1 00:00:00,000 --> 00:00:11,120 Welcome to the Canvas Data Science Meetup Group. 2 00:00:11,120 --> 00:00:13,280 You're in for a treat today. 3 00:00:13,280 --> 00:00:16,400 It could easily be the best meetup to date. 4 00:00:16,400 --> 00:00:18,760 So we'll cover a lot of ground today. 5 00:00:18,760 --> 00:00:22,280 So fasten your seatbelts because it's going to be a good one. 6 00:00:22,280 --> 00:00:24,920 So before we kick it off though, 7 00:00:24,920 --> 00:00:27,440 Kay and Travis. 8 00:00:27,440 --> 00:00:32,400 So Kay just told me that you're interested in data science, 9 00:00:32,400 --> 00:00:35,600 particularly R. Travis, 10 00:00:35,600 --> 00:00:38,280 would you be interested in introducing yourself and 11 00:00:38,280 --> 00:00:42,040 what questions about cannabis and the cannabis industry 12 00:00:42,040 --> 00:00:46,000 you may be interested in answering with data science? 13 00:00:47,040 --> 00:00:50,400 Hi. I'm just trying to get into the industry 14 00:00:50,400 --> 00:00:52,360 and getting to be in a data scientist. 15 00:00:52,360 --> 00:00:55,960 I just want to learn as much as I can right now. 16 00:00:55,960 --> 00:00:59,560 Awesome. You're in the right place. 17 00:00:59,560 --> 00:01:02,600 Well, without further ado, 18 00:01:02,600 --> 00:01:08,120 I typically end up just trying to present some of my latest work, 19 00:01:08,120 --> 00:01:11,960 and then just would like to pause frequently throughout 20 00:01:11,960 --> 00:01:14,240 to answer any questions you may have, 21 00:01:14,240 --> 00:01:16,880 and hopefully it spark a discussion. 22 00:01:16,880 --> 00:01:21,600 So I'll go ahead and share my screen. 23 00:01:21,600 --> 00:01:24,280 Awesome. So welcome to the meetup. 24 00:01:24,280 --> 00:01:30,880 You can find all of the source code on GitHub here. 25 00:01:30,880 --> 00:01:35,640 Here's a good regular Varus Graham who will be joining momentarily. 26 00:01:37,640 --> 00:01:40,280 Welcome Graham. It's good to see you today. 27 00:01:40,280 --> 00:01:42,280 Hey. 28 00:01:42,280 --> 00:01:46,880 Well, we've got a lot of questions. 29 00:01:46,880 --> 00:01:53,000 Well, we've got a lot of questions. 30 00:02:16,880 --> 00:02:21,880 Awesome. So just making sure everybody's on board. 31 00:02:21,880 --> 00:02:24,480 So here's the source code. 32 00:02:24,480 --> 00:02:30,280 And essentially thought we could start with a question 33 00:02:30,280 --> 00:02:33,960 and spend most of the day doing data science today. 34 00:02:33,960 --> 00:02:39,520 So Graham will be getting to producers and processors 35 00:02:39,520 --> 00:02:42,040 in next week's meetup. 36 00:02:42,040 --> 00:02:47,120 For today, I thought we could answer the question, 37 00:02:47,120 --> 00:02:50,560 how can you run a profitable lab? 38 00:02:50,560 --> 00:02:56,120 So we've got this rich data set of lab analyses in Washington state. 39 00:02:56,120 --> 00:03:01,920 And so I was thinking, is there any way we can use this data set to help others 40 00:03:01,920 --> 00:03:04,600 either in Washington state or across the country, 41 00:03:04,600 --> 00:03:08,320 think about how they can run a profitable lab? 42 00:03:08,320 --> 00:03:12,160 So always helps to start with a good question. 43 00:03:12,160 --> 00:03:14,160 So this is our question of the day. 44 00:03:18,160 --> 00:03:27,000 So without further ado, the first step is really just to read in the data. 45 00:03:27,000 --> 00:03:34,160 So I've already read in the data just to save us a minute or so. 46 00:03:34,160 --> 00:03:43,760 But it's simple enough to read in the packages. 47 00:03:43,760 --> 00:03:51,240 And here I just defined a few useful helper functions. 48 00:03:51,240 --> 00:03:55,960 And then to save ourselves some bandwidth, 49 00:03:55,960 --> 00:03:59,080 I'm just reading in one variable. 50 00:03:59,080 --> 00:04:06,880 So we've got this awesome data set of lab results in Washington state. 51 00:04:06,880 --> 00:04:11,960 And so you can find the link here above. 52 00:04:11,960 --> 00:04:16,360 And just to show you where I've stored the data, 53 00:04:16,360 --> 00:04:21,640 so in the cannabis data science repository, 54 00:04:21,640 --> 00:04:24,840 you can create a folder at your root. 55 00:04:24,840 --> 00:04:29,040 Called data sets, dot data sets. 56 00:04:29,040 --> 00:04:36,240 And here I've stored these three sets of lab results, 57 00:04:36,240 --> 00:04:38,440 as well as the licensees data. 58 00:04:38,440 --> 00:04:46,800 So as you can see, we're working with about 2.2 or so gigabytes of data. 59 00:04:46,800 --> 00:04:48,320 So a nice good chunk. 60 00:04:48,320 --> 00:04:51,920 And there's so many data points here. 61 00:04:51,920 --> 00:04:57,680 We've got all these solvents and all the pesticide data is blank. 62 00:04:57,680 --> 00:05:00,240 So you can't use it anyways. 63 00:05:00,240 --> 00:05:02,680 You do have the cannabinoid data. 64 00:05:02,680 --> 00:05:04,720 So rich data set here. 65 00:05:04,720 --> 00:05:10,760 But I'll show you what you can do with just this one data point, the global ID. 66 00:05:10,760 --> 00:05:14,520 So here I've got the 67 00:05:14,520 --> 00:05:19,720 global ID, which is this one data point, the global ID. 68 00:05:19,720 --> 00:05:26,840 So we'll do a lot of legwork today with just this one variable. 69 00:05:26,840 --> 00:05:32,120 So basically, I've already read in the data here. 70 00:05:32,120 --> 00:05:39,400 This doesn't take very long, but maybe 30 seconds or so to read in this data set. 71 00:05:39,400 --> 00:05:42,640 And then we clean it. 72 00:05:42,640 --> 00:05:50,320 So we can identify all of the labs in Washington state. 73 00:05:50,320 --> 00:05:53,760 And this is sort of supplementary. 74 00:05:53,760 --> 00:05:56,560 We don't really need this step. 75 00:05:56,560 --> 00:06:02,520 But you can also find out information about the various labs 76 00:06:02,520 --> 00:06:07,360 by running it, by reading in the licensees data. 77 00:06:07,360 --> 00:06:15,440 That way we can, for example, look at specific factors. 78 00:06:15,440 --> 00:06:20,760 And then here, just add day, month, year variables. 79 00:06:20,760 --> 00:06:25,440 So a lot of cleaning and organizing to do. 80 00:06:25,440 --> 00:06:32,120 But we can get onto the data because that's what's more interesting to look at. 81 00:06:32,120 --> 00:06:36,040 So we can calculate our first data point. 82 00:06:36,040 --> 00:06:39,600 So just using the total number of samples, 83 00:06:39,600 --> 00:06:47,160 we can see which lab tested each sample based on the global ID. 84 00:06:47,160 --> 00:07:00,680 So for example, you can see that the global ID contains the lab's identifier. 85 00:07:00,680 --> 00:07:06,680 So these five lab results were tested by lab 10. 86 00:07:06,680 --> 00:07:15,960 So we can simply just do a count of how many samples all the various labs were testing. 87 00:07:15,960 --> 00:07:21,160 And since we know the total number of samples tested, 88 00:07:21,160 --> 00:07:25,120 we can see the proportion that each lab is testing. 89 00:07:25,120 --> 00:07:28,320 And so I just call that the market share here, 90 00:07:28,320 --> 00:07:33,280 so that the number of samples that lab tested versus the total. 91 00:07:33,280 --> 00:07:40,920 So without further ado, let's go ahead and print out our first statistics for the day. 92 00:07:40,920 --> 00:07:44,120 And here we have them. 93 00:07:44,120 --> 00:07:56,280 So these are a handful of laboratories that have or are operating in Washington state. 94 00:07:56,280 --> 00:08:04,240 And I just printed out the city, so that way you can just get a geographical perspective 95 00:08:04,240 --> 00:08:06,800 of where the labs may be. 96 00:08:06,800 --> 00:08:12,600 And just to sort these, I'm just going to essentially sort them 97 00:08:12,600 --> 00:08:16,040 by their total number of samples. 98 00:08:16,040 --> 00:08:21,240 That way, we can see what's the market share by laboratory 99 00:08:21,240 --> 00:08:24,600 and where in the state are they located. 100 00:08:24,600 --> 00:08:31,680 So you can see, so Poles bow is what you would call Western Washington. 101 00:08:31,680 --> 00:08:34,840 So it's interesting the breakdown of Washington. 102 00:08:34,840 --> 00:08:40,320 It's really a different climate in Eastern versus Western Washington. 103 00:08:40,320 --> 00:08:46,720 And so I always thought it was interesting about where producers may be located. 104 00:08:46,720 --> 00:08:52,200 And I thought, OK, maybe laboratories can locate anywhere. 105 00:08:52,200 --> 00:08:59,480 But of course, people prefer a shorter distance for sending their samples. 106 00:08:59,480 --> 00:09:06,360 So there is a laboratory in Poles bow, and they have the highest market share. 107 00:09:06,360 --> 00:09:16,080 Bellingham is in Northeastern Washington, so they're capturing that segment of the market. 108 00:09:16,080 --> 00:09:19,360 Redmond is near Seattle. 109 00:09:19,360 --> 00:09:25,600 And then, of course, Olympia is on the board and in a couple of places. 110 00:09:25,600 --> 00:09:35,440 So just a statistic, there's the market share and there's the cities they're in. 111 00:09:35,440 --> 00:09:42,280 So we always like to try to do some plots here. 112 00:09:42,280 --> 00:09:50,640 So here is simply a plot of samples by day. 113 00:09:50,640 --> 00:09:53,800 Just checking, am I still coming in OK for everyone? 114 00:09:58,320 --> 00:10:00,680 Just making sure we're still on the view. 115 00:10:00,680 --> 00:10:01,480 Awesome. 116 00:10:01,480 --> 00:10:10,320 So like I said, we've got a lot of ground to cover, so sorry if I talk a lot. 117 00:10:10,320 --> 00:10:19,680 So here we have samples by day, by lab. 118 00:10:19,680 --> 00:10:24,400 Quite noisy, so a little too noisy. 119 00:10:24,400 --> 00:10:29,720 So here I'm just doing the same plot, but by month. 120 00:10:29,720 --> 00:10:35,040 So we're just aggregating all the samples by month. 121 00:10:35,040 --> 00:10:46,880 So here you can see we've got labs operating all over the board with some testing, 122 00:10:46,880 --> 00:10:57,120 you know, almost some months more, but, you know, some up to around 3,000 samples a month. 123 00:10:57,120 --> 00:10:59,640 And then we can plot by year. 124 00:10:59,640 --> 00:11:05,440 Just to reduce a bit of the noise even further. 125 00:11:05,440 --> 00:11:11,960 And so here you see samples tested by lab by year. 126 00:11:11,960 --> 00:11:21,440 Just really just sort of adding statistics upon statistics here. 127 00:11:21,440 --> 00:11:24,920 So, right, we started with, OK, what's the total number of samples? 128 00:11:24,920 --> 00:11:30,720 Tested by lab, now we want to try to break that out by year. 129 00:11:30,720 --> 00:11:41,880 So doing that, and, you know, we're seeing, right, we've got some labs testing around a little less than 30,000 samples per year. 130 00:11:41,880 --> 00:11:49,680 Looks like the blue lab, light blue, is, you know, on the trajectory towards 30,000. 131 00:11:49,680 --> 00:11:57,680 Then you've got a handful of labs testing around 15,000 or so samples a year. 132 00:11:57,680 --> 00:12:10,840 And then we've got another handful that are testing below 10,000 per year with this gray on a slight positive trajectory. 133 00:12:10,840 --> 00:12:14,280 So nothing too interesting yet. 134 00:12:14,280 --> 00:12:20,080 Just, you know, these are just the samples being tested. 135 00:12:20,080 --> 00:12:26,880 So where things get interesting in data science is when you can start adding supplementary statistics. 136 00:12:26,880 --> 00:12:33,880 So for those of you that are ambitious, I'll let you calculate these extra statistics on your own. 137 00:12:33,880 --> 00:12:39,880 So this is you can start to break down samples tested by the various analyses. 138 00:12:39,880 --> 00:12:44,880 That they may be doing, or you could even look at the failure rate. 139 00:12:44,880 --> 00:12:55,880 So these are quality assurance or quality control tests that samples need to go through to make it onto the shelves. 140 00:12:55,880 --> 00:13:03,480 So it would be interesting from a producer processor point of view to know what the failure rate may be. 141 00:13:03,480 --> 00:13:12,480 So that may be for a bit more for next week. So next week, we'll look at the profitability of producers and processors. 142 00:13:12,480 --> 00:13:17,480 So that'll come into play then. 143 00:13:17,480 --> 00:13:28,080 So we'll start supplementing this with data here in a second, but we can start using some of our statistical techniques. 144 00:13:28,080 --> 00:13:34,680 So, for example, we've done forecasting with ARIMA models. 145 00:13:34,680 --> 00:13:41,880 So here we can essentially forecast the number of samples tested by lab. 146 00:13:41,880 --> 00:13:45,680 And I was thinking. 147 00:13:45,680 --> 00:13:53,080 We've been talking about the long term here, so I figured let's just go ahead and do a long term forecast. 148 00:13:53,080 --> 00:13:58,280 So we'll forecast today through the next five years. 149 00:13:58,280 --> 00:14:02,280 So through 2027. 150 00:14:02,280 --> 00:14:09,280 Cecilia joining. 151 00:14:09,280 --> 00:14:17,480 So I'll go ahead and run this forecast since this forecasting model takes a second to run. 152 00:14:17,480 --> 00:14:22,080 And I'll talk about some of the next data that we'll collect. 153 00:14:22,080 --> 00:14:26,080 So where am I going with all of this? 154 00:14:26,080 --> 00:14:33,880 Right. We're getting the samples tested per year, but I'm essentially interested in, OK, how do you run a profitable lab? 155 00:14:33,880 --> 00:14:39,880 Just because you're testing a lot of samples doesn't necessarily mean you're profitable or not. 156 00:14:39,880 --> 00:14:46,080 So I was thinking we could supplement this with some of the data that we've collected. 157 00:14:46,080 --> 00:14:52,880 And we could supplement this with the price data in Washington state. 158 00:14:52,880 --> 00:15:01,480 So essentially what I've done is I've just collected a big data set here of laboratories. 159 00:15:01,480 --> 00:15:15,280 And I just went to the laboratory's website, went to their pricing page and. 160 00:15:15,280 --> 00:15:18,280 And grabbed. 161 00:15:18,280 --> 00:15:26,280 The price of an eye 502 panel, so that would be your quality assurance testing panel. 162 00:15:26,280 --> 00:15:34,680 Some of the labs have different prices, depending on if it's a flower or if it's a concentrate or what have you. 163 00:15:34,680 --> 00:15:38,080 So I just chose the simplest number. 164 00:15:38,080 --> 00:15:41,280 So that's sort of an imperfection. 165 00:15:41,280 --> 00:15:48,080 So the price may be overstated because, for example. 166 00:15:48,080 --> 00:15:54,480 And the products such as sodas, I believe. 167 00:15:54,480 --> 00:16:01,680 I may walk that back, but certain products don't need as many tests as others, so their prices are cheaper. 168 00:16:01,680 --> 00:16:07,680 So I generally got the price for a flower or concentrate test. 169 00:16:07,680 --> 00:16:11,380 And so I collected all the prices for the various labs. 170 00:16:11,380 --> 00:16:17,680 And since can litics is here to make all this data easily accessible for you. 171 00:16:17,680 --> 00:16:22,680 You can actually access all of this data. 172 00:16:22,680 --> 00:16:37,080 Through an API, so you can ping can litics dot com forward slash API forward slash labs and you can get all of the lab data. 173 00:16:37,080 --> 00:16:43,480 And haven't built in very many queries yet, but we can at least query by the state. 174 00:16:43,480 --> 00:16:50,680 So we'll do that here momentarily, but it looks like our forecasts are finished. 175 00:16:50,680 --> 00:17:02,980 So here's a forecast for the next five years of samples tested by the various laboratories in Washington state. 176 00:17:02,980 --> 00:17:12,880 So as you can see, we're forecasting there is a group of laboratories that are testing around 500 or so samples a month. 177 00:17:12,880 --> 00:17:19,380 Another set testing between a thousand and two thousand samples per month. 178 00:17:19,380 --> 00:17:27,580 And then there's a couple of labs here at the top that are testing around 2500 samples per month. 179 00:17:27,580 --> 00:17:33,880 But keep in mind this these are samples tested, not necessarily profitability. 180 00:17:33,880 --> 00:17:45,880 So we needed to keep in mind price and cost if we want to really get a deep dive on profitability. 181 00:17:45,880 --> 00:17:50,180 So let's read in this place price data. 182 00:17:50,180 --> 00:17:58,580 So simple enough. So here I'm just making a request to can litics API. 183 00:17:58,580 --> 00:18:10,180 Forward slash labs in passing the parameter that I want the state to be Washington since we just want to look at Washington labs here. 184 00:18:10,180 --> 00:18:13,980 So just going to make a quick request. 185 00:18:13,980 --> 00:18:17,280 And it should already have executed. 186 00:18:17,280 --> 00:18:21,780 And there are all of our labs. 187 00:18:21,780 --> 00:18:30,880 And in particular, we just want to look at the prices that the labs are charging for their panels. 188 00:18:30,880 --> 00:18:32,780 And so. 189 00:18:32,780 --> 00:18:36,180 Here you see, you know, we've got a range. 190 00:18:36,180 --> 00:18:40,380 I couldn't find prices for all the labs. That's OK. 191 00:18:40,380 --> 00:18:44,880 We'll just look at the labs where we do have price data. 192 00:18:44,880 --> 00:18:49,380 As you can see, we've got a range here, 80 to 120. 193 00:18:49,380 --> 00:18:54,880 It looks like around 100 is around the average. 194 00:18:54,880 --> 00:19:03,380 Not certain if we can just take a quick average here. 195 00:19:03,380 --> 00:19:06,880 Yeah, so the average is is 99. 196 00:19:06,880 --> 00:19:26,180 So this is a lot lower of a price than you'll see in many other states because Washington state doesn't mandate yet pesticide or heavy metal testing, whereas a lot of the other states do. 197 00:19:26,180 --> 00:19:29,980 And I'll explain further on. 198 00:19:29,980 --> 00:19:36,680 As to a reason why, but it's really hard to change the rules once you've implemented them. 199 00:19:36,680 --> 00:19:44,380 So a lot of states, their initial rules required heavy metal or pesticide testing. 200 00:19:44,380 --> 00:19:55,780 Washington State, one of the earlier labs, I mean, one of the earlier states to start laboratory testing just started with some of the more basic tests. 201 00:19:55,780 --> 00:20:02,980 So cannabinoids that screen for micro toxins and microbiology, foreign matter. 202 00:20:02,980 --> 00:20:07,580 And they do test concentrates for residual solvents. 203 00:20:07,580 --> 00:20:12,480 So not as many tests that are required. 204 00:20:12,480 --> 00:20:18,880 So you're looking at around $100 for a quality assurance test. 205 00:20:18,880 --> 00:20:30,680 Well, assuming that all of the labs keep their prices the same, then what would their revenue be for the next five years? 206 00:20:30,680 --> 00:20:33,180 So this is where things get interesting, right? 207 00:20:33,180 --> 00:20:42,180 So we now get to combine price data with data on the total number of samples tested. 208 00:20:42,180 --> 00:20:45,980 So we can do just that. 209 00:20:45,980 --> 00:20:53,580 And we can have a forecast of monthly. 210 00:20:53,580 --> 00:21:01,280 This should actually read monthly revenue by labs in Washington. 211 00:21:01,280 --> 00:21:05,780 So let's go ahead and make that quick. 212 00:21:05,780 --> 00:21:09,180 Adjustment. OK, so here we have it. 213 00:21:09,180 --> 00:21:22,480 So revenue is ranging anywhere from below 50,000 a month to above 300 or not above certain months. 214 00:21:22,480 --> 00:21:31,080 But really, like above 250,000 for certain laboratories here. 215 00:21:31,080 --> 00:21:42,580 A lot of times we would like to think about this on a year to year basis so we can plot this by year. 216 00:21:42,580 --> 00:21:47,980 So we've talked about our forecasting model in the past. 217 00:21:47,980 --> 00:22:00,480 I don't think this package PMD-AutoAReMI is handling month effects correctly. 218 00:22:00,480 --> 00:22:04,980 And so I'm having to include all of the months. 219 00:22:04,980 --> 00:22:09,080 And so you can't really interpret the regression. 220 00:22:09,080 --> 00:22:17,880 So long story short, just keep in mind that I don't have full confidence in our forecasting model here. 221 00:22:17,880 --> 00:22:22,780 But some forecasts are better than no forecasts. 222 00:22:22,780 --> 00:22:28,280 So keep that in mind. Just want to bring that up that. 223 00:22:28,280 --> 00:22:34,980 This isn't estimating how I would prefer to estimate our forecast. 224 00:22:34,980 --> 00:22:44,480 So hopefully I'll find some time this spring or later this winter to redo this forecast. 225 00:22:44,480 --> 00:22:51,280 But long story short, we can still get some approximations for today. 226 00:22:51,280 --> 00:22:56,580 So we've got our prices, we've got our year forecast. 227 00:22:56,580 --> 00:23:06,980 So we're forecasting some labs will be cresting 3 million in revenue per year. 228 00:23:06,980 --> 00:23:14,180 Where, you know, you're looking at, you know, this lab. 229 00:23:14,180 --> 00:23:20,280 It's just a smaller lab and they've just they're going to have a little less than half a million in revenue. 230 00:23:20,280 --> 00:23:27,580 And then the second that we can measure a little less than one million in revenue. 231 00:23:27,580 --> 00:23:31,680 And then you see where the others fall out as well. 232 00:23:31,680 --> 00:23:42,880 The only main major prediction so far is that this lab, lab number 22, 233 00:23:42,880 --> 00:23:53,880 for which for whatever reason, our forecasting model predicts that their revenue is greatly going to diminish after 2023. 234 00:23:53,880 --> 00:24:00,280 So that would be the only real thing I would really point out so far is, you know, 235 00:24:00,280 --> 00:24:09,680 if you're running lab 22, I would look at your fundamentals and see maybe it's just the forecasting model. 236 00:24:09,680 --> 00:24:15,980 Something's odd with the forecasting model, but that's that's sort of. 237 00:24:15,980 --> 00:24:23,080 Something's going on there because you see the other forecasts are relatively staffed. 238 00:24:23,080 --> 00:24:29,780 15 million, three lab three around 14.6 million and then down the line. 239 00:24:29,780 --> 00:24:33,180 So long story short. 240 00:24:33,180 --> 00:24:45,980 I find it interesting that the lab with the second highest predicted revenue doesn't actually have the second highest forecasted total samples tested. 241 00:24:45,980 --> 00:24:52,180 So this would be sort of my my second major. 242 00:24:52,180 --> 00:25:01,480 Observation is. Lab number 21 is forecasted to test almost. 243 00:25:01,480 --> 00:25:11,780 150,000 samples over the next five years, but they have the lowest price. 244 00:25:11,780 --> 00:25:21,080 And. We need to do a bit more research, but it's not clear to me that price. 245 00:25:21,080 --> 00:25:25,780 Is a major factor for boosting revenue. 246 00:25:25,780 --> 00:25:37,580 So, for example, these are mandated quality control tests. So producers and processors demand for tests, 247 00:25:37,580 --> 00:25:43,980 maybe fairly not maybe I mean, to a large extent, it could. 248 00:25:43,980 --> 00:25:48,880 We will include the maybe it may be fairly inelastic. 249 00:25:48,880 --> 00:25:56,380 Right, you have to get the tests one way or the other, so you pretty much have to pay your price. 250 00:25:56,380 --> 00:26:03,940 And, you know, maybe you can go and travel to another lab, but these labs are. 251 00:26:03,940 --> 00:26:10,980 You saw they have different geographic locations, so you may not necessarily want to. 252 00:26:10,980 --> 00:26:17,680 Have to work out logistics to send your samples to a lab across the state. 253 00:26:17,680 --> 00:26:22,580 So that could play a factor. So there's a lot of factors that. 254 00:26:22,580 --> 00:26:26,980 May suggest there may be price inelasticity. 255 00:26:26,980 --> 00:26:32,140 For quality control testing, so. 256 00:26:32,140 --> 00:26:39,080 I think there's more research to be done, but I would naively suggest to lab number 21, 257 00:26:39,080 --> 00:26:45,980 you know that they may want to think about increasing their price, right? Because. 258 00:26:45,980 --> 00:26:53,440 The top two labs have higher prices and. 259 00:26:53,440 --> 00:27:00,780 You know, it's not clear that lowering their price is, you know, increasing their revenue. 260 00:27:00,780 --> 00:27:08,180 It may, you know, their low price may be one of the reasons why they're testing a lot of samples. 261 00:27:08,180 --> 00:27:14,840 I just think it would be interesting to see, OK, you know, if they did raise their price, you know, could they? 262 00:27:14,840 --> 00:27:17,480 So long story short. 263 00:27:17,480 --> 00:27:22,640 I'm not convinced that price necessarily shakes out with revenue. 264 00:27:22,640 --> 00:27:31,340 It does seem like there's a pretty strong correlation here between the number of samples you're testing and your revenue. 265 00:27:31,340 --> 00:27:36,740 But once again, not a perfect correlation there. 266 00:27:36,740 --> 00:27:38,440 So. 267 00:27:38,440 --> 00:27:44,920 Now on to I think the most interesting statistic of the day. 268 00:27:44,920 --> 00:27:52,740 And that is profitability, and so we'll need to take into account fixed costs. 269 00:27:52,740 --> 00:28:05,180 And what I was really trying to do was collect instrument data for various instruments that you would expect to see at a laboratory. 270 00:28:05,180 --> 00:28:14,020 However, it can be tricky to find prices for instruments, so prices vary and you may get a. 271 00:28:14,020 --> 00:28:19,140 It wouldn't surprise me if you get a lab by lab quote on a price. 272 00:28:19,140 --> 00:28:30,680 So long story short. Don't have a good measure on instrument prices, so I'm just going to have to do my best. 273 00:28:30,680 --> 00:28:45,180 And use my Bayesian prior on what total fixed costs may be, and we'll sort of do like they do in finance and do a couple of different projections, a low, a medium and a high. 274 00:28:45,180 --> 00:28:54,840 So I think, you know, was there a question or comment? 275 00:28:54,840 --> 00:28:58,740 So long story short, we'll start with the low projection. 276 00:28:58,740 --> 00:29:07,520 So keep in mind, Washington State doesn't have pesticide and heavy metals mandated, which would increase the cost of testing. 277 00:29:07,520 --> 00:29:14,240 So if you need those additional instruments, your fixed costs are going to increase. 278 00:29:14,240 --> 00:29:19,480 But let's start with the low projection. Just say, OK, let's say. 279 00:29:19,480 --> 00:29:27,880 It costs one million to increase to to open a lab. 280 00:29:27,880 --> 00:29:37,880 Well, essentially, what I've done here is I've taken the projected revenue that you would. 281 00:29:37,880 --> 00:29:42,720 The lab would earn over the next five years. 282 00:29:42,720 --> 00:29:49,480 Assuming that the instruments fully depreciate over five years. 283 00:29:49,480 --> 00:29:52,980 So here we've factored in the total fixed costs. 284 00:29:52,980 --> 00:30:03,780 And so that may not necessarily be the case. These instruments do see a lot of wear and tear in the technology is always increasing. 285 00:30:03,780 --> 00:30:09,040 People use scientific instruments from way back, though. 286 00:30:09,040 --> 00:30:14,620 So it's not uncommon to get a used scientific instrument. 287 00:30:14,620 --> 00:30:23,580 So there could be some resale value of these, but we'll just assume they fully depreciate over five years. 288 00:30:23,580 --> 00:30:34,180 So what I'm basically estimating here is the maximum variable cost per sample. 289 00:30:34,180 --> 00:30:38,580 So we're subtracting out fixed costs. 290 00:30:38,580 --> 00:30:50,620 So that means the rest is you basically write you need to keep your revenue above your fixed costs plus your variable costs. 291 00:30:50,620 --> 00:31:00,720 So here is a measure of what a lab's maximum variable cost could be to remain profitable. 292 00:31:00,720 --> 00:31:09,160 And keep in mind, we don't have a measure of prices for these labs, so just ignore these. 293 00:31:09,160 --> 00:31:14,660 So these are what we're focusing on. So. 294 00:31:14,660 --> 00:31:19,620 So long story short. 295 00:31:19,620 --> 00:31:26,320 The this would include all of your variable costs, so all of your supplies, 296 00:31:26,320 --> 00:31:32,860 say, plus your instrument fixes, so say your instrument breaks down and needs a repair. 297 00:31:32,860 --> 00:31:37,500 That would be a variable cost plus all of your labor. 298 00:31:37,500 --> 00:31:48,260 So you need a well-educated labor force to run your laboratory. 299 00:31:48,260 --> 00:31:54,760 So that can often come with high hourly wages for your employees. 300 00:31:54,760 --> 00:32:07,800 So long story short, you know, these labs will need to keep their variable costs per sample under these dollar amounts to remain profitable. 301 00:32:07,800 --> 00:32:12,360 So as you can see. As. 302 00:32:12,360 --> 00:32:16,660 Labs can test more and more samples. 303 00:32:16,660 --> 00:32:23,460 The. Room for increased variable costs increases. 304 00:32:23,460 --> 00:32:30,060 So this has a couple implications here. 305 00:32:30,060 --> 00:32:39,960 One profit, right? So if let's say all of the labs have just picking out a number, 306 00:32:39,960 --> 00:32:45,920 say they're all operating with sixty dollars variable cost per sample. 307 00:32:45,920 --> 00:32:55,160 Well. You know, lab 18 wouldn't be profitable if variable costs are sixty dollars per sample, 308 00:32:55,160 --> 00:33:00,560 whereas, you know, lab, you know, for. 309 00:33:00,560 --> 00:33:11,160 At sixty dollars per sample cost, they could, you know, you know, they could borderline. 310 00:33:11,160 --> 00:33:20,660 Well, I guess they're just testing a lot fewer samples, but they could have a higher profit margin than lab 22. 311 00:33:20,660 --> 00:33:26,600 So, you know, they've got, you know, a bit bit more room for their variable costs there. 312 00:33:26,600 --> 00:33:30,120 We're just going to say, let's just redo these. 313 00:33:30,120 --> 00:33:34,060 These estimates with a bit higher of a fixed cost. 314 00:33:34,060 --> 00:33:41,920 So here we'll assume a fixed cost of five million dollars to open a laboratory. 315 00:33:41,920 --> 00:33:47,360 You know, today, so say all of these labs had to expend five million dollars today. 316 00:33:47,360 --> 00:33:54,100 You know, what would their variable costs need to be over the next five years? 317 00:33:54,100 --> 00:34:02,100 So we see that if each lab had to expend five million dollars today, 318 00:34:02,100 --> 00:34:10,960 then, you know, lab four and 18, you know, they couldn't be profitable in over the next five years. 319 00:34:10,960 --> 00:34:20,160 That's a high fixed cost. But it just would be, you know, sort of a lesson to people thinking about opening a lab. 320 00:34:20,160 --> 00:34:24,660 You know, if you're looking at fixed costs of around five million, you know, 321 00:34:24,660 --> 00:34:33,160 you're going to need to test, you know, more than, you know, more than 30,000 samples a year. 322 00:34:33,160 --> 00:34:43,660 And, you know, you may want to start looking at, you know, at these labs as to what may be feasible. 323 00:34:43,660 --> 00:34:47,860 Keep in mind that they need to keep their variable costs low. 324 00:34:47,860 --> 00:34:56,420 So, you know, if they had five million dollars of fixed costs, you know, they'll need to keep their variable costs, 325 00:34:56,420 --> 00:35:01,720 you know, below, you know, twenty four and twenty eight dollars per sample. 326 00:35:01,720 --> 00:35:09,120 So that may be getting pretty tight with all the labor you need to employ. 327 00:35:09,120 --> 00:35:16,720 And as you can see, you know, these labs have slightly, lab 21 has slightly higher. 328 00:35:16,720 --> 00:35:23,800 And then, you know, lab three and seven have a good, you know, a good room there. 329 00:35:23,800 --> 00:35:34,200 So, you know, the lower they can keep their variable costs, the higher their profit will be. 330 00:35:34,200 --> 00:35:42,600 And then, you know, for example, in I've heard an estimate that in Massachusetts, 331 00:35:42,600 --> 00:35:47,400 you know, it would cost 15 million to open a laboratory. 332 00:35:47,400 --> 00:35:58,000 And, you know, at 15 million, really, none of the laboratories in Washington would be profitable at their given price. 333 00:35:58,000 --> 00:36:03,160 So keep in mind that in Massachusetts, they also have. 334 00:36:03,160 --> 00:36:06,500 Pesticide and heavy metal screening. 335 00:36:06,500 --> 00:36:14,800 So if you did need high capital expenditure to do heavy metal and pesticide screening, 336 00:36:14,800 --> 00:36:23,500 then that's that's why labs in states such as Massachusetts charge higher higher prices. 337 00:36:23,500 --> 00:36:27,620 So so that's the low, medium and high. 338 00:36:27,620 --> 00:36:35,300 And we'll just stick with the medium forecast for now. 339 00:36:35,300 --> 00:36:45,400 And we've been doing a lot of numbers, so we'll just look at a few plots here and then I'll share my takeaway with you. 340 00:36:45,400 --> 00:36:51,900 And we can start to talk about this because I know it was a lot of number crunching for one day. 341 00:36:51,900 --> 00:36:55,700 So sorry if it's been. 342 00:36:55,700 --> 00:37:02,960 Been a little abstract or much, but here is just a quick regression. 343 00:37:02,960 --> 00:37:09,300 We only have seven observations, so, you know, make of it what you will. 344 00:37:09,300 --> 00:37:18,340 But these were the four tests that we made for the various labs over the next five years. 345 00:37:18,340 --> 00:37:21,760 Regressed on their price. 346 00:37:21,760 --> 00:37:27,500 And basically what I can show you is right here, you know, two labs, you know, 347 00:37:27,500 --> 00:37:33,160 where they each have a price of $100 per test. 348 00:37:33,160 --> 00:37:41,300 But, you know, one lab is forecasted to make less than 2.5 million over the next five years, 349 00:37:41,300 --> 00:37:45,360 where the other lab is projected to make. 350 00:37:45,360 --> 00:37:49,660 You know, around 15 million in the next five years. 351 00:37:49,660 --> 00:38:02,060 So long story short is I don't know how much price competition will work in boosting your sales. 352 00:38:02,060 --> 00:38:04,700 So. 353 00:38:04,700 --> 00:38:14,560 So long story short, you know, if I was doing pricing, I mean, why not, you know, just crank it up, right? 354 00:38:14,560 --> 00:38:23,460 Because that's what it looks like some labs have done, and they've just they've just chosen a price of 120. 355 00:38:23,460 --> 00:38:27,900 Because it looks like they've just maybe inelastic demand. 356 00:38:27,900 --> 00:38:29,620 I'm not telling anyone what the price. 357 00:38:29,620 --> 00:38:35,700 I'm just saying, you know, if I was running a lab. 358 00:38:35,700 --> 00:38:42,500 So you can't compete on price or at least I'm not certain you can. 359 00:38:42,500 --> 00:38:46,100 So at least not to a great extent. 360 00:38:46,100 --> 00:38:50,860 So, you know, what are some of the other factors you can compete on? 361 00:38:50,860 --> 00:39:03,020 Well, here is essentially, you know, the maximum variable cost per sample regressed against revenue. 362 00:39:03,020 --> 00:39:07,760 And so here we see a positive correlation. 363 00:39:07,760 --> 00:39:18,560 Keep in mind, we only have seven observations, but essentially what I'm starting to kind of gather from this data is, you know, 364 00:39:18,560 --> 00:39:26,700 it looks like, you know, if you're able to sort of give yourself a bit more cushion and, you know, 365 00:39:26,700 --> 00:39:37,700 have room to have a higher variable cost and still make a profit, then that can increase your revenue. 366 00:39:37,700 --> 00:39:45,200 And just to kind of show a few more figures here to flesh out the story. 367 00:39:45,200 --> 00:39:52,000 Here is total samples against variable costs. 368 00:39:52,000 --> 00:40:03,360 So it looks like, you know, if you are able to have a higher variable cost, it looks like you're able to test more samples. 369 00:40:03,360 --> 00:40:12,660 And it also looks, I mean, I would assume there is sort of simultaneous causation here in that. 370 00:40:12,660 --> 00:40:17,240 The more samples you test. 371 00:40:17,240 --> 00:40:26,060 That's not the best plot ever, but the more samples you test, the. 372 00:40:26,060 --> 00:40:31,660 You know, the higher your variable cost can be. 373 00:40:31,660 --> 00:40:36,760 So let's just print this back out here. 374 00:40:36,760 --> 00:40:39,440 And I'll try to get to my takeaway here. 375 00:40:39,440 --> 00:40:44,240 So so basically. 376 00:40:44,240 --> 00:40:53,360 We're starting to get a little abstract and a little imperfect here, but basically what I'm trying to argue is. 377 00:40:53,360 --> 00:40:57,060 You know, to run a profitable lab. 378 00:40:57,060 --> 00:41:05,600 I think focusing on your variable costs is one of your most important. 379 00:41:05,600 --> 00:41:14,000 Factors that's under your control, so you can only do so many pricing gains. 380 00:41:14,000 --> 00:41:22,800 And if you look at the laboratory's websites, you see that they are engaged in pricing strategies. 381 00:41:22,800 --> 00:41:26,860 They do, for example, they do bulk deals. 382 00:41:26,860 --> 00:41:34,300 There's various discounts and packages, so they're engaged in pricing strategies, 383 00:41:34,300 --> 00:41:39,600 but I'm not certain that those are effective strategies. 384 00:41:39,600 --> 00:41:44,760 And. I would like to just introduce a concept to you. 385 00:41:44,760 --> 00:41:47,660 Price as a signal. 386 00:41:47,660 --> 00:41:56,500 So, you know, we think, oh, yes, just increasing the price or decreasing the price, and that'll change quantity. 387 00:41:56,500 --> 00:42:02,520 But retailers are hip to this and. 388 00:42:02,520 --> 00:42:08,920 It may make sense for a laboratory to think about this price can be a signal of quality. 389 00:42:08,920 --> 00:42:20,860 So if you charge a higher price, you know that that could just be a signal to others that your higher quality. 390 00:42:20,860 --> 00:42:30,500 And so, you know, if you're charging a price of 80, you know, and other people are charging a price of 120. 391 00:42:30,500 --> 00:42:38,600 That may be signaling to people that you're doing poor quality tests, 392 00:42:38,600 --> 00:42:45,900 whereas you're just trying that pricing strategy out because you think, you know, it may be profitable. 393 00:42:45,900 --> 00:42:50,460 So so just, you know, that's a factor. 394 00:42:50,460 --> 00:42:58,060 Of course, you know, the higher your price is, you know, the more wiggle room you have on. 395 00:42:58,060 --> 00:43:06,560 You know, to keep your variable costs under control, under control while making a profit. 396 00:43:06,560 --> 00:43:13,200 So so that's sort of my main takeaway from today is. 397 00:43:13,200 --> 00:43:25,900 And that was sort of the lesson I was trying to say is, you know, it's important to keep your variable costs right. 398 00:43:25,900 --> 00:43:31,760 Gloves, solvents, disposables, it's important to keep those low. 399 00:43:31,760 --> 00:43:36,800 And in economics, there's a mathematical proof to it. 400 00:43:36,800 --> 00:43:47,160 But essentially, there's an argument that cost minimization results in the same outcome as profit maximization. 401 00:43:47,160 --> 00:43:54,100 So instead of just. Question or comment. 402 00:43:54,100 --> 00:43:59,360 Yes, there was a question about the recording. Yes, there will be a recording available. 403 00:43:59,360 --> 00:44:05,300 I'm sorry. I'm sorry, actually, you know, I'm running into two calls and I wanted to ask a lot of questions there. 404 00:44:05,300 --> 00:44:08,760 But just because of like my office work results there. 405 00:44:08,760 --> 00:44:12,560 So I could not. But at the last, like I thought, OK, maybe it's only 10 minutes. 406 00:44:12,560 --> 00:44:18,640 And if you guys disconnect, I need to have that video like audio, some recording maybe. 407 00:44:18,640 --> 00:44:24,360 Because next week I will come to, you know, join this call. 408 00:44:24,360 --> 00:44:27,500 And with a lot of questions. Awesome. 409 00:44:27,500 --> 00:44:30,960 We'll love to have you next week and I'll get the recording up. 410 00:44:30,960 --> 00:44:37,760 And then next week we'll start talking. 411 00:44:37,760 --> 00:44:45,000 Yes, so bring your bring your questions next time and we'll make sure to spend a ton of time answering your questions. 412 00:44:45,000 --> 00:44:49,600 And we'll be focusing on producers and processors next time. 413 00:44:49,600 --> 00:45:00,100 But long story short is, you know, just instead of just being obsessed with profit and prices and, 414 00:45:00,100 --> 00:45:05,440 you know, how many total samples are you testing and this and that, 415 00:45:05,440 --> 00:45:11,340 you could also be just as successful by trying to be efficient. 416 00:45:11,340 --> 00:45:19,540 So this is where I say, you know, just if you just have this laser focus on efficiencies and optimizing your workflow, 417 00:45:19,540 --> 00:45:24,960 then you can, you know, you can run a profitable lab here. 418 00:45:24,960 --> 00:45:33,960 Right. I mean, you know, it's hard to imagine that these laboratories aren't going to be profitable. 419 00:45:33,960 --> 00:45:39,540 It's possible. Right. It's expensive to operate a laboratory. 420 00:45:39,540 --> 00:45:49,540 So, for example, you know, lab 21, you know, they could bring in 12 million dollars, you know, in the next five years. 421 00:45:49,540 --> 00:45:59,340 But if they can't keep their variable costs under forty six dollars per sample, then they're not going to be profitable. 422 00:45:59,340 --> 00:46:12,640 And forty six dollars per sample, I mean, we're talking about things like disposables and variable costs include things like your employee labor. 423 00:46:12,640 --> 00:46:24,800 Then, you know, every, you know, centers matters, especially on things like your like, you know, your disposables. 424 00:46:24,800 --> 00:46:34,040 And basically, you know, I do know, you know, there's no need to cut corners because it does look like. 425 00:46:34,040 --> 00:46:44,060 You can make a profit with room to spare. But but that's sort of my my main takeaway is, 426 00:46:44,060 --> 00:46:55,640 you know, rather than, you know, just try to just be laser focused on price or, you know, how just just thinking, oh, how can we bring in more and more samples? 427 00:46:55,640 --> 00:47:07,280 You know, maybe you should think about costs, you know, you know, maybe there are ways you can improve your workflow or, you know, 428 00:47:07,280 --> 00:47:14,960 I always think there is endless opportunities for just becoming more and more efficient, just just doing things better and better and better. 429 00:47:14,960 --> 00:47:20,240 And that's why I started, you know, analytics. 430 00:47:20,240 --> 00:47:30,440 So, you know, just quick shout out to you know, analytics for making this all possible. And, you know, if any of you are interested in helping laboratories 431 00:47:30,440 --> 00:47:39,080 keep their variable costs as low as possible, then check out the analytics software, the analytics engine. 432 00:47:39,080 --> 00:47:48,320 So here is software specifically designed to help laboratories keep their costs as low as possible. 433 00:47:48,320 --> 00:47:57,120 Right. If you're having your analyst enter data, that's increasing your variable costs. 434 00:47:57,120 --> 00:48:01,360 And there's many better things that your analysts can be doing. 435 00:48:01,360 --> 00:48:17,440 So I've seen in the past where people have gotten their time freed up and they're able to enroll in courses part time and they just become more educated and just their careers blossom. 436 00:48:17,440 --> 00:48:24,120 So, you know, there's a lot of ways that you can just become more efficient over time. 437 00:48:24,120 --> 00:48:28,640 And then, you know, that way, you know, you don't have to compete on price. 438 00:48:28,640 --> 00:48:37,560 And, you know, if you get more efficient, I would argue that you're going to be able to test more samples. 439 00:48:37,560 --> 00:48:45,320 You can test more samples. That'll give you a bit more wiggle room and you can do a better and better job. 440 00:48:45,320 --> 00:48:50,520 So they can actually increase the quality of their tests. Right. 441 00:48:50,520 --> 00:48:58,440 So if you've got a bit more room, you can do even more stringent quality control. 442 00:48:58,440 --> 00:49:03,000 So I think it's just a win win for everyone. 443 00:49:03,000 --> 00:49:10,560 So focus on costs, test more samples, test your samples better. 444 00:49:10,560 --> 00:49:15,440 Increase your price because you're doing quality work. Right. 445 00:49:15,440 --> 00:49:23,200 It's just that that's how I would go about running a profitable lab. 446 00:49:23,200 --> 00:49:31,640 So sort of a long winded long winded story for today, but. 447 00:49:31,640 --> 00:49:34,880 But I think it was an interesting one. 448 00:49:34,880 --> 00:49:44,320 So are there any thoughts, comments? I'll just put up a plot. 449 00:49:44,320 --> 00:49:56,640 But any thoughts or comments about the labs in Washington state or across the country or really any questions whatsoever? 450 00:49:56,640 --> 00:50:03,800 Maybe maybe not related to maybe not to the Washington state thing. 451 00:50:03,800 --> 00:50:08,080 I was, you know, I'm back from my office. I'm sorry. 452 00:50:08,080 --> 00:50:12,720 I winded up like, OK, they wind up for five minutes. So I have a lot of questions. 453 00:50:12,720 --> 00:50:16,400 I'm joining this session from last three consecutive times. 454 00:50:16,400 --> 00:50:19,520 I'm trying to get a hold of this thing. 455 00:50:19,520 --> 00:50:25,760 Number one is like, OK, this cannabis is a kind of like the commercial tool or what? 456 00:50:25,760 --> 00:50:29,920 Or we are we are most of time talking about like, OK, these labs, 457 00:50:29,920 --> 00:50:36,840 because I am trying to where I'm trying to pick up your brain is I'm into the data science. 458 00:50:36,840 --> 00:50:46,120 Just just stepping into this one. And I am finding like, OK, any of the existing forums which are talking about more of like 459 00:50:46,120 --> 00:50:50,880 like today you were talking about the forecasting and all that stuff and definitely maybe not the forecasting. 460 00:50:50,880 --> 00:50:56,680 And you for for for only for this one, it will be applicable to anything else as well. 461 00:50:56,680 --> 00:51:04,800 If it is my right understanding, like this this forum is specific to a certain need or is it really the generalized version of it, 462 00:51:04,800 --> 00:51:10,760 like which we can implement anywhere else in the industry? 463 00:51:10,760 --> 00:51:15,480 Really, you know, these are tools you can use really anywhere. 464 00:51:15,480 --> 00:51:21,840 So this forecasting model I used is if you want to learn more, 465 00:51:21,840 --> 00:51:28,440 tune in, I'd recommend checking out Saturday morning statistics when we can go in depth on this. 466 00:51:28,440 --> 00:51:32,960 But we really just use what's called a REMA forecasting. 467 00:51:32,960 --> 00:51:38,040 And so this just uses historic observations. 468 00:51:38,040 --> 00:51:42,280 So we literally just use. 469 00:51:42,280 --> 00:51:52,400 The historic right. So these are the historic, the historic trends we've seen from the various labs. 470 00:51:52,400 --> 00:51:58,680 And with the REMA forecasting, all we use is the one time series. 471 00:51:58,680 --> 00:52:03,360 I always also use the month so you can add in other variables. 472 00:52:03,360 --> 00:52:11,640 But really, all you're using is the past historic trends of these labs. 473 00:52:11,640 --> 00:52:18,320 And you're forecasting it forward. So to not only to you can. 474 00:52:18,320 --> 00:52:26,240 That's what I love a REMA forecasting is you can forecast any time series, 475 00:52:26,240 --> 00:52:34,880 any data that you can track over time, you can forecast with the REMA. 476 00:52:34,880 --> 00:52:44,880 There's a few stipulations like it's good to have like a consistent interval and it's best not to have missing data. 477 00:52:44,880 --> 00:52:53,640 So there's and there's like some statistical checks you can do to make sure. 478 00:52:53,640 --> 00:52:59,880 You're not leaving any variation left on the table, so to speak. 479 00:52:59,880 --> 00:53:07,080 But it's it's powerful because a lot of people stop here. 480 00:53:07,080 --> 00:53:13,400 Right. I see a lot of people, they they do like historic analysis, which is useful. 481 00:53:13,400 --> 00:53:20,480 Right. It's it's awesome to know how many samples were tested by month by the labs. 482 00:53:20,480 --> 00:53:29,160 But you can just play it forward. Right. You can just use statistical techniques, forecast this forward. 483 00:53:29,160 --> 00:53:35,000 It's not going to be perfect. We're not going to hit everything perfectly. 484 00:53:35,000 --> 00:53:39,960 But, you know, at least now you have a forecast for the future. 485 00:53:39,960 --> 00:53:47,480 And, you know, like I was saying. You know, yes. 486 00:53:47,480 --> 00:53:51,280 Yeah. So one more question here. So I get your point. 487 00:53:51,280 --> 00:53:56,520 Like when you said, OK, it is that's good for forecasting. 488 00:53:56,520 --> 00:54:06,600 Let me let me say, like, in other words, like what my expectations, if it is like if it is useful for me, would be is 489 00:54:06,600 --> 00:54:10,600 we deal with a lot of volume of data, like which is coming daily. 490 00:54:10,600 --> 00:54:16,040 And you make you can say like, OK, millions of records. 491 00:54:16,040 --> 00:54:21,680 Right. And let's say it's the claims data, like millions of claims we are getting day in and out. 492 00:54:21,680 --> 00:54:31,320 Right. And I want to be in a position like where I can say, OK, hey, for because let's say I for this year, 493 00:54:31,320 --> 00:54:40,160 I think we churned out like almost close to maybe 30 million of records. 494 00:54:40,160 --> 00:54:43,160 We churned out that one. And that is just growing. 495 00:54:43,160 --> 00:54:47,840 And we are just started like three years back. It is not like two old data. 496 00:54:47,840 --> 00:54:54,800 We don't have much data from the from the longevity perspective, but we have like, OK, within the short time, 497 00:54:54,800 --> 00:55:05,040 we have grown enormously, like from like, OK, from few hundred thousands from few thousands towards the hundred thousands. 498 00:55:05,040 --> 00:55:14,240 And now we are churning out approximately 300 to 400 thousand in a day and which will be giving us a huge amount. 499 00:55:14,240 --> 00:55:21,600 I don't know where we are going. And I want to see be in a position like, OK, hey, I have the, you know, 500 00:55:21,600 --> 00:55:27,800 a prediction that based on like whatever the data from the last three years. 501 00:55:27,800 --> 00:55:34,120 Because you were talking about the samples as well. I was just since I was into the two calls, 502 00:55:34,120 --> 00:55:37,960 like I was thinking this sample, which you're talking about, like the sample of the population. 503 00:55:37,960 --> 00:55:42,280 And I think you were talking about the sample about like the lab testing. Right. 504 00:55:42,280 --> 00:55:51,280 So somewhat I could differentiate. But then I think, OK, maybe maybe I need a dedicated call for this one. 505 00:55:51,280 --> 00:55:59,880 Maybe with you. And questions were how often you guys are doing this session so that I can be regular here 506 00:55:59,880 --> 00:56:06,480 and I can pick up some brain for like, OK, these kind of things can be applied at my side. 507 00:56:06,480 --> 00:56:12,320 Second thing, is it a commercial tool or is it just like an open source Python like kind of library? 508 00:56:12,320 --> 00:56:17,240 Awesome questions. So it's all open source. So it's under the MIT license. 509 00:56:17,240 --> 00:56:22,760 So if you just, you know, give credit to Caneletics and that's awesome. 510 00:56:22,760 --> 00:56:26,800 And then to answer your second your other questions, we do meet weekly. 511 00:56:26,800 --> 00:56:33,320 So there's a group Saturday morning statistics where we specifically do statistics. 512 00:56:33,320 --> 00:56:38,000 And then we do the Wednesday cannabis data science each week. 513 00:56:38,000 --> 00:56:46,960 And then as to your question with data, the time series forecasting can be done. 514 00:56:46,960 --> 00:56:55,600 Wait, did I do that here? Yes, this model, it can be done on any time series. 515 00:56:55,600 --> 00:57:08,000 So here you see this data. It's just. It's, you know, it's yearly data. 516 00:57:08,000 --> 00:57:16,800 I wonder. But but as long. 517 00:57:16,800 --> 00:57:26,960 You will be lost, too, I guess. Hello. 518 00:57:26,960 --> 00:57:32,600 Is it only me or anybody else? OK, another three more people. 519 00:57:32,600 --> 00:57:40,560 Yes, can you listen to me, Dharma? Paul, I can hear you, but I might as well answer the question for Keegan. 520 00:57:40,560 --> 00:57:47,480 I think you're you described like huge data sets, like millions of data points. 521 00:57:47,480 --> 00:57:55,320 And if you're looking for humongous data sets with millions of data points, you're in the wrong space. 522 00:57:55,320 --> 00:58:06,720 But because cannabis has very few data points, a lot of this situation is primarily if we're going to go into data science realm, 523 00:58:06,720 --> 00:58:15,320 unsupervised learning where you have the bootstrap data samples and supplement data samples and everything like that. 524 00:58:15,320 --> 00:58:26,040 They're the only big data sets within the cannabis space is held in private businesses because cannabis is big business. 525 00:58:26,040 --> 00:58:37,280 And it particularly relies with cannabis genome sequencing and to tie it into this cannabis conversation. 526 00:58:37,280 --> 00:58:43,200 That is particularly why Eastern Washington. 527 00:58:43,200 --> 00:58:55,400 Is very well known for their testing sample, because that is the forefront of cannabis research in terms of growing and testing and sampling. 528 00:58:55,400 --> 00:58:58,040 Got it. Appreciate that. Thank you. 529 00:58:58,040 --> 00:59:03,560 And this sample which you have created, like is it available somewhere in the GitHub or somewhere? 530 00:59:03,560 --> 00:59:07,880 Or is it shareable? I mean, first thing I should ask, like, is it shareable? 531 00:59:07,880 --> 00:59:18,120 Yes. So check out the Cannabix Data Science GitHub repository. 532 00:59:18,120 --> 00:59:27,760 And I've got a link here in this latest script lab profitability to the data sets. 533 00:59:27,760 --> 00:59:34,680 So that way you can download those. Can you please put into the message? 534 00:59:34,680 --> 00:59:39,960 Just fork the repo, dude. If you're a real data scientist, you know what a fork on GitHub. 535 00:59:39,960 --> 00:59:42,240 Here we are. 536 00:59:42,240 --> 00:59:54,240 So. Thank you. And here's the link to the actual data. 537 00:59:54,240 --> 01:00:00,320 And the largest data there is going to be the sales data. And Graham was right. 538 01:00:00,320 --> 01:00:04,480 We've got to work with the data points we can get. 539 01:00:04,480 --> 01:00:12,720 So but I would just just push back slightly in that I think the the models, the ARIMA models, 540 01:00:12,720 --> 01:00:16,640 they work on small and large data sets. 541 01:00:16,640 --> 01:00:23,440 I think as long as you've got a timestamp, you can start to aggregate by time. 542 01:00:23,440 --> 01:00:29,040 So say you're looking at sales over time or you were looking at claims. 543 01:00:29,040 --> 01:00:33,920 So you can just look at number of claims over time. 544 01:00:33,920 --> 01:00:42,080 And any frequency, right? So here. 545 01:00:42,080 --> 01:00:47,520 Right, we already showed you one with yearly and this should be monthly. 546 01:00:47,520 --> 01:00:52,760 So really anything that you can put a timestamp on. 547 01:00:52,760 --> 01:00:57,960 And so you could run your ARIMA model. 548 01:00:57,960 --> 01:01:06,040 Minute by minute, hour by hour, so you could get, you know, same day predictions. 549 01:01:06,040 --> 01:01:11,960 I found they're useful for like week, like one week ahead, one month ahead forecasts. 550 01:01:11,960 --> 01:01:16,440 But if you're ambitious and you want to do them in real time with your data, 551 01:01:16,440 --> 01:01:18,680 there's really no reason why you can't. 552 01:01:18,680 --> 01:01:27,320 All you just all you need is just a series of values that you're keeping track of over time. 553 01:01:27,320 --> 01:01:30,360 And you're off to the races. 554 01:01:30,360 --> 01:01:35,480 And one more question. So this data cleanup, the data which you have used, 555 01:01:35,480 --> 01:01:36,920 like you would send me both of the links. 556 01:01:36,920 --> 01:01:42,760 So the data is already cleaned up or you are doing it like in your program somewhere? 557 01:01:42,760 --> 01:01:46,440 Oh, yes. All the cleanings done here. 558 01:01:46,440 --> 01:01:52,360 So basically, so here I just read in all of the data. 559 01:01:52,360 --> 01:02:00,360 And if you're reading it in, you can just read it in sort of bit by bit at a time. 560 01:02:00,360 --> 01:02:05,160 And in here, you know, you can pick and choose your variables. 561 01:02:05,160 --> 01:02:06,920 Got it. 562 01:02:06,920 --> 01:02:10,760 I just scraped the tip of the iceberg. 563 01:02:10,760 --> 01:02:17,160 But like I said, if you're feeling ambitious, then check out the sales data. 564 01:02:17,160 --> 01:02:22,760 There is. I really appreciate that. 565 01:02:22,760 --> 01:02:29,560 I will not say like, OK, the millions of the data points, I would say the data is too much. 566 01:02:29,560 --> 01:02:31,320 But yeah, the features are limited. 567 01:02:31,320 --> 01:02:34,040 Like, OK, it's not expanding any further. 568 01:02:34,040 --> 01:02:37,640 It's like merely like 10 to 20 features, I would be saying. 569 01:02:37,640 --> 01:02:47,000 But yes, there is a huge diversification in terms of like one data is segmented into 570 01:02:47,000 --> 01:02:48,280 this type of process. 571 01:02:48,280 --> 01:02:50,120 Second is into this type of process. 572 01:02:50,120 --> 01:02:51,880 Like, as I said, like the claims. 573 01:02:51,880 --> 01:02:55,560 So it could be like, OK, this guy is having this kind of claim. 574 01:02:55,560 --> 01:02:57,480 That guy is having that kind of claim. 575 01:02:57,480 --> 01:03:02,760 So in a nutshell, like I'm getting like the variety of the claims basically, 576 01:03:03,400 --> 01:03:05,560 maybe 10 to 20 types of that one. 577 01:03:05,560 --> 01:03:08,840 And then eventually it turns into like millions of records. 578 01:03:08,840 --> 01:03:15,560 So I was thinking like, OK, like to be very honest, like I even started this one like 579 01:03:15,560 --> 01:03:21,160 some time back and then I think, OK, what actually like would be would be helpful for me? 580 01:03:21,160 --> 01:03:25,240 And I am honestly like going from kind of the door to door. 581 01:03:25,240 --> 01:03:27,480 People say, OK, if you have too much of data, OK, go ahead. 582 01:03:27,480 --> 01:03:28,440 Do the deep learning. 583 01:03:28,440 --> 01:03:31,560 OK, if you are like, OK, this no, no, no, you do this supervised learning. 584 01:03:31,560 --> 01:03:35,320 And today, again, like, OK, it's not going anywhere. 585 01:03:35,320 --> 01:03:40,840 So let me let me start with some point like how at least they start handling these things. 586 01:03:41,480 --> 01:03:44,440 And, you know, that's where I am. 587 01:03:44,440 --> 01:03:49,720 And I appreciate that, like you guys, at least, you know, sharing this information so that I can 588 01:03:49,720 --> 01:03:55,080 next time probably I will see you with a lot of questions what I have and what I don't. 589 01:03:55,080 --> 01:03:59,960 Or probably I'll try to grab some sample, maybe not the sample from the original data, 590 01:03:59,960 --> 01:04:04,760 but I will just create a replica of like with some dummy values so that we can, 591 01:04:04,760 --> 01:04:09,560 if you guys can, you know, think about like, yeah, you can you can guide me. 592 01:04:09,560 --> 01:04:13,800 We lost you again. 593 01:04:13,800 --> 01:04:17,800 Yes. 594 01:04:17,800 --> 01:04:23,800 I don't really yet. 595 01:04:23,800 --> 01:04:27,800 We lost you for for last 15 seconds. 596 01:04:27,800 --> 01:04:29,000 You have to repeat what I said. 597 01:04:29,000 --> 01:04:29,800 I'm sorry. 598 01:04:29,800 --> 01:04:30,840 Oh, yes. 599 01:04:30,840 --> 01:04:33,400 So my background is in economics. 600 01:04:33,400 --> 01:04:34,840 So that's why I use. 601 01:04:34,840 --> 01:04:40,120 A lot of tools that economists may use. 602 01:04:40,120 --> 01:04:44,120 So I'm just as eager as you are to learn some of these interesting, 603 01:04:44,120 --> 01:04:48,120 say, machine learning tools or other data science tools. 604 01:04:48,120 --> 01:04:52,120 So just because I haven't talked about it doesn't mean it's not useful. 605 01:04:52,120 --> 01:04:57,640 So if you find an awesome, cool data science tool, machine learning tool, 606 01:04:57,640 --> 01:05:00,120 then please share it because I'd be. 607 01:05:00,120 --> 01:05:00,600 Yeah. 608 01:05:00,600 --> 01:05:05,000 Drilled to explore it and apply it to some of this cannabis data. 609 01:05:05,000 --> 01:05:06,040 Yep. 610 01:05:06,040 --> 01:05:06,680 For sure. 611 01:05:06,680 --> 01:05:07,960 Thank you very much. 612 01:05:07,960 --> 01:05:08,520 Awesome. 613 01:05:08,520 --> 01:05:10,680 Well, thank you all for coming. 614 01:05:10,680 --> 01:05:13,720 I'm going to go ahead and stop the presentation for today. 615 01:05:13,720 --> 01:05:18,520 I hope I hope you all were able to get something out of it. 616 01:05:18,520 --> 01:05:20,120 It was a little long. 617 01:05:20,120 --> 01:05:21,720 It was a little long winded today. 618 01:05:21,720 --> 01:05:26,200 So next week, I'll I'll try to make it a bit more simple and a bit more handy. 619 01:05:26,200 --> 01:05:29,240 So I'll try to make it a little bit more simple and a bit more handy. 620 01:05:29,240 --> 01:05:32,200 I'll try to make it a little bit more simple and a bit more hands on, 621 01:05:32,200 --> 01:05:33,400 a bit more discussion based. 622 01:05:33,400 --> 01:05:35,800 So I think next week will be a good one, too. 623 01:05:35,800 --> 01:05:36,600 I appreciate that. 624 01:05:36,600 --> 01:05:37,160 Thank you. 625 01:05:37,160 --> 01:05:37,720 Thank you. 626 01:05:37,720 --> 01:05:38,360 Great. 627 01:05:38,360 --> 01:05:39,080 Awesome. 628 01:05:39,080 --> 01:05:40,520 Thank you all for coming. 629 01:05:40,520 --> 01:06:00,200 Have a productive week.