1 00:00:00,000 --> 00:00:08,120 Welcome to the Cannabis Data Science Meetup Group. 2 00:00:08,120 --> 00:00:13,460 Definitely in for a treat today, as always, going to cover a lot of data, should be some 3 00:00:13,460 --> 00:00:16,080 good research ideas for you. 4 00:00:16,080 --> 00:00:23,060 I'll talk about a couple that are really kind of picking my brain and then some ways that 5 00:00:23,060 --> 00:00:30,240 you could potentially help with this data wrangling if you're so inclined. 6 00:00:30,240 --> 00:00:35,820 But I think we're pushing some new frontier today. 7 00:00:35,820 --> 00:00:44,060 People have been asking for a lot of sales data, finally, finally wrangled north of a 8 00:00:44,060 --> 00:00:47,420 million sales receipts. 9 00:00:47,420 --> 00:00:51,000 So this is going to be good data to look at. 10 00:00:51,000 --> 00:00:59,360 I'll share with you all the trials and tribulations I went through in the data curation phase, 11 00:00:59,360 --> 00:01:05,600 which was non-trivial and would love for you to reproduce it if you wish, because that 12 00:01:05,600 --> 00:01:11,640 would add value just to see, yes, to verify that, yes, you can get the same results. 13 00:01:11,640 --> 00:01:19,780 And also another set of eyes will help maybe find any improvements that could be had. 14 00:01:19,780 --> 00:01:22,900 So there's a couple ways that code can be improved upon, and I'll point that out to 15 00:01:22,900 --> 00:01:23,900 you today. 16 00:01:23,900 --> 00:01:28,320 And then, of course, we'll get to some cool statistics. 17 00:01:28,320 --> 00:01:35,480 I know we were talking about HomeGrow last week and how that could provide an interesting 18 00:01:35,480 --> 00:01:36,700 incentive. 19 00:01:36,700 --> 00:01:44,700 But what are your thoughts before I go too, too, too much further into the data that I've 20 00:01:44,700 --> 00:01:46,480 been pawing around at? 21 00:01:46,480 --> 00:01:53,760 So long time classic, Candice, feel free to share about anything that you want, and then 22 00:01:53,760 --> 00:01:56,280 Luis, I'll get to you. 23 00:01:56,280 --> 00:02:04,560 But any, I owe you a worksheet from last time, and I actually penned it up, but I haven't 24 00:02:04,560 --> 00:02:05,560 shared it yet. 25 00:02:05,560 --> 00:02:11,320 So if you have any ideas there, but anything that you're working on? 26 00:02:11,320 --> 00:02:19,640 I'm still working with the private GPT on my GPUs. 27 00:02:19,640 --> 00:02:27,920 This is going to be a good project because, one, you've got the hardware, and then, two, 28 00:02:27,920 --> 00:02:37,760 I've got a sneaking suspicion that we may need a lot of kind of custom tailored code 29 00:02:37,760 --> 00:02:43,800 to kind of make the output fruitful. 30 00:02:43,800 --> 00:02:47,680 But I'll have some more thoughts to share on that momentarily. 31 00:02:47,680 --> 00:02:50,600 But how about you, Luis? 32 00:02:50,600 --> 00:02:52,760 Happy to see you at the group today. 33 00:02:52,760 --> 00:02:56,400 We'd love to hear about some of the things that you're interested in, especially, you 34 00:02:56,400 --> 00:02:58,280 know, Canvas data related. 35 00:02:58,280 --> 00:02:59,760 Thank you. 36 00:02:59,760 --> 00:03:02,680 It's good to be here. 37 00:03:02,680 --> 00:03:03,680 Good morning. 38 00:03:03,680 --> 00:03:16,880 I'm Luis Coyote, and I have been apparently an outlier in my ability to sort of successfully 39 00:03:16,880 --> 00:03:29,160 convince my state to not only make all Canvas certificate of analysis data public by default, 40 00:03:29,160 --> 00:03:33,640 but also to integrate it into the state's data portal. 41 00:03:33,640 --> 00:03:40,560 So that it no longer requires any sort of manual lever pulling in order to get access 42 00:03:40,560 --> 00:03:43,080 to that data. 43 00:03:43,080 --> 00:03:55,600 Unfortunately, access, as we know, does not come with guaranteed data quality or trustworthiness 44 00:03:55,600 --> 00:03:57,280 of data. 45 00:03:57,280 --> 00:04:04,280 So that's kind of the angle that I find myself attacking these days. 46 00:04:04,280 --> 00:04:10,080 Unfortunately it means I've been doing a lot more policy work and a lot less data science. 47 00:04:10,080 --> 00:04:16,520 You know, it's, I could come up with all the hypotheses in the world and test them all 48 00:04:16,520 --> 00:04:24,400 I want, but what good is that if I don't trust the inputs of the data into, you know, those 49 00:04:24,400 --> 00:04:27,480 regressions I'm running into whatever else I might be doing. 50 00:04:27,480 --> 00:04:40,640 So I think for me, what I'm most interested in is the continuing efforts of folks in other 51 00:04:40,640 --> 00:04:49,400 markets, particularly mass, to achieve the same parity of, you know, data access while 52 00:04:49,400 --> 00:04:56,360 also, you know, understanding that it goes hand in glove with the quality factor. 53 00:04:56,360 --> 00:05:08,040 And I'm also becoming more interested in centralized aggregations of data from different markets 54 00:05:08,040 --> 00:05:15,280 for purposes of sort of cross-market analysis. 55 00:05:15,280 --> 00:05:20,280 And I had been meaning to get involved with this group much sooner and I'm glad to finally 56 00:05:20,280 --> 00:05:22,000 be here. 57 00:05:22,000 --> 00:05:27,600 I've admired and respected the work that both of you have done. 58 00:05:27,600 --> 00:05:35,840 And so mostly, I guess I'll mostly be a kind of a fly on the wall at least to start, but 59 00:05:35,840 --> 00:05:43,880 I really would like to start using that, you know, the data science muscles instead of 60 00:05:43,880 --> 00:05:50,800 the policy crafting muscles because I'm a little bit burnt out on the policy side of 61 00:05:50,800 --> 00:05:58,040 things at this point and I'd like to get back into what I consider to be the more fun stuff. 62 00:05:58,040 --> 00:06:02,480 So maybe that requires a little bit of suspension of disbelief on my part in terms of the quality 63 00:06:02,480 --> 00:06:06,480 of the data that I'm working with currently, but maybe I'm through this group, I'll be 64 00:06:06,480 --> 00:06:12,280 open, my eyes will be opened to additional sources, additional outlets, different avenues 65 00:06:12,280 --> 00:06:16,560 that those pursuits can take that are not necessarily reliant solely on my own state's 66 00:06:16,560 --> 00:06:17,560 data. 67 00:06:17,560 --> 00:06:21,680 So once again, thank you for having me and I appreciate the kindness. 68 00:06:21,680 --> 00:06:23,400 Absolutely. 69 00:06:23,400 --> 00:06:24,400 Love it, Lou. 70 00:06:24,400 --> 00:06:29,120 And you brought up a thousand and one phenomenal points. 71 00:06:29,120 --> 00:06:30,920 Where to even begin? 72 00:06:30,920 --> 00:06:38,040 And well, we'll actually be doing just that today because like you said, it's easy to 73 00:06:38,040 --> 00:06:44,480 get caught up in the weeds and the arguments of the policies, but we're data scientists 74 00:06:44,480 --> 00:06:45,480 here. 75 00:06:45,480 --> 00:06:48,680 So what is our comparative advantage? 76 00:06:48,680 --> 00:06:55,560 It's actually calculating statistics or facts. 77 00:06:55,560 --> 00:07:06,960 So that it may seem irrelevant at times, but I think it adds a lot of value because, you 78 00:07:06,960 --> 00:07:13,600 know, these are statistics that then people can then actually ground their say policy 79 00:07:13,600 --> 00:07:22,160 arguments in because that's often an easy criticism if you're arguing for one point 80 00:07:22,160 --> 00:07:26,360 and then somebody says, oh, do you have any facts to back that up? 81 00:07:26,360 --> 00:07:32,840 Well, you want to have at least a few statistics in your back pocket. 82 00:07:32,840 --> 00:07:37,000 And so that's kind of what we're going to provide today. 83 00:07:37,000 --> 00:07:42,840 And then you said, why even have access to this data? 84 00:07:42,840 --> 00:07:46,440 Alex, welcome to the group. 85 00:07:46,440 --> 00:07:53,080 Just giving a quick synopsis of why it's even important to have cannabis data. 86 00:07:53,080 --> 00:07:57,840 And then what's the point of even calculating these statistics? 87 00:07:57,840 --> 00:08:04,880 And so basically, some of the states are publishing aggregates like aggregate statistics. 88 00:08:04,880 --> 00:08:08,680 So we've seen, oh, there's total sales by month. 89 00:08:08,680 --> 00:08:11,680 And that was actually something that I was going to point you out to. 90 00:08:11,680 --> 00:08:18,760 There is a like a Yahoo Finance article that was talking about the amount consumed by people 91 00:08:18,760 --> 00:08:19,760 in Seattle. 92 00:08:19,760 --> 00:08:25,080 And I'll get you the I can get you the link right now. 93 00:08:25,080 --> 00:08:29,360 And so they're talking about the amount consumed by people in Seattle. 94 00:08:29,360 --> 00:08:37,720 And I thought, well, as good data scientists, we can basically try to replicate their statistic. 95 00:08:37,720 --> 00:08:44,800 And then similarly for the state aggregates, because Washington state publishes the total 96 00:08:44,800 --> 00:08:51,920 amount sold, the total amount of cannabis sold, and so do other states like Massachusetts. 97 00:08:51,920 --> 00:08:55,960 But you kind of have to take it at face value. 98 00:08:55,960 --> 00:09:03,000 And what's cool is in Washington state, we can do a public records request, get the entire 99 00:09:03,000 --> 00:09:06,160 population of cannabis sales. 100 00:09:06,160 --> 00:09:12,480 So every single line item on every single receipt. 101 00:09:12,480 --> 00:09:15,760 And we can validate, right? 102 00:09:15,760 --> 00:09:19,760 We were talking about the importance of, you know, verifying statistics already. 103 00:09:19,760 --> 00:09:26,440 Well, we could potentially verify the Washington state totals, right? 104 00:09:26,440 --> 00:09:30,800 So they're telling us that certain amount is being sold per month. 105 00:09:30,800 --> 00:09:35,040 Well, we've got every single receipt. 106 00:09:35,040 --> 00:09:42,400 So it's just almost an accounting endeavor to just go through every single receipt and 107 00:09:42,400 --> 00:09:44,360 add them all up. 108 00:09:44,360 --> 00:09:49,280 And as I'll demonstrate today, it's non trivial. 109 00:09:49,280 --> 00:09:55,520 So whenever something's non trivial in statistics, you kind of want people to double check your 110 00:09:55,520 --> 00:09:56,520 work. 111 00:09:56,520 --> 00:10:00,720 And so that's, I think, going to be the big argument today for, say, states like Massachusetts 112 00:10:00,720 --> 00:10:08,800 is it's nice for them to publish, say, summary statistics. 113 00:10:08,800 --> 00:10:13,840 But you really, really want to see the raw data because there's so many assumptions that 114 00:10:13,840 --> 00:10:19,840 go into calculating aggregates. 115 00:10:19,840 --> 00:10:20,840 So it's worthwhile. 116 00:10:20,840 --> 00:10:32,120 And while I find this article, Alex, would you want to say a word for yourself? 117 00:10:32,120 --> 00:10:34,240 Maybe what brought you to the group? 118 00:10:34,240 --> 00:10:41,800 And maybe, you know, what do you hope to get out of the marrying of cannabis plus data 119 00:10:41,800 --> 00:10:44,560 science? 120 00:10:44,560 --> 00:10:45,560 Thanks for having me. 121 00:10:45,560 --> 00:10:46,560 Sorry for running behind. 122 00:10:46,560 --> 00:10:52,520 But, yeah, I just I've seen your posts on LinkedIn and I love cannabis and I love data. 123 00:10:52,520 --> 00:10:58,000 I'm a business intelligence student, online student, and just hoping to learn. 124 00:10:58,000 --> 00:10:59,920 I'll follow along the parts I can follow. 125 00:10:59,920 --> 00:11:05,120 And if it's above my level, I'll just listen and figure it out next time, hopefully. 126 00:11:05,120 --> 00:11:08,120 Thank you. 127 00:11:08,120 --> 00:11:16,160 Phenomenal and we cover the gamut as far as cannabis data goes. 128 00:11:16,160 --> 00:11:18,880 So I started out working at a laboratory. 129 00:11:18,880 --> 00:11:26,280 So my heart is a lot in the chemistry side of the cannabis industry. 130 00:11:26,280 --> 00:11:34,520 So I love learning more about, you know, what actually the chemical constituents of the 131 00:11:34,520 --> 00:11:40,280 cannabis plant and then I've becoming increasingly interested in the agriculture side. 132 00:11:40,280 --> 00:11:44,880 So how do people actually cultivate and grow the plant? 133 00:11:44,880 --> 00:11:49,400 Actually increasingly interested in hemp because it's still the cannabis plant at the end of 134 00:11:49,400 --> 00:11:51,760 the day. 135 00:11:51,760 --> 00:11:55,760 The processing side is the part I know the least about. 136 00:11:55,760 --> 00:11:57,520 And then retail. 137 00:11:57,520 --> 00:12:02,200 I come at this from an economics point of view. 138 00:12:02,200 --> 00:12:08,800 So sometimes I maybe get too lost in the weeds when we really start talking about sales and 139 00:12:08,800 --> 00:12:10,680 market structure and things like that. 140 00:12:10,680 --> 00:12:12,680 But some people find it interesting. 141 00:12:12,680 --> 00:12:17,400 So tons of people find different things of interest. 142 00:12:17,400 --> 00:12:25,480 But one thing I love to do is see sort of what people are talking about in various news 143 00:12:25,480 --> 00:12:28,360 outlets and in the mainstream. 144 00:12:28,360 --> 00:12:33,240 Like I said, try to replicate some of their statistics sometimes, always for fun. 145 00:12:33,240 --> 00:12:36,760 But welcome to the group, John. 146 00:12:36,760 --> 00:12:42,720 We'd love to hear what's your angle and what do you hope to get out of the group? 147 00:12:42,720 --> 00:12:43,720 Yeah. 148 00:12:43,720 --> 00:12:45,840 So hi, all. 149 00:12:45,840 --> 00:12:51,920 So I'm a company called Genetica and we do cannabis strain matching. 150 00:12:51,920 --> 00:12:55,720 So matching people's effects to the strains. 151 00:12:55,720 --> 00:13:04,160 And I've been watching your work for a while now and I just really love your open source 152 00:13:04,160 --> 00:13:07,680 tools and the videos and the code. 153 00:13:07,680 --> 00:13:17,200 And it helps provide data sources that we can leverage to really understand, to your 154 00:13:17,200 --> 00:13:24,100 point, the science behind the mix of like, what are all these cannabinoids and terpenes 155 00:13:24,100 --> 00:13:29,000 and how much do they actually influence the total effect that somebody feels? 156 00:13:29,000 --> 00:13:31,640 And doing that on a granular level. 157 00:13:31,640 --> 00:13:37,160 And there's a lot of data out there about effects to a strain, but like not down to 158 00:13:37,160 --> 00:13:38,920 the to the COA level. 159 00:13:38,920 --> 00:13:46,560 So really trying to connect a batch to a review and to a strain and really understand all 160 00:13:46,560 --> 00:13:53,160 the causal factors to, you know, we work with understanding and we work with companies where 161 00:13:53,160 --> 00:14:00,460 we can look at the DNA of people and of the strain to make a match to that level as well. 162 00:14:00,460 --> 00:14:05,960 So that's really kind of where my head's at is right. 163 00:14:05,960 --> 00:14:11,040 There's retail and the things that make you money and do stuff like that. 164 00:14:11,040 --> 00:14:17,080 But my passion's in the science and really understanding the data and what's behind it. 165 00:14:17,080 --> 00:14:19,960 It's great to be here. 166 00:14:19,960 --> 00:14:26,280 I absolutely love it and you've definitely got some big things coming in your space because 167 00:14:26,280 --> 00:14:33,300 I think what's held back a lot of the research is there is not a lot of funding for things 168 00:14:33,300 --> 00:14:38,600 like clinical trials because none of the. 169 00:14:38,600 --> 00:14:45,080 Actually, I don't want to say none because there are some people at academic institutes 170 00:14:45,080 --> 00:14:50,400 doing various types of research, but I don't know how much and how much funding they have. 171 00:14:50,400 --> 00:14:57,160 But a lot of the academic institutions shy away from cannabis research because they don't 172 00:14:57,160 --> 00:15:05,120 want to get their funding stripped away and probably some of the nonprofits, too, for 173 00:15:05,120 --> 00:15:07,000 that matter. 174 00:15:07,000 --> 00:15:13,240 So I love the work that you're doing, and that's really going to be an exciting frontier 175 00:15:13,240 --> 00:15:19,840 and you're going to end up taking it so much further than anything that I could do. 176 00:15:19,840 --> 00:15:22,640 Because like I said, I studied economics. 177 00:15:22,640 --> 00:15:28,400 I wish I had studied chemistry or I still can and still trying, but it's a difficult 178 00:15:28,400 --> 00:15:29,400 subject. 179 00:15:29,400 --> 00:15:34,460 And then, like you said, then you're trying to get into the biochemistry. 180 00:15:34,460 --> 00:15:43,160 So now you're going to need to know all about the human body and extensive biology. 181 00:15:43,160 --> 00:15:51,800 And so it's it's a complicated subject, but I love love your work. 182 00:15:51,800 --> 00:15:57,960 Yeah, no, it's yeah, it's there's no end to the learning that you can do. 183 00:15:57,960 --> 00:16:04,560 And there's always something more that you can, I guess, kind of use the sharp in your 184 00:16:04,560 --> 00:16:08,040 sword because there's a lot of angles to it. 185 00:16:08,040 --> 00:16:15,520 And everything from social determinants of health to like the mental psychology aspects. 186 00:16:15,520 --> 00:16:18,680 I mean, there's a lot of a lot of legs where it could go. 187 00:16:18,680 --> 00:16:20,960 And it's helping people. 188 00:16:20,960 --> 00:16:24,240 So that's what I'm excited about. 189 00:16:24,240 --> 00:16:25,240 Love it, John. 190 00:16:25,240 --> 00:16:26,240 Keep up the good work. 191 00:16:26,240 --> 00:16:32,360 And you're always welcome to use the Canvas Data Science Meetup as sort of a platform. 192 00:16:32,360 --> 00:16:38,800 So a lot of times I'll talk about my latest research ideas or some of the things that 193 00:16:38,800 --> 00:16:40,080 I'm tinkering on. 194 00:16:40,080 --> 00:16:45,520 And you're always welcome to, you know, you're even welcome to prepare a presentation and 195 00:16:45,520 --> 00:16:47,240 come take the floor. 196 00:16:47,240 --> 00:16:49,280 So I love that. 197 00:16:49,280 --> 00:16:50,960 Yeah, absolutely. 198 00:16:50,960 --> 00:16:58,000 Well, well, Jatin, just let me know if I mispronounce your name for any reason. 199 00:16:58,000 --> 00:17:00,320 But welcome to the group. 200 00:17:00,320 --> 00:17:02,200 Happy to have you here. 201 00:17:02,200 --> 00:17:08,560 Would love to hear your angle and what do you hope to walk away at the end of the day 202 00:17:08,560 --> 00:17:10,840 from the Canvas Data Science Meetup? 203 00:17:10,840 --> 00:17:13,920 Hi, morning, everybody. 204 00:17:13,920 --> 00:17:15,940 I completely agree with you. 205 00:17:15,940 --> 00:17:23,960 So I'm a PhD student and I haven't seen any data or analytics on cannabis. 206 00:17:23,960 --> 00:17:27,960 And I was intrigued when I saw this meetup and I was like, that's the kind of meetup 207 00:17:27,960 --> 00:17:35,080 that I would definitely want to go for, academia always shies away from this kind of topics. 208 00:17:35,080 --> 00:17:36,720 And I'm not sure. 209 00:17:36,720 --> 00:17:42,760 And even I try to talk to professors about this, like, why do we go away from these topics 210 00:17:42,760 --> 00:17:47,160 or even like with different, I go to different meetups as well. 211 00:17:47,160 --> 00:17:51,120 So I generally talk to people in health institutions as well. 212 00:17:51,120 --> 00:17:52,120 Why do we go away? 213 00:17:52,120 --> 00:17:53,900 Why don't we talk about this? 214 00:17:53,900 --> 00:17:56,840 Why there's not much research? 215 00:17:56,840 --> 00:17:58,880 So I haven't been having those kind of answers. 216 00:17:58,880 --> 00:18:00,920 So I was like, so forget it. 217 00:18:00,920 --> 00:18:03,960 If I'm not going to get answers from academia, I'll just make some of mine. 218 00:18:03,960 --> 00:18:06,160 I'll just go and do some work on my own. 219 00:18:06,160 --> 00:18:10,480 So that's why I was like, OK, it should be a good meetup. 220 00:18:10,480 --> 00:18:11,480 I love it. 221 00:18:11,480 --> 00:18:14,480 That's the true academic spirit, right? 222 00:18:14,480 --> 00:18:18,880 Just, you know, go where no one's willing to go before. 223 00:18:18,880 --> 00:18:19,880 Exactly. 224 00:18:19,880 --> 00:18:25,840 Well, as promised, I'll get some good data in your hands. 225 00:18:25,840 --> 00:18:36,280 So here's basically a rich gold vein of data that we keep mining. 226 00:18:36,280 --> 00:18:39,400 So this is just the Washington state data. 227 00:18:39,400 --> 00:18:49,920 But this is a good demonstration of how, you know, if you keep at it, exactly, then it's 228 00:18:49,920 --> 00:18:53,920 kind of remarkable how much data is really there. 229 00:18:53,920 --> 00:19:02,120 And in fact, if you're interested, maybe I can share my screen with all of you and kind 230 00:19:02,120 --> 00:19:08,320 of just give a quick demo of what we've been able to pull off with this data, maybe a 15, 231 00:19:08,320 --> 00:19:11,240 20 minute demo, and then we can talk about it afterwards. 232 00:19:11,240 --> 00:19:12,240 Sure, sure. 233 00:19:12,240 --> 00:19:13,240 Go ahead. 234 00:19:13,240 --> 00:19:14,240 OK. 235 00:19:14,240 --> 00:19:15,240 So. 236 00:19:15,240 --> 00:19:24,360 It looks like. 237 00:19:24,360 --> 00:19:33,160 OK, so if for any reason you can't see my screen, you know, just let me know. 238 00:19:33,160 --> 00:19:37,640 It looks like they kind of toggle the user interface a little. 239 00:19:37,640 --> 00:19:38,640 OK. 240 00:19:38,640 --> 00:19:42,720 Here, I'll just start up something new here. 241 00:19:42,720 --> 00:19:45,600 OK, so what are we even working with here? 242 00:19:45,600 --> 00:19:54,320 So if you do a public records request to the Washington State Liquor and Cannabis Board, 243 00:19:54,320 --> 00:19:57,840 just let me know if you need to see the actual language. 244 00:19:57,840 --> 00:20:06,340 But many people do this request, so they have a nice link prepared and they'll give you 245 00:20:06,340 --> 00:20:09,280 a nice zip file. 246 00:20:09,280 --> 00:20:18,400 You know, this one is the latest, you know, seven gigabytes zipped and then many, many 247 00:20:18,400 --> 00:20:27,040 gigabytes, maybe north of 60, so 63 gigabytes of data here. 248 00:20:27,040 --> 00:20:31,400 And we've done many cool things here, like looking at string names. 249 00:20:31,400 --> 00:20:38,920 We've done a lot of analysis of lab results, so there's a lot there. 250 00:20:38,920 --> 00:20:47,360 Just to be frank, one of the hardest times I'm having right now is basically. 251 00:20:47,360 --> 00:20:52,240 And I'll post all this code today and point you in the direction of it. 252 00:20:52,240 --> 00:20:55,680 But we're basically trying to. 253 00:20:55,680 --> 00:21:04,080 Actually, I don't think I've created a diagram yet of a actually. 254 00:21:04,080 --> 00:21:07,640 Yeah, we have a diagram here. 255 00:21:07,640 --> 00:21:14,000 I don't know if I'm going to be able to find it. 256 00:21:14,000 --> 00:21:21,320 So yeah, since I won't be able to find the diagram here, I'll just point you. 257 00:21:21,320 --> 00:21:23,120 I'll share it with you afterwards. 258 00:21:23,120 --> 00:21:28,120 But we've got. 259 00:21:28,120 --> 00:21:31,200 We looked at the low hanging fruit, right? 260 00:21:31,200 --> 00:21:35,440 Looking at licensee details, lab results, strains. 261 00:21:35,440 --> 00:21:40,520 And I finally wanted to get around to sales and so just wanted to be frank with you about 262 00:21:40,520 --> 00:21:47,440 like all the difficulties of putting these sales items together. 263 00:21:47,440 --> 00:21:54,120 That way, maybe you can have some insights and do a better job than I. 264 00:21:54,120 --> 00:21:56,680 Can I interrupt you just for a second? 265 00:21:56,680 --> 00:21:57,680 Yes, please. 266 00:21:57,680 --> 00:22:04,520 So this might be a silly question and I apologize if so, but I'm just trying to understand in 267 00:22:04,520 --> 00:22:13,040 terms of the tooling, I see you've got VS code connected to CDS. 268 00:22:13,040 --> 00:22:23,080 So I'm just wondering if you could like, you know, go through like how you set up that 269 00:22:23,080 --> 00:22:28,600 within the within VS code. 270 00:22:28,600 --> 00:22:32,800 Set up this interactive terminal. 271 00:22:32,800 --> 00:22:35,080 Yeah, yeah. 272 00:22:35,080 --> 00:22:36,080 OK. 273 00:22:36,080 --> 00:22:40,560 Or if it's if that's out of scope for what we're talking about right now, if you push 274 00:22:40,560 --> 00:22:45,520 control shift and P. 275 00:22:45,520 --> 00:22:52,120 That once again, I may have toggled my VS code, so if this doesn't look like how yours 276 00:22:52,120 --> 00:22:57,080 does, then you know, feel free to email me afterwards and we can make sure to set up 277 00:22:57,080 --> 00:22:58,080 your environment. 278 00:22:58,080 --> 00:23:05,560 But the way I go about running code is I make a Jupiter window so that you can type. 279 00:23:05,560 --> 00:23:09,240 I think you can just type up here Jupiter. 280 00:23:09,240 --> 00:23:13,880 So these are Jupiter notebooks that OK, all right. 281 00:23:13,880 --> 00:23:14,880 Exactly. 282 00:23:14,880 --> 00:23:16,840 That was the piece I was missing there. 283 00:23:16,840 --> 00:23:17,840 Thank you. 284 00:23:17,840 --> 00:23:28,040 So that's that's just kind of how I like to play around with code and develop just to. 285 00:23:28,040 --> 00:23:30,560 You'll you'll see my process here momentarily. 286 00:23:30,560 --> 00:23:36,360 I just kind of sometimes like to. 287 00:23:36,360 --> 00:23:39,960 Write code on the fly, but. 288 00:23:39,960 --> 00:23:46,800 I just you can just create an interactive window, so that should be one of your commands. 289 00:23:46,800 --> 00:23:51,520 And then this will connect to your. 290 00:23:51,520 --> 00:23:54,760 Your your Python environment in here. 291 00:23:54,760 --> 00:23:59,640 I've actually set up a. 292 00:23:59,640 --> 00:24:04,480 A virtual environment where. 293 00:24:04,480 --> 00:24:11,520 Which actually may not necessarily be necessary for for the code we're running today, but. 294 00:24:11,520 --> 00:24:14,760 But you can set up a virtual environment. 295 00:24:14,760 --> 00:24:18,440 Yeah, I didn't want to derail the conversation. 296 00:24:18,440 --> 00:24:22,800 Maybe be on the scope for today, but like I said, if you need any help getting the code 297 00:24:22,800 --> 00:24:27,360 to run, I'm more than happy to help with that, especially happy to help with that. 298 00:24:27,360 --> 00:24:29,280 In fact, awesome, thank you. 299 00:24:29,280 --> 00:24:35,960 So but that's a big ugly piece of code, and so this is a smaller script. 300 00:24:35,960 --> 00:24:40,440 And so basically. 301 00:24:40,440 --> 00:24:42,440 I've gone through. 302 00:24:42,440 --> 00:24:49,640 Yes, this is what we're running here and just started to tie together all of these data 303 00:24:49,640 --> 00:24:50,640 sets. 304 00:24:50,640 --> 00:25:01,080 So we've got like basically every line item on the receipt is called a like a sales detail. 305 00:25:01,080 --> 00:25:04,720 So if you actually look at that. 306 00:25:04,720 --> 00:25:06,760 So. 307 00:25:06,760 --> 00:25:12,880 Look at the latest one. 308 00:25:12,880 --> 00:25:18,600 And keep in mind, so this is basically what they so they give you a bunch of their their 309 00:25:18,600 --> 00:25:21,000 data sets. 310 00:25:21,000 --> 00:25:29,960 So there's, you know, 102 sales details. 311 00:25:29,960 --> 00:25:35,000 And if you each one is large, so I'll show you what this looks like. 312 00:25:35,000 --> 00:25:43,600 And basically what the script I've written does is it just iterates through all of the 313 00:25:43,600 --> 00:25:51,840 details and just matches it with all of the pertinent details that are necessary. 314 00:25:51,840 --> 00:25:59,400 Because if you look at this, this is what they they give you for a sales receipt. 315 00:25:59,400 --> 00:26:03,800 So they just give you the ID. 316 00:26:03,800 --> 00:26:07,520 Then you know you have the inventory ID. 317 00:26:07,520 --> 00:26:09,700 Which is critical. 318 00:26:09,700 --> 00:26:17,200 And then you know they tell you the how much you sold for any discount. 319 00:26:17,200 --> 00:26:19,960 And then the sales tax. 320 00:26:19,960 --> 00:26:24,440 And I think this is the excise tax. 321 00:26:24,440 --> 00:26:28,720 So those are the mission critical pieces here. 322 00:26:28,720 --> 00:26:32,480 But you know what are people always interested in? 323 00:26:32,480 --> 00:26:42,040 There's they want to know how much flowers sold or how many edibles were sold or which 324 00:26:42,040 --> 00:26:45,960 producer has the highest sales. 325 00:26:45,960 --> 00:26:53,000 And so we have to now do is that information is not here. 326 00:26:53,000 --> 00:26:59,600 But if you kind of look at how the data is structured, oh, like you can match the sales 327 00:26:59,600 --> 00:27:03,960 detail with the header. 328 00:27:03,960 --> 00:27:08,840 And so the header is basically all the information on the receipt. 329 00:27:08,840 --> 00:27:12,440 So this is you can think of this as a receipt. 330 00:27:12,440 --> 00:27:18,720 And each one of these is a line item on the receipt. 331 00:27:18,720 --> 00:27:21,840 Okay so that's you know one match you have to make. 332 00:27:21,840 --> 00:27:29,560 And then you can you know then you find out okay who actually sold this item. 333 00:27:29,560 --> 00:27:30,560 Wonderful. 334 00:27:30,560 --> 00:27:41,880 And then then you know you have to match the inventory with the inventory item by the inventory 335 00:27:41,880 --> 00:27:44,960 ID. 336 00:27:44,960 --> 00:27:48,400 And then then it's just the big game of matching. 337 00:27:48,400 --> 00:27:55,120 Because then maybe you're interested in knowing what was the most popular strain that sold 338 00:27:55,120 --> 00:27:57,080 in Washington state. 339 00:27:57,080 --> 00:28:05,960 Well now you've got to match the strain ID with the strain data to get the strain name. 340 00:28:05,960 --> 00:28:12,720 So it's just this really kind of convoluted matching game that we have to play. 341 00:28:12,720 --> 00:28:23,760 And basically the one thing that I cannot get matched is I you should be from my understanding 342 00:28:23,760 --> 00:28:33,480 you can and from what we've done it sure appears that you can match the lab results to a subset 343 00:28:33,480 --> 00:28:36,640 of the inventory. 344 00:28:36,640 --> 00:28:40,360 Which makes sense. 345 00:28:40,360 --> 00:28:44,340 People are getting the end products tested. 346 00:28:44,340 --> 00:28:48,160 So any intermediary product may not get tested. 347 00:28:48,160 --> 00:28:55,120 But the idea is they test the end product and then maybe that gets sold. 348 00:28:55,120 --> 00:29:02,080 But I'm not I haven't quite figured out from my understanding this actually may be the 349 00:29:02,080 --> 00:29:08,040 problem and this is kind of how what Washington state licensees think about it. 350 00:29:08,040 --> 00:29:15,040 I think they think about it is in they send in what's called like a parent lot to get 351 00:29:15,040 --> 00:29:16,160 tested. 352 00:29:16,160 --> 00:29:18,400 So that's one inventory item. 353 00:29:18,400 --> 00:29:24,040 And then I think they then split it up into a bunch of children lots. 354 00:29:24,040 --> 00:29:30,080 And so it's basically like they send in an inventory item representing a five pound lot 355 00:29:30,080 --> 00:29:32,280 or what have you. 356 00:29:32,280 --> 00:29:38,360 And then that comes back and then they then divvy that up into a bunch of different inventory 357 00:29:38,360 --> 00:29:40,240 items maybe. 358 00:29:40,240 --> 00:29:43,200 They then go off and get sold. 359 00:29:43,200 --> 00:29:47,080 And maybe that's where the link is breaking. 360 00:29:47,080 --> 00:29:53,600 But but basically you've got an inventory ID here. 361 00:29:53,600 --> 00:29:58,480 And so if you look at the lab results there's an inventory ID there. 362 00:29:58,480 --> 00:30:07,560 So one would think that you could match lab results to inventory items and then consequently 363 00:30:07,560 --> 00:30:11,400 match that with sales. 364 00:30:11,400 --> 00:30:13,920 Did I tried everything under this. 365 00:30:13,920 --> 00:30:16,440 Well maybe maybe I haven't tried everything under this. 366 00:30:16,440 --> 00:30:22,080 Like I said I think there's still something that you could think of here where I think 367 00:30:22,080 --> 00:30:30,520 there's you have to almost match the parent inventory with the child inventory or there's 368 00:30:30,520 --> 00:30:35,480 another data set here products. 369 00:30:35,480 --> 00:30:39,600 I haven't quite figured out how that fits in the mix. 370 00:30:39,600 --> 00:30:48,120 But that's really what's needed because when you buy a product at the store you get the 371 00:30:48,120 --> 00:30:56,880 receipt and then on the label they have you know the THC and CBD percentage. 372 00:30:56,880 --> 00:31:06,360 So for us to get a complete observation here right we need to know the unit price and then 373 00:31:06,360 --> 00:31:14,400 we also have to know the like the THC and CBD content. 374 00:31:14,400 --> 00:31:17,480 And so I'll show you how far I was able to take this. 375 00:31:17,480 --> 00:31:23,600 And so basically I've just been what I call augmenting all of these. 376 00:31:23,600 --> 00:31:33,240 And so right if you just look at this same same data set but here I've augmented every 377 00:31:33,240 --> 00:31:39,320 data point that I can. 378 00:31:39,320 --> 00:31:45,760 So once again other people may have different approaches. 379 00:31:45,760 --> 00:31:52,520 So if you're an SQL wizard then you may actually have better luck just putting all of this 380 00:31:52,520 --> 00:32:03,480 in an SQL database and you know taking it from there. 381 00:32:03,480 --> 00:32:18,180 And in fact some of the so you see some of these actually most of these are incomplete. 382 00:32:18,180 --> 00:32:25,800 So for example you know like out of all of these sales I was only really able to map 383 00:32:25,800 --> 00:32:34,960 like match back you know a small percentage to strain. 384 00:32:34,960 --> 00:32:41,200 And it you know theoretically you know there you know theoretically every single one of 385 00:32:41,200 --> 00:32:50,800 these should should map back to a complete observation. 386 00:32:50,800 --> 00:33:00,280 So what I'm going to basically show you today is just a tiny tiny subset of what I've been 387 00:33:00,280 --> 00:33:03,440 able to wrangle. 388 00:33:03,440 --> 00:33:12,480 So here let me open one more of these just to see if maybe maybe and maybe for some reason 389 00:33:12,480 --> 00:33:18,320 that tail end one had a low match rate. 390 00:33:18,320 --> 00:33:22,480 But let's maybe not maybe there is just low matches all around. 391 00:33:22,480 --> 00:33:37,960 So let me open one more and then I'll share with you the analysis. 392 00:33:37,960 --> 00:33:43,680 Because basically this is the so it looks like it's low matching all across the board. 393 00:33:43,680 --> 00:33:50,920 So this is basically the final like the final frontier or you know the front the final hurdle 394 00:33:50,920 --> 00:34:00,720 that we have to to get over here is is there any way to you know completely flesh out this 395 00:34:00,720 --> 00:34:03,960 data set here. 396 00:34:03,960 --> 00:34:06,440 I didn't realize how sparse it was. 397 00:34:06,440 --> 00:34:14,280 So basically I've you know been running this script and this is sort of the one of the 398 00:34:14,280 --> 00:34:19,720 first times I've really done a really in-depth look and you're in wow we're really missing 399 00:34:19,720 --> 00:34:25,600 a ton of licensees here. 400 00:34:25,600 --> 00:34:34,640 So that's unfortunate but the jar half full view is you know the ones that we are able 401 00:34:34,640 --> 00:34:36,880 to match back. 402 00:34:36,880 --> 00:34:41,600 Well check this out we can actually start doing some some cool things here. 403 00:34:41,600 --> 00:34:48,640 So so here's just just a random licensee. 404 00:34:48,640 --> 00:34:57,640 So basically we can just start you know looking at you know all the you know the sales that 405 00:34:57,640 --> 00:35:07,760 actually that one is new. 406 00:35:07,760 --> 00:35:09,920 So this one looks a bit more complete. 407 00:35:09,920 --> 00:35:12,480 So one two three four five six seven. 408 00:35:12,480 --> 00:35:23,200 So so June should have been the last full month and you know so this is you know a bit 409 00:35:23,200 --> 00:35:31,460 better of data you know if you just sort of look at you know a single licensee. 410 00:35:31,460 --> 00:35:45,360 So maybe for however this licensee operates maybe they don't do the the parent child relationship 411 00:35:45,360 --> 00:35:47,220 with their inventory. 412 00:35:47,220 --> 00:35:53,320 Maybe for whatever reason the way this licensee likes to operate once again I'm conjecturing. 413 00:35:53,320 --> 00:35:59,080 Maybe they just like just a pure one to one relationship all the way through. 414 00:35:59,080 --> 00:36:06,640 Who knows or maybe we're only capturing a tiny subset of sales for this licensee. 415 00:36:06,640 --> 00:36:11,560 But basically we have a benchmark right. 416 00:36:11,560 --> 00:36:18,960 We have the total sales published by the Washington State Liquor and Cannabis Board. 417 00:36:18,960 --> 00:36:27,060 So basically you know the idea here is you know can we just basically go through sum 418 00:36:27,060 --> 00:36:34,480 up all the sales and compare it to their total. 419 00:36:34,480 --> 00:36:43,880 But let's just see here real quick if there's anything of interest here. 420 00:36:43,880 --> 00:36:55,280 So this looks like this was Firehouse in Ellensburg and you know lots of good data there. 421 00:36:55,280 --> 00:37:01,280 And like I said this could potentially provide you know an ambitious data scientist with 422 00:37:01,280 --> 00:37:08,600 you know an endless amount of work right because you know maybe you can dive into and once 423 00:37:08,600 --> 00:37:14,360 again this may not be their complete amount of sales but you can start you know doing 424 00:37:14,360 --> 00:37:18,800 analysis you know on a retail by retail basis. 425 00:37:18,800 --> 00:37:23,120 And I've always you know had the grand idea of oh you know wouldn't it be cool to you 426 00:37:23,120 --> 00:37:31,240 know reach out to some of these retailers if you find you know some I don't know you 427 00:37:31,240 --> 00:37:39,080 know quirk in their data that you know you can say like oh maybe there is a you're you 428 00:37:39,080 --> 00:37:46,280 know you're selling a lot more liquid edibles than everybody else at a lower price or you 429 00:37:46,280 --> 00:37:53,360 know or maybe you're not selling as many liquid edibles as everybody else or you know who 430 00:37:53,360 --> 00:38:03,400 knows that's you know for a data scientist to look at their questions at hand. 431 00:38:03,400 --> 00:38:13,920 Okay so that's the preface that this is still messy work in progress but the idea is once 432 00:38:13,920 --> 00:38:23,440 you have all of these curated and you know here I'm just going to read in all of the 433 00:38:23,440 --> 00:38:26,000 the licensee sales. 434 00:38:26,000 --> 00:38:35,800 So these are just all the sales that are you know fairly well identified from receipt to 435 00:38:35,800 --> 00:38:39,960 to seller and to inventory. 436 00:38:39,960 --> 00:38:45,760 So I'll just start reading these in and so this could take a hot second. 437 00:38:45,760 --> 00:38:49,200 Oops need some packages. 438 00:38:49,200 --> 00:38:57,840 So I'll just be reading this in here and then the classic like the baking show where you 439 00:38:57,840 --> 00:39:04,320 know you already have everything baked. 440 00:39:04,320 --> 00:39:08,200 Here I've already read it in all of the data. 441 00:39:08,200 --> 00:39:10,160 So here it is. 442 00:39:10,160 --> 00:39:14,400 It shouldn't start chugging along here in a second but like I said it'll take about 443 00:39:14,400 --> 00:39:16,400 a minute or two. 444 00:39:16,400 --> 00:39:23,920 But and once again just to show you what this looks like you can basically just take a random 445 00:39:23,920 --> 00:39:30,660 sample and just look at one of these observations here. 446 00:39:30,660 --> 00:39:43,320 So this was usable cannabis sold at Zips cannabis in Tacoma. 447 00:39:43,320 --> 00:39:55,080 It was sold the January 30th 2023. 448 00:39:55,080 --> 00:40:05,600 And it was this lemon cherry gelato pre-roll you know and then you you've got the strain 449 00:40:05,600 --> 00:40:09,600 lemon cherry gelato. 450 00:40:09,600 --> 00:40:15,960 It looks like this may have cost Zips maybe 450. 451 00:40:15,960 --> 00:40:22,000 Once again I wouldn't read too much into these numbers and then Lou this is where you talked 452 00:40:22,000 --> 00:40:25,800 about data quality. 453 00:40:25,800 --> 00:40:38,880 You have to be super careful about your insights because you can't rely too too much on or 454 00:40:38,880 --> 00:40:45,720 maybe that's sort of the art where data science goes is maybe you may have to just exclude 455 00:40:45,720 --> 00:40:57,080 any licensees who aren't accurately recording their costs or this or that because you see 456 00:40:57,080 --> 00:40:58,640 wild things here right. 457 00:40:58,640 --> 00:41:01,480 You see negative number right. 458 00:41:01,480 --> 00:41:13,760 So for example we could just say I wonder if we can I wonder how this works. 459 00:41:13,760 --> 00:41:18,600 Something else worth mentioning is that it might be one of these like steel sharp and 460 00:41:18,600 --> 00:41:22,740 steel scenarios with you know the overall data set. 461 00:41:22,740 --> 00:41:30,400 You may be able to use some of the data to expose quality issues with other parts of 462 00:41:30,400 --> 00:41:32,520 the data. 463 00:41:32,520 --> 00:41:34,960 So that's always worth keeping in mind. 464 00:41:34,960 --> 00:41:35,960 Exactly. 465 00:41:35,960 --> 00:41:42,920 But you see you know there are some negatives there so just you know be wary about that. 466 00:41:42,920 --> 00:41:45,680 You know if you're summing up costs then there's a negative. 467 00:41:45,680 --> 00:41:49,560 How do you handle that? 468 00:41:49,560 --> 00:41:57,820 And you'll see a similar thing with discounts where the discount field is also a really 469 00:41:57,820 --> 00:41:59,320 messy field. 470 00:41:59,320 --> 00:42:03,600 But there could be insights there. 471 00:42:03,600 --> 00:42:10,760 You know like what like you know there could be there could be really critical insights 472 00:42:10,760 --> 00:42:12,040 there. 473 00:42:12,040 --> 00:42:20,600 I mean that's what economists are always studying is price and they're talking about real price. 474 00:42:20,600 --> 00:42:23,920 You know what's the real price the consumer pays. 475 00:42:23,920 --> 00:42:29,280 So you're doing cost studies price studies. 476 00:42:29,280 --> 00:42:33,960 You definitely want to have a nice accurate measure of the discounts. 477 00:42:33,960 --> 00:42:41,080 As I said I haven't been able to map back the lab results yet. 478 00:42:41,080 --> 00:42:45,960 And this is something that of course everybody's super interested in. 479 00:42:45,960 --> 00:42:54,640 And the one question that has been on everybody's mind and we've been asking this question now 480 00:42:54,640 --> 00:43:01,800 for over two years is how much does THC matter. 481 00:43:01,800 --> 00:43:07,960 First did the I mean first the question is does THC matter. 482 00:43:07,960 --> 00:43:10,100 Just a binary yes or no. 483 00:43:10,100 --> 00:43:15,040 You can basically just do and we've said all of these those two questions. 484 00:43:15,040 --> 00:43:17,960 Does does THC matter. 485 00:43:17,960 --> 00:43:28,960 And if so how much does THC matter can be run can be estimated by just a simple regression 486 00:43:28,960 --> 00:43:37,560 basically of price on. 487 00:43:37,560 --> 00:43:42,760 And total THC look they're even right beside each other. 488 00:43:42,760 --> 00:43:44,600 So that that's all you need. 489 00:43:44,600 --> 00:43:49,600 But unfortunately you know the data should be there. 490 00:43:49,600 --> 00:43:53,600 But I cried pretty extensively. 491 00:43:53,600 --> 00:43:59,000 And so like I said so this is where you could potentially have a breakthrough because if 492 00:43:59,000 --> 00:44:05,960 you can match the lab results to which which we have. 493 00:44:05,960 --> 00:44:06,960 Right. 494 00:44:06,960 --> 00:44:13,840 So you have these nice curated lab results. 495 00:44:13,840 --> 00:44:15,520 So you're all show you the two ones. 496 00:44:15,520 --> 00:44:21,400 So here's lab results that haven't been matched inventory. 497 00:44:21,400 --> 00:44:28,080 And then these looks like there's some problems. 498 00:44:28,080 --> 00:44:32,760 And then here are some lab results that have been matched to inventory. 499 00:44:32,760 --> 00:44:37,000 So I'll show you those two. 500 00:44:37,000 --> 00:44:44,800 So I think what's causing this error is I think one of the fields somewhere in here 501 00:44:44,800 --> 00:44:52,360 write some product description or strain name I think begins with an equal sign in Excel 502 00:44:52,360 --> 00:44:54,340 doesn't like them. 503 00:44:54,340 --> 00:45:00,480 So once again if you can find you know what strain or product name that is or implement 504 00:45:00,480 --> 00:45:04,320 a fix that's a necessary fix. 505 00:45:04,320 --> 00:45:12,560 And but but but but but long story short the this is basically if you just want the raw 506 00:45:12,560 --> 00:45:16,140 lab results we have those. 507 00:45:16,140 --> 00:45:23,240 And so as I said they've got an inventory ID and then you can get all the cool things 508 00:45:23,240 --> 00:45:27,960 about these particular samples right there. 509 00:45:27,960 --> 00:45:44,040 The Delta 9 THC CBD CBD the moisture content once again completely under analyzed field 510 00:45:44,040 --> 00:45:49,720 here when from my perspective I don't think the labs. 511 00:45:49,720 --> 00:45:53,320 Hey well maybe things have kind of changed. 512 00:45:53,320 --> 00:45:55,720 Actually it's different in different states. 513 00:45:55,720 --> 00:46:04,440 Some states people report dry weight where they basically it's actually an inflation 514 00:46:04,440 --> 00:46:09,960 so they actually inflate the number kind of relative to the moisture content. 515 00:46:09,960 --> 00:46:13,000 So in some states it really really matters. 516 00:46:13,000 --> 00:46:22,880 Washington State I'm fairly certain reports wet weight numbers and maybe somebody trying 517 00:46:22,880 --> 00:46:25,160 to join your quick. 518 00:46:25,160 --> 00:46:30,200 Hopefully hopefully some psychosis. 519 00:46:30,200 --> 00:46:44,560 Long story short there may or may not be much that you can actually get like knowledge wise 520 00:46:44,560 --> 00:46:47,080 out of moisture and water activity. 521 00:46:47,080 --> 00:46:51,200 But any data is better than no data. 522 00:46:51,200 --> 00:46:58,560 And as I pointed out hopefully to people's attention is you know it's worthwhile looking 523 00:46:58,560 --> 00:47:02,480 at pesticides and residual solvents. 524 00:47:02,480 --> 00:47:08,400 There's actually almost I mean there may be some but I want to say there's either none 525 00:47:08,400 --> 00:47:13,320 or almost no heavy metal detections. 526 00:47:13,320 --> 00:47:17,680 Let's actually confirm real quick. 527 00:47:17,680 --> 00:47:25,680 So let's just look at heavy metals and I want to say there's actually I take that back. 528 00:47:25,680 --> 00:47:27,960 It looks like. 529 00:47:27,960 --> 00:47:30,160 Right. 530 00:47:30,160 --> 00:47:41,560 So out of what is there out of like seventy five thousand lab tests there were about one 531 00:47:41,560 --> 00:47:45,960 hundred that that heavy metal was detected in. 532 00:47:45,960 --> 00:47:49,560 So that's that's pretty serious. 533 00:47:49,560 --> 00:48:02,280 You know once again these are outliers but sometimes it's interesting to study the outliers 534 00:48:02,280 --> 00:48:03,280 right. 535 00:48:03,280 --> 00:48:05,060 And there's only so much time in the day. 536 00:48:05,060 --> 00:48:10,720 So maybe this is a cool pet project for one of you data scientists is you know why don't 537 00:48:10,720 --> 00:48:13,840 you look at. 538 00:48:13,840 --> 00:48:17,480 So here we could actually do the same thing and we can get more information so we could 539 00:48:17,480 --> 00:48:23,760 actually see you know is there any sort of you know pattern right. 540 00:48:23,760 --> 00:48:27,080 That's what humans are all about right is looking for patterns. 541 00:48:27,080 --> 00:48:33,940 So is there any pattern to to the samples that failed for heavy metals. 542 00:48:33,940 --> 00:48:40,720 So once again I wouldn't be surprised if there if there's some interesting. 543 00:48:40,720 --> 00:48:53,320 Looks like they're yeah we are thinking that they may be some sort of concentrates because. 544 00:48:53,320 --> 00:48:57,200 You can get sort of all sorts of background contamination right. 545 00:48:57,200 --> 00:49:03,720 So if you're like like say you're making edibles in your you know you're grinding it up or 546 00:49:03,720 --> 00:49:08,880 something like maybe you're using a metal grinder or something like that and you're 547 00:49:08,880 --> 00:49:11,160 getting but they're getting hits. 548 00:49:11,160 --> 00:49:14,280 So maybe who knows right. 549 00:49:14,280 --> 00:49:19,800 So I don't want to actually make any I don't want to make any I guess they'd be hypotheses 550 00:49:19,800 --> 00:49:23,440 but I don't think this stage they're more of conjectures. 551 00:49:23,440 --> 00:49:28,480 So I don't really want to just take any stabs at this but. 552 00:49:28,480 --> 00:49:33,040 But I'm just kind of pointing this out is there's really cool. 553 00:49:33,040 --> 00:49:43,480 Data still to be had by adding the lab results here. 554 00:49:43,480 --> 00:49:47,200 Final thing is. 555 00:49:47,200 --> 00:49:50,440 You know what what's the point of all of this. 556 00:49:50,440 --> 00:50:00,120 Well at the end of the day the point is calculating statistics and once again. 557 00:50:00,120 --> 00:50:06,120 And or neglect or. 558 00:50:06,120 --> 00:50:10,880 Don't take into consideration the statistics I'm about to show you because as I pointed 559 00:50:10,880 --> 00:50:17,880 out I don't think I've matched all of the licensees to sales. 560 00:50:17,880 --> 00:50:24,400 So this is only the total sales for the people I have identified. 561 00:50:24,400 --> 00:50:30,800 And so this is going to explain why the numbers look really really low. 562 00:50:30,800 --> 00:50:34,480 But. 563 00:50:34,480 --> 00:50:40,760 But we can find out the you know we can start to basically these are basically. 564 00:50:40,760 --> 00:50:47,600 So I may have oversold the statistics that I was telling you about today but as I said. 565 00:50:47,600 --> 00:50:48,600 Here's the code. 566 00:50:48,600 --> 00:50:55,720 I'll have this posted to GitHub today for you to peruse through and see if there's any 567 00:50:55,720 --> 00:50:58,840 improvements we can have here. 568 00:50:58,840 --> 00:51:03,600 But the whole point is. 569 00:51:03,600 --> 00:51:10,440 Can't we now take this a step further than just look at aggregate sales because I mean 570 00:51:10,440 --> 00:51:15,960 how meaningful is that like say there's a hundred million. 571 00:51:15,960 --> 00:51:20,560 Per month or 50 million per month. 572 00:51:20,560 --> 00:51:23,520 Different states it's in it's in that ballpark. 573 00:51:23,520 --> 00:51:28,160 I mean what what does that even mean you know and that's where we've started to say like 574 00:51:28,160 --> 00:51:34,360 oh well maybe sales per retailer is more relevant. 575 00:51:34,360 --> 00:51:41,960 But then we ran into the dish the distribution problem where you know not all retailers are 576 00:51:41,960 --> 00:51:50,960 are operating the same right there and they're not all located in the same place. 577 00:51:50,960 --> 00:51:57,080 The classic hoteling problem so you know the long story short you can start doing fun things 578 00:51:57,080 --> 00:52:00,920 like you know. 579 00:52:00,920 --> 00:52:06,680 Calculating how many products are sold you can look at how many products are sold by 580 00:52:06,680 --> 00:52:09,880 retailer and here's a hint. 581 00:52:09,880 --> 00:52:20,320 In the past we we did run a regression where we did find a positive correlation between 582 00:52:20,320 --> 00:52:27,640 the amount of products I want to say sold but it could have been stopped but I want 583 00:52:27,640 --> 00:52:28,640 to say sold. 584 00:52:28,640 --> 00:52:33,200 So just the total amount of different products you were selling in your revenue. 585 00:52:33,200 --> 00:52:39,360 And this kind of goes in line with something someone told me once about a retailer and 586 00:52:39,360 --> 00:52:45,080 he says you just want to have your product your shelves filled with a bunch of different 587 00:52:45,080 --> 00:52:48,800 types of products and you want to keep cycling those. 588 00:52:48,800 --> 00:52:55,280 And once again this is just here say this is just what this this person recommends. 589 00:52:55,280 --> 00:53:01,000 But he says you know you want to keep your customers seeing new products you know they're 590 00:53:01,000 --> 00:53:04,240 going to get bored if they only see the same things on the shelves. 591 00:53:04,240 --> 00:53:11,760 Once again he may be right he may be wrong but it's kind of fun right cannabis. 592 00:53:11,760 --> 00:53:17,400 They're always coming up with new strain names so maybe that maybe people maybe they find 593 00:53:17,400 --> 00:53:24,520 that fun maybe they're looking for the new fun strain name this this month. 594 00:53:24,520 --> 00:53:29,440 And but but this is basically where I. 595 00:53:29,440 --> 00:53:38,760 I started to realize my numbers were completely wrong because I'm only I if you only account 596 00:53:38,760 --> 00:53:49,360 everything that can be matched back to inventory I can only account for about less than 24 597 00:53:49,360 --> 00:53:52,200 million in sales. 598 00:53:52,200 --> 00:54:00,800 But I want to say once again we've calculated this in the past and so I'm just going to 599 00:54:00,800 --> 00:54:07,200 have to go off of my Bayesian prior aka my best recollection. 600 00:54:07,200 --> 00:54:14,080 But my best recollection is maybe say like I said I think sales in Washington maybe around 601 00:54:14,080 --> 00:54:17,040 50 to 100 million a month. 602 00:54:17,040 --> 00:54:23,080 So I think we've only accounted for a small fraction not a small but definitely only a 603 00:54:23,080 --> 00:54:26,320 fraction of sales. 604 00:54:26,320 --> 00:54:29,560 But but you know what are we after here. 605 00:54:29,560 --> 00:54:36,880 You know we can start doing things like oh you know we can try to calculate the effective 606 00:54:36,880 --> 00:54:37,960 tax rate. 607 00:54:37,960 --> 00:54:38,960 So you know what. 608 00:54:38,960 --> 00:54:45,600 What are after you add up all the taxes you know how much your people are paying. 609 00:54:45,600 --> 00:54:52,720 And here is actually a statistic that could maybe be extrapolated out. 610 00:54:52,720 --> 00:55:00,080 But once again we've got a non random sample out of our population of sales or non random 611 00:55:00,080 --> 00:55:06,080 sample is the sales items I've been able to curate. 612 00:55:06,080 --> 00:55:10,680 But this is a statistic that I found interesting. 613 00:55:10,680 --> 00:55:11,680 Right. 614 00:55:11,680 --> 00:55:17,400 So we're just going to look at the proportion of products sold. 615 00:55:17,400 --> 00:55:25,240 I mean so this this statistic may or may not change as we we take into consideration all 616 00:55:25,240 --> 00:55:30,120 of the other sales items. 617 00:55:30,120 --> 00:55:40,800 But long story short is you know by Bayesian prior was that flower was around 60 percent 618 00:55:40,800 --> 00:55:43,280 of all sales. 619 00:55:43,280 --> 00:55:51,600 But it's looking like either things have changed in the Washington market. 620 00:55:51,600 --> 00:56:00,520 These statistics are incorrect or potentially people are purchasing more of the some of 621 00:56:00,520 --> 00:56:06,720 these other types of products like the liquid edibles and solid edibles. 622 00:56:06,720 --> 00:56:13,000 We'll have to double check but you know those are each approaching. 623 00:56:13,000 --> 00:56:17,240 They're still shy each of about 10 percent of sales. 624 00:56:17,240 --> 00:56:21,680 But that may be up. 625 00:56:21,680 --> 00:56:22,680 That may be up. 626 00:56:22,680 --> 00:56:29,800 I want to say in the past maybe liquid and edibles combined were around 10 percent. 627 00:56:29,800 --> 00:56:38,920 And so you know if it increases from around 10 percent to around 15 percent of total sales 628 00:56:38,920 --> 00:56:42,840 those sales have to come from somewhere. 629 00:56:42,840 --> 00:56:45,800 So who knows. 630 00:56:45,800 --> 00:56:50,200 Maybe people are doing some sort of substitution there. 631 00:56:50,200 --> 00:57:03,360 But anywho I've kind of lost my confidence in my own statistics for better or for worse. 632 00:57:03,360 --> 00:57:08,360 So to make of that what you will. 633 00:57:08,360 --> 00:57:11,800 But I don't know what what do you make of it. 634 00:57:11,800 --> 00:57:19,720 Do you do you think we need to to refine some of this data curation a little bit. 635 00:57:19,720 --> 00:57:26,000 Use some of the other low hanging fruit or you know work on this cherry picker. 636 00:57:26,000 --> 00:57:29,600 You know what are your thoughts. 637 00:57:29,600 --> 00:57:31,480 Please. 638 00:57:31,480 --> 00:57:37,840 Data cleaning is is great. 639 00:57:37,840 --> 00:57:44,320 And then also to you know if we started putting in tests little tests you know to to make 640 00:57:44,320 --> 00:57:49,760 sure you know if there are empty values you know check on that. 641 00:57:49,760 --> 00:57:52,440 And I don't know. 642 00:57:52,440 --> 00:57:55,640 Awesome job Keegan. 643 00:57:55,640 --> 00:57:58,920 Well it will be awesome before long. 644 00:57:58,920 --> 00:58:05,240 So as I said if we are somehow able to match everything up. 645 00:58:05,240 --> 00:58:12,920 So basically if we're able to match every single sales item to a piece of inventory 646 00:58:12,920 --> 00:58:20,120 and then we should be able to match every thing that was sold to a lab result via the 647 00:58:20,120 --> 00:58:27,680 inventory then we'll have a complete picture of the Washington state market. 648 00:58:27,680 --> 00:58:35,280 And this is kind of what we've been chasing and I feel like we're so so so close. 649 00:58:35,280 --> 00:58:42,160 And I think this would really just I think this would finally be like the big breakthrough 650 00:58:42,160 --> 00:58:48,600 everybody's been looking for because you could finally start to answer some of these really 651 00:58:48,600 --> 00:58:52,440 interesting questions. 652 00:58:52,440 --> 00:58:57,800 For example is you know just to play a devil's advocate. 653 00:58:57,800 --> 00:59:02,600 People are saying oh there's a problem with high THC cannabis. 654 00:59:02,600 --> 00:59:08,840 Well you could then look at the zip codes where there's a lot of high THC cannabis being 655 00:59:08,840 --> 00:59:12,640 sold you know see if that correlates with anything negative. 656 00:59:12,640 --> 00:59:14,960 So that's one thing you could do. 657 00:59:14,960 --> 00:59:22,560 As I said I'd personally would just love to see how much THC matters. 658 00:59:22,560 --> 00:59:24,360 If it even matters at all. 659 00:59:24,360 --> 00:59:25,360 Right. 660 00:59:25,360 --> 00:59:29,320 For example for edibles it may not matter at all. 661 00:59:29,320 --> 00:59:32,440 Or it may. 662 00:59:32,440 --> 00:59:38,440 So I'm you. 663 00:59:38,440 --> 00:59:39,440 I don't know. 664 00:59:39,440 --> 00:59:44,080 I just feel like I just feel like the sky's the limit. 665 00:59:44,080 --> 00:59:48,200 Once as Candice points out the data is you know finally clean. 666 00:59:48,200 --> 00:59:55,840 So I don't think it's I don't think it's quite at that stage but it's I almost am calling 667 00:59:55,840 --> 00:59:59,760 on you for help because it's almost as far as I can take it. 668 00:59:59,760 --> 01:00:06,160 So I really need what some one of you to have like almost a brilliant insight here in you 669 01:00:06,160 --> 01:00:12,640 know how can we map like match this inventory together. 670 01:00:12,640 --> 01:00:19,400 So I strongly encourage you to help on this because as I said I've got a sneaking suspicion 671 01:00:19,400 --> 01:00:28,880 that there's either like some sort of like parent child inventory relationship or maybe 672 01:00:28,880 --> 01:00:32,240 the product data is the missing piece. 673 01:00:32,240 --> 01:00:38,000 But there's just I feel like there's just this one final piece that'll bring everything 674 01:00:38,000 --> 01:00:41,560 together. 675 01:00:41,560 --> 01:00:43,900 So I'm sharing this with you. 676 01:00:43,900 --> 01:00:49,480 So I hope you're excited about this and the potential. 677 01:00:49,480 --> 01:00:51,080 That's awesome Keegan. 678 01:00:51,080 --> 01:00:58,720 So for everybody to can litics on that hub there is a cannabis data science that you 679 01:00:58,720 --> 01:01:05,840 can you know clone and also to there is a high can litics dot com. 680 01:01:05,840 --> 01:01:11,320 There's a Slack chance and because that's what I'm gathering right Keegan that you know 681 01:01:11,320 --> 01:01:17,080 we're looking for other eyeballs and you know certainly to you know Keegan if you're busy 682 01:01:17,080 --> 01:01:22,840 to when people need some help with setting up environments or you know running code to 683 01:01:22,840 --> 01:01:28,680 can litics you know if you want you can toss it over to me and if anybody appreciates my 684 01:01:28,680 --> 01:01:35,040 help you don't have to pay me just donate to cannabis data science and can litics. 685 01:01:35,040 --> 01:01:36,780 So it's so good. 686 01:01:36,780 --> 01:01:41,560 Thank you Keegan. 687 01:01:41,560 --> 01:01:43,080 I absolutely love it. 688 01:01:43,080 --> 01:01:50,240 Thanks for the shout out Candace and I'm a little behind on committing the code but today 689 01:01:50,240 --> 01:01:54,120 is the day right. 690 01:01:54,120 --> 01:01:58,920 You know the Python philosophy you know now is better than ever and you know today is 691 01:01:58,920 --> 01:02:00,300 finally the now. 692 01:02:00,300 --> 01:02:03,800 So I'm going to get all the code posted today. 693 01:02:03,800 --> 01:02:10,660 As I mentioned it's going to be messy and there's an invitation to the Slack channel 694 01:02:10,660 --> 01:02:15,440 so you're welcome to join us there and then as I said I'm going to be doing a better job 695 01:02:15,440 --> 01:02:23,600 at you know filling out issues on GitHub making sure it's a bit more you know generally approachable 696 01:02:23,600 --> 01:02:29,560 because there's some real cool data curation scripts here because the idea is you can take 697 01:02:29,560 --> 01:02:36,160 that say public records request from the Washington State LCB and then you can sort of run these 698 01:02:36,160 --> 01:02:38,800 duration scripts one by one. 699 01:02:38,800 --> 01:02:46,920 So the lab result one after a lot of fine tuning it used to take two hours to run and 700 01:02:46,920 --> 01:02:49,960 we've now have it down to eight minutes. 701 01:02:49,960 --> 01:02:57,560 So you can now curate that data set of lab results I was showing you in about eight minutes 702 01:02:57,560 --> 01:03:03,200 which there's a lot of a lot a lot you can poke around that there. 703 01:03:03,200 --> 01:03:06,600 Like I was just showing you with the heavy metals. 704 01:03:06,600 --> 01:03:12,120 I mean that's a whole research question of its own that people are super interested in. 705 01:03:12,120 --> 01:03:18,520 People are I mean different states are requiring heavy metal testing and I mean a big question 706 01:03:18,520 --> 01:03:21,220 is is you know is it worth it right. 707 01:03:21,220 --> 01:03:28,440 It's a really really expensive test and so if there's if all of the detections are just 708 01:03:28,440 --> 01:03:36,520 oddballs out then is it worth it or maybe they snagged some really hard finds. 709 01:03:36,520 --> 01:03:41,280 So so that's a really interesting research question. 710 01:03:41,280 --> 01:03:46,880 And then the sales items it takes about two hours. 711 01:03:46,880 --> 01:03:52,440 It takes about once again maybe on a nice rig like Candice's you may make short work 712 01:03:52,440 --> 01:04:05,000 of this but for me it takes about two to four hours per sales data file and there are I 713 01:04:05,000 --> 01:04:09,800 have GPU so you know let me know too if I get kind of a little lost in my Lang Chang 714 01:04:09,800 --> 01:04:16,480 and trying out the different free open source LLMs you know help bring me back down to earth 715 01:04:16,480 --> 01:04:21,920 and just put in the put what you want run you know want me to run and download and and 716 01:04:21,920 --> 01:04:25,800 you know I should be able to you know get you results you know within a hopefully a 717 01:04:25,800 --> 01:04:29,920 quick you know GPU amount of time. 718 01:04:29,920 --> 01:04:38,800 It's phenomenal and we're kind of going under this sort of the blockchain philosophy here 719 01:04:38,800 --> 01:04:46,360 in that we could ideally distribute this in that if we all follow the same curation principles 720 01:04:46,360 --> 01:04:50,400 you can almost say check if one was curated the same way. 721 01:04:50,400 --> 01:04:55,920 So we basically just create a hash of the whole data set and if we each get the same 722 01:04:55,920 --> 01:04:58,580 hash then we know it's the same. 723 01:04:58,580 --> 01:05:08,680 So maybe I curate 101 and 102 and then Candice calculates you know 80 through 101 and then 724 01:05:08,680 --> 01:05:15,320 you know I can check that 101 was right so yes I you know you can kind of verify that 725 01:05:15,320 --> 01:05:22,640 and is this process was right and then yeah you know Candice can do a lot of the work. 726 01:05:22,640 --> 01:05:33,320 So I just had my computer running for the past week or two so like I said it took you 727 01:05:33,320 --> 01:05:38,960 know 200 to 400 hours of computing but actually I had like I was running it sometimes like 728 01:05:38,960 --> 01:05:50,200 in five parallel processes sometimes so maybe 50 to 60 hours with some breaks there to let 729 01:05:50,200 --> 01:05:52,120 my computer cool down. 730 01:05:52,120 --> 01:06:00,480 So it takes a long time to curate the data and I made a mistake so we're going to have 731 01:06:00,480 --> 01:06:06,280 to fix the mistake and then you know recreate all of them but that's okay like I said it 732 01:06:06,280 --> 01:06:12,320 was a good learning process to just see how much of the data we actually can match how 733 01:06:12,320 --> 01:06:20,480 much is still to be matched and then you know I'll get all this in your hands as much as 734 01:06:20,480 --> 01:06:29,160 I can and yeah all hands on deck you know put your favorite data science tools to use 735 01:06:29,160 --> 01:06:34,040 and let's uncover some knowledge. 736 01:06:34,040 --> 01:06:41,760 So want to go ahead and thank you all for coming I can't thank you enough it's your 737 01:06:41,760 --> 01:06:47,400 eyes your ears your attention that's really moving things forward you're the people who 738 01:06:47,400 --> 01:06:50,760 are making the cannabis data science meetup happen. 739 01:06:50,760 --> 01:06:57,880 I hope we're advancing the cannabis space and cannabis science if even only a molecule 740 01:06:57,880 --> 01:07:18,160 at a time I'm happy with that so I hope you're happy and had a little fun too.