amber Lin: Put that down also, because I think that’s what the client cares about the most. It’s just because they’re going through M. And A, and there’s a lot of valuation stuff. And they want to show, okay, based on our revenue. This is what we’re going to project our valuation to be
290 00:27:50.000 ⇒ 00:27:50.770 Payas Parab: Yep.
291 00:27:51.950 ⇒ 00:27:52.550 Payas Parab: Yeah.
292 00:27:53.500 ⇒ 00:27:55.070 Amber Lin: Value, yeah.
293 00:27:55.372 ⇒ 00:27:59.610 Payas Parab: Well, so that that’s this. This one right page one is the revenue forecast
294 00:28:00.458 ⇒ 00:28:05.389 Amber Lin: Plus financial metrics. I have got to go. I will stay on. I will listen in
295 00:28:05.520 ⇒ 00:28:11.110 Payas Parab: No worries, no worries at all. We can end this quickly, I think that’s everything.
296 00:28:12.380 ⇒ 00:28:17.180 Payas Parab: Yeah. So I would look at the state level. And then, yeah, for the integration and deployment. I think
297 00:28:18.660 ⇒ 00:28:39.790 Payas Parab: streamlit app makes the most sense. We can debate this cost thing right? So Bo, if you can. Can you make, like your code as functional as possible for the forecast, like, if there’s like a define like, define a function to build the financial forecast as opposed to just doing the multiplication. If that makes sense. So you could theoretically feed in a cost assumption
298 00:28:40.790 ⇒ 00:28:45.219 Yoon: A cost assumption for oh, I’m sorry! What can you
299 00:28:45.220 ⇒ 00:28:49.609 Payas Parab: Like, you know, I’m like the forecast here. I I can. I can help do this. By the way, it’s
300 00:28:49.850 ⇒ 00:28:59.289 Payas Parab: we’re saying, like, we want to be able to make sure we have, like some sort of like, in a cost assumption, right? To be able to forecast the financials directly.
301 00:29:01.210 ⇒ 00:29:06.289 Payas Parab: we may want to make it so he can plug it in and see what happens. Does that make sense
302 00:29:07.240 ⇒ 00:29:12.029 Yoon: Oh, so, so, so, so add like a add like an input box. So here
303 00:29:12.030 ⇒ 00:29:40.919 Payas Parab: And input box exactly like St. Dot, you know, whatever text. Input and then take that text input and then multiply it by a demand forecast to do like move the revenue forecast dynamically. That’s that’s my view. That might just be like a fun thing for him, like you know what I mean, like it also helps show off like, look we built this tool to like, do forecasting, and we can do forecasting and projections for everybody that could make him look really sick to the private equity firm
304 00:29:42.130 ⇒ 00:29:42.940 Yoon: Yeah.
305 00:29:43.370 ⇒ 00:29:53.509 Payas Parab: So let’s let’s let’s if you want to. What would be really helpful is if you save the code and like Github and push it there. So I can also like review, and I think I can help like modularize some of this. So
306 00:29:53.510 ⇒ 00:29:53.850 Yoon: Okay.
307 00:29:54.000 ⇒ 00:30:04.720 Payas Parab: Like, build that streamlit app and also be building these forecasts in the in the background. But yeah, that’s all my feedback. I don’t want to take up everyone’s more time. But does this all make sense? Is this helpful
308 00:30:04.910 ⇒ 00:30:06.849 Yoon: Yeah. Yeah. Totally.
309 00:30:06.910 ⇒ 00:30:17.419 Amber Lin: Yeah, this is super helpful. And I think we really know a lot more of what we’re gonna do next. Can you just help define the immediate next steps
310 00:30:17.770 ⇒ 00:30:22.839 Payas Parab: Yes, absolutely so immediate. Next steps, I think, is
311 00:30:23.640 ⇒ 00:30:29.459 Payas Parab: so. You ran the forecast already. Right, Bo, for each each thing, each part of us
312 00:30:29.460 ⇒ 00:30:30.480 Yoon: Protocols, yeah
313 00:30:30.750 ⇒ 00:30:40.489 Payas Parab: Okay, so why don’t you just share out those results somewhere? And then let’s also try and build out the immediate next step would be figuring out their average cost. For the revenue.
314 00:30:40.690 ⇒ 00:30:51.679 Payas Parab: That, I think, is the the 1st immediate step is like, what’s the best way to determine average cost. Like, basically, how many months back should we look? And is the average a reliable thing? Does that make sense
315 00:30:51.930 ⇒ 00:30:52.700 Yoon: Yeah.
316 00:30:52.980 ⇒ 00:31:00.669 Payas Parab: These average like those 2 questions, if you can answer, that would be like, the immediate next step is like, if we’re going to take these quantity forecasts you already have in the state that they’re in.
317 00:31:00.920 ⇒ 00:31:10.959 Payas Parab: Let’s try. And the the next step, like ideally, would just be like, what is the average cost. And can we get an average cost that’s reliable for multiplying for our forecast to get a financial forecast
318 00:31:11.940 ⇒ 00:31:13.560 Yoon: Yeah, yeah, sure.
319 00:31:13.560 ⇒ 00:31:15.130 Payas Parab: Let me write that down.
320 00:31:23.700 ⇒ 00:31:25.699 Payas Parab: okay, we’re just gonna review those.
321 00:31:27.720 ⇒ 00:31:29.040 Payas Parab: Build out the
322 00:31:36.610 ⇒ 00:31:39.369 Payas Parab: view. And Bo, are you able to do that? Or do you need?
323 00:31:39.810 ⇒ 00:31:43.359 Payas Parab: I am a little swamped with other work, but
324 00:31:44.070 ⇒ 00:31:46.730 Yoon: Yeah, yeah, I can. I can. I can do that today.
325 00:31:46.730 ⇒ 00:31:52.109 Payas Parab: Awesome. Great, all right. So I think those are immediate. Next steps, initial forecast review with Adam.
326 00:31:52.980 ⇒ 00:31:54.150 Payas Parab: The Xg boost.
327 00:31:54.540 ⇒ 00:32:01.740 Payas Parab: I I like, I’m okay not to like. Look at actually boost. That’s just my view. But api versus dashboard
328 00:32:01.740 ⇒ 00:32:06.820 Yoon: We also need some kind of metric to evaluate the model right
329 00:32:09.100 ⇒ 00:32:13.360 Payas Parab: Yeah, I’m trying to think. What do you have any preliminary thoughts on that? I think
330 00:32:13.360 ⇒ 00:32:17.329 Yoon: Yeah, I mean, I mean, there is a a metric that we can use
331 00:32:17.970 ⇒ 00:32:22.149 Yoon: within the the profit library. So I can. I can take a look at that
332 00:32:23.830 ⇒ 00:32:28.869 Payas Parab: Great. So then, let’s do that. I cause I’m just how many years of data do we have? How many years of data
333 00:32:28.870 ⇒ 00:32:31.940 Yoon: We? I think we had
334 00:32:32.120 ⇒ 00:32:38.430 Yoon: 5 years, but the 1st 2 was had too much noise, so it was better to just exclude it
335 00:32:38.740 ⇒ 00:32:39.849 Yoon: from the data
336 00:32:40.260 ⇒ 00:32:44.269 Payas Parab: Yeah, I I get, yeah, that makes sense. The 1st 2 years we exclude. I’m
337 00:32:44.750 ⇒ 00:32:50.720 Payas Parab: like in my head, I’m like, could we cross validate it right like, could we do like an in in?
338 00:32:51.740 ⇒ 00:33:00.609 Payas Parab: I think if you can find like a mean, absolute error for even just the training data like. I don’t think the cross value you’ll be able to like meaningfully. Come up with something
339 00:33:00.820 ⇒ 00:33:01.200 Yoon: Yeah.
340 00:33:01.200 ⇒ 00:33:05.554 Payas Parab: 3 years of data. That’s just my view. I I think
341 00:33:06.520 ⇒ 00:33:13.050 Payas Parab: scoring metric. What we might just look at right is in sample.
342 00:33:15.220 ⇒ 00:33:16.840 Yoon: Just a back test
343 00:33:16.840 ⇒ 00:33:20.329 Payas Parab: Yeah, yeah, exactly. In sample, dash back. Test
344 00:33:20.500 ⇒ 00:33:21.210 Yoon: Okay.
345 00:33:21.210 ⇒ 00:33:29.360 Payas Parab: Of, I’m thinking, mean absolute error, because it’s the most like meaningful to someone that isn’t
346 00:33:29.800 ⇒ 00:33:32.360 Payas Parab: like. I don’t think an R. Squared would help a lot
347 00:33:33.950 ⇒ 00:33:36.300 Yoon: Yeah, it should be. It should be a mean, absolute error.
348 00:33:36.300 ⇒ 00:33:45.029 Payas Parab: Yeah, student mean absolute error, basically like our. And that’s something that like Dan can also react to is like, if we can forecast it within the bounds of this
349 00:33:45.030 ⇒ 00:33:45.630 Yoon: Yeah.
350 00:33:45.871 ⇒ 00:33:59.630 Payas Parab: So let’s do. And and I would just look at in sample. I just don’t think an out of sample test like a cross validated out of sample test is gonna be really helpful. I’m okay with an in sample. There also isn’t a ton of data for regression in general. And you know what I mean.
351 00:33:59.810 ⇒ 00:34:00.600 Yoon: Yeah.
352 00:34:00.600 ⇒ 00:34:06.280 Payas Parab: There isn’t a ton of like individual data points. So okay.
353 00:34:06.530 ⇒ 00:34:12.870 Payas Parab: it looks like the confidence interval is like reasonable, though it’s not like. At least it’s not like a fat like the whole thing, right?
354 00:34:13.210 ⇒ 00:34:20.314 Payas Parab: Look like it has a reasonable confidence interval like if it was like showing much wider ones, I’d be concerned.
355 00:34:21.630 ⇒ 00:34:28.569 Payas Parab: But yeah, let’s I think that’s let’s do that alright. So and then this one is this done, Beau, can I Mark? This is done here
356 00:34:29.860 ⇒ 00:34:31.699 Yoon: Yeah, this, this, yeah, that’s fine.
357 00:34:32.480 ⇒ 00:34:33.150 Payas Parab: Awesome.
358 00:34:33.460 ⇒ 00:34:39.519 Payas Parab: So review initial forecast client, mock up streamload app. I’m gonna put Tbd.
359 00:34:39.780 ⇒ 00:34:43.609 Payas Parab: and then this one amber I’m gonna assign to you
360 00:34:43.870 ⇒ 00:34:46.370 Payas Parab: to talk to Dan about. Does that work
361 00:34:51.010 ⇒ 00:34:53.769 Payas Parab: awesome? Alright cool. Is that all team
362 00:34:54.880 ⇒ 00:34:56.659 Payas Parab: don’t wanna take up everyone’s time
363 00:34:56.989 ⇒ 00:35:04.050 Payas Parab: helpful. Bo. Sorry, Brian dash Luke. Sorry if you just got like, had to just sit through all this apologies
364 00:35:04.050 ⇒ 00:35:04.540 Yoon: Oh, no worries.
365 00:35:04.540 ⇒ 00:35:05.510 Luke Daque: Yeah.
366 00:35:05.860 ⇒ 00:35:09.490 Payas Parab: Alright awesome. Alright, bye, guys, we’ll talk soon.
367 00:35:10.390 ⇒ 00:35:11.230 Luke Daque: Bye.
368 00:35:11.590 ⇒ 00:35:12.170 Yoon: You too.