1
00:00:00,000 --> 00:00:13,200
Good morning.

2
00:00:13,200 --> 00:00:23,160
And so I guess as you know, I'm doing a venture here in Oklahoma City or Oklahoma in general

3
00:00:23,160 --> 00:00:26,960
for a little bit, essentially doing some outreach to the labs here.

4
00:00:26,960 --> 00:00:32,240
So I'll show you on the map the places I've been and you know, that you know, CanLytics

5
00:00:32,240 --> 00:00:39,480
is spreading the word about our, you know, Canvas testing solutions and I'll tell you

6
00:00:39,480 --> 00:00:42,160
a little bit about the data here.

7
00:00:42,160 --> 00:00:46,200
But before that, like, how are things going with you Charles?

8
00:00:46,200 --> 00:00:48,800
Pretty good.

9
00:00:48,800 --> 00:00:49,800
I think I have a job.

10
00:00:49,800 --> 00:00:50,800
Oh.

11
00:00:50,800 --> 00:00:54,840
I think I'm going to start like in a week or two.

12
00:00:54,840 --> 00:00:58,120
Well, please, please tell me.

13
00:00:58,120 --> 00:01:08,600
It's working with like 3D imaging and making measurements off of 3D image scans using,

14
00:01:08,600 --> 00:01:14,080
using like, you know, using machine learning.

15
00:01:14,080 --> 00:01:16,680
Interesting.

16
00:01:16,680 --> 00:01:22,400
So have you done much machine learning work?

17
00:01:22,400 --> 00:01:28,160
So there is actually a lab that CanLytics is interested in the work CanLytics is doing

18
00:01:28,160 --> 00:01:33,520
and one of the things they're interested in is finding uses for machine learning.

19
00:01:33,520 --> 00:01:36,800
So how do you go about even using it?

20
00:01:36,800 --> 00:01:40,600
You get the problem first or do you have a good model in mind?

21
00:01:40,600 --> 00:01:44,200
So how do you even go about things?

22
00:01:44,200 --> 00:01:51,120
Well, I mean, yeah, you need to figure out, you need to figure out what they want to solve

23
00:01:51,120 --> 00:01:56,840
and then, you know, and then try and find a model that will work with that or a lot of

24
00:01:56,840 --> 00:02:01,600
times you have to build your own model, you know, and you have to do some feature engineering

25
00:02:01,600 --> 00:02:11,560
and it's, you know, it's a lot of trial and error and kind of just sort of knowing, you

26
00:02:11,560 --> 00:02:21,360
know, having some kind of intuition as to what to try.

27
00:02:21,360 --> 00:02:28,560
So say, so the problem we have at hand is, say like a lab wanted to like maybe almost

28
00:02:28,560 --> 00:02:34,720
predict the chance of a sample failing as it comes in the door.

29
00:02:34,720 --> 00:02:45,720
So that way, you know, you can just have an eye out, you know, in case and you can like,

30
00:02:45,720 --> 00:02:50,680
you know, run those samples right away in case there's a failure, you can, you know,

31
00:02:50,680 --> 00:02:52,320
let people know right away.

32
00:02:52,320 --> 00:03:00,320
So is that a reasonable use for machine learning?

33
00:03:00,320 --> 00:03:05,000
And if so, how would you grow about even starting?

34
00:03:05,000 --> 00:03:08,480
Yeah, I mean, you could do that.

35
00:03:08,480 --> 00:03:12,520
I thought about trying to do that with the Washington data.

36
00:03:12,520 --> 00:03:15,880
There aren't a lot of failures in there, although I would imagine in most cases there aren't

37
00:03:15,880 --> 00:03:19,280
a lot of failures.

38
00:03:19,280 --> 00:03:24,480
But so you're probably going to, the only thing you're really going to know ahead of

39
00:03:24,480 --> 00:03:34,560
time, right, is the producer and the strain, right?

40
00:03:34,560 --> 00:03:40,120
And so it was real interesting.

41
00:03:40,120 --> 00:03:45,760
So at KanaCon, there was a presentation about heavy metals.

42
00:03:45,760 --> 00:03:53,200
And what's interesting is if you look at the distribution of, you know, heavy metals that

43
00:03:53,200 --> 00:04:01,040
have been tested, say in water, soil, even in this case, they were looking at human blood

44
00:04:01,040 --> 00:04:03,960
samples.

45
00:04:03,960 --> 00:04:06,280
It's not evenly distributed.

46
00:04:06,280 --> 00:04:12,160
So there's, you know, there's hot spots, you know, certain places are sort of cleaner

47
00:04:12,160 --> 00:04:14,920
and others are dirtier than others.

48
00:04:14,920 --> 00:04:23,120
So I think geographic region, I mean, I've got a hunch it may play a factor, but I don't

49
00:04:23,120 --> 00:04:25,560
know how well it would explain.

50
00:04:25,560 --> 00:04:32,600
Yeah, you can use geographic region as one of the inputs, although that would be correlated

51
00:04:32,600 --> 00:04:36,240
to the producer.

52
00:04:36,240 --> 00:04:39,000
But you might have a producer you've never seen before.

53
00:04:39,000 --> 00:04:43,400
So yeah, you might just want to, instead of going by the producer, you could go by where

54
00:04:43,400 --> 00:04:51,400
they're located, time of year, maybe time of year that has some effect on it.

55
00:04:51,400 --> 00:04:58,160
So yeah, you would just try and find things that you're going to know ahead of time and

56
00:04:58,160 --> 00:05:03,280
try and predict, you know, the outcome.

57
00:05:03,280 --> 00:05:07,040
It seems like a problem you could, you know, you could at least try.

58
00:05:07,040 --> 00:05:12,200
Whether you have enough data and whether that data would give you enough information to

59
00:05:12,200 --> 00:05:18,440
predict something is another question.

60
00:05:18,440 --> 00:05:24,240
Do you think that's even the most value added use?

61
00:05:24,240 --> 00:05:27,480
Or like, do you think it would even be that worthwhile?

62
00:05:27,480 --> 00:05:39,200
Or are you thinking, like, so for example, can you just discover gold with machine learning,

63
00:05:39,200 --> 00:05:45,280
just find avenues that you hadn't thought of, or do you really have to have a, you know,

64
00:05:45,280 --> 00:05:49,240
great value added idea to begin with?

65
00:05:49,240 --> 00:05:53,680
You can find, I mean, you can find things that you don't know about.

66
00:05:53,680 --> 00:05:56,560
You could do some sort of like clustering algorithm.

67
00:05:56,560 --> 00:05:58,280
I mean, that's something else.

68
00:05:58,280 --> 00:06:04,520
You could, yeah, you could do a clustering algorithm and see, you know, if certain, if

69
00:06:04,520 --> 00:06:12,000
certain regions or producers or strains and that, you know, fall, fall into a group that

70
00:06:12,000 --> 00:06:15,440
fails more.

71
00:06:15,440 --> 00:06:21,680
So, see if there's any similar characteristics.

72
00:06:21,680 --> 00:06:22,680
Yeah.

73
00:06:22,680 --> 00:06:23,680
Yeah.

74
00:06:23,680 --> 00:06:30,240
I mean, I would be interested in working on something like that.

75
00:06:30,240 --> 00:06:35,080
That would be, that would be pretty cool.

76
00:06:35,080 --> 00:06:40,400
Well, with the Washington State data, we could technically do that, right?

77
00:06:40,400 --> 00:06:49,080
So we've got their geographic location and potentially the failure rates, right?

78
00:06:49,080 --> 00:06:55,840
So, so I think we may need to add that to the list.

79
00:06:55,840 --> 00:07:02,840
So I could, I guess work on getting those data points, just the, well, we already have

80
00:07:02,840 --> 00:07:04,120
several of these data points.

81
00:07:04,120 --> 00:07:07,400
So we've got essentially their latitude and longitude.

82
00:07:07,400 --> 00:07:12,960
So may just want to do, I like to do counties, but we could get more granular if we can get

83
00:07:12,960 --> 00:07:20,280
as granular as we would like.

84
00:07:20,280 --> 00:07:29,160
And so the next step would just be to calculate, you know, failure rates of the variable of

85
00:07:29,160 --> 00:07:30,680
interest that we're interested in.

86
00:07:30,680 --> 00:07:50,360
So, so the main things that you could test for failing would be the microbials.

87
00:07:50,360 --> 00:07:54,120
And like you said, those may have a real low failure rate.

88
00:07:54,120 --> 00:08:01,560
And then really the only other thing in Washington State is the residual solvents, right?

89
00:08:01,560 --> 00:08:10,400
Because pesticides and heavy metals aren't, there's not regular testing on that, on those

90
00:08:10,400 --> 00:08:12,120
compounds.

91
00:08:12,120 --> 00:08:20,720
So do you think like residual solvents like from processors would vary from location to

92
00:08:20,720 --> 00:08:22,800
location or?

93
00:08:22,800 --> 00:08:33,400
It might, I mean, isn't that part of the problem is that some producers just, you know, heavily,

94
00:08:33,400 --> 00:08:38,240
you know, like heavily spray their crops and others don't.

95
00:08:38,240 --> 00:08:48,960
And you know, I mean, like, yeah, I mean, it's certainly worth investigating to see.

96
00:08:48,960 --> 00:08:59,720
Yeah, so the residual solvents will be like your butanes, your propanes, and other extraction

97
00:08:59,720 --> 00:09:02,840
compounds.

98
00:09:02,840 --> 00:09:11,560
So yeah, I mean, it could be, you know, the type of equipment that they use to do the

99
00:09:11,560 --> 00:09:16,160
extraction, how experienced they are.

100
00:09:16,160 --> 00:09:17,600
Here's actually a good thought.

101
00:09:17,600 --> 00:09:26,720
So at Canacon, right, the, they had a wide range of equipment, right?

102
00:09:26,720 --> 00:09:38,400
So if you're, when you do processing, you have, you can invest quite a bit in extraction

103
00:09:38,400 --> 00:09:39,480
equipment.

104
00:09:39,480 --> 00:09:50,480
So once you've done your, you know, your propane or butane or your hydrocarbon extraction,

105
00:09:50,480 --> 00:09:58,720
you can, one, you can invest a lot in an expensive extraction equipment, and then you're going

106
00:09:58,720 --> 00:10:03,760
to need to essentially, you know, evaporate off all the hydrocarbons.

107
00:10:03,760 --> 00:10:10,120
And so, you know, people have different budgets, right?

108
00:10:10,120 --> 00:10:18,280
So a smaller processor is going to use, you know, less efficient equipment than, you know,

109
00:10:18,280 --> 00:10:20,840
a large scale processor.

110
00:10:20,840 --> 00:10:28,520
So process, so large scale processors may be more effective, you know, they may be more

111
00:10:28,520 --> 00:10:37,080
effective at removing unwanted hydrocarbons from their products.

112
00:10:37,080 --> 00:10:39,960
And so they may have a lower failure rate.

113
00:10:39,960 --> 00:10:49,000
And so that would actually be worthwhile, you know, knowing.

114
00:10:49,000 --> 00:10:57,580
So, you know, if you like, if you scale, does your failure rate go down as well?

115
00:10:57,580 --> 00:11:01,560
Is there, you know, scaling for cost saving reasons?

116
00:11:01,560 --> 00:11:05,080
Well, you know, economies of scale, essentially.

117
00:11:05,080 --> 00:11:10,960
And so it'd be interesting to see if that applies to failure rates.

118
00:11:10,960 --> 00:11:20,000
Yeah, that would be interesting to see.

119
00:11:20,000 --> 00:11:23,560
That would be, you know, that's the kind of thing I've been trying to figure out, like

120
00:11:23,560 --> 00:11:28,680
what kind of predictive machine learning type things can we do with this data?

121
00:11:28,680 --> 00:11:30,200
And that would be an interesting thing.

122
00:11:30,200 --> 00:11:41,120
And I've seen, there was like, I think it was a story on the news about people doing

123
00:11:41,120 --> 00:11:48,680
extractions, you know, and like, having small operations out of their garage and stuff.

124
00:11:48,680 --> 00:11:55,160
So yeah, I wonder, you know, I would imagine that the type of equipment that you use has

125
00:11:55,160 --> 00:11:56,560
some effect on it.

126
00:11:56,560 --> 00:12:01,080
And hopefully, if you're spending more money on equipment, you're getting higher quality

127
00:12:01,080 --> 00:12:04,120
product out of it.

128
00:12:04,120 --> 00:12:06,960
Well, exactly.

129
00:12:06,960 --> 00:12:13,920
And that's what they're saying is, you know, you're getting a more pure product.

130
00:12:13,920 --> 00:12:22,080
And so one would hypothesize that you would have a, you know, a lower failure rate.

131
00:12:22,080 --> 00:12:27,500
We can potentially quantify.

132
00:12:27,500 --> 00:12:33,380
So the way I think you can potentially measure the processors.

133
00:12:33,380 --> 00:12:39,560
So there may just be a one size fits all processing license in Washington state.

134
00:12:39,560 --> 00:12:53,120
So what if you, you know, proxy the size or the amount that they spend on their equipment

135
00:12:53,120 --> 00:12:54,840
by their sales.

136
00:12:54,840 --> 00:13:09,200
So you would just look at it, say, total sales by processor, say by month.

137
00:13:09,200 --> 00:13:17,760
And then look at the failure rate for that processor for that month and see if those

138
00:13:17,760 --> 00:13:20,960
are correlated.

139
00:13:20,960 --> 00:13:30,680
So you'd run a regression of the failure rate on total sales.

140
00:13:30,680 --> 00:13:39,840
So one would think that the larger processors have larger sales, thus they can spend more

141
00:13:39,840 --> 00:13:42,360
on their equipment.

142
00:13:42,360 --> 00:13:49,320
And we would hypothesize have a lower failure rate.

143
00:13:49,320 --> 00:13:53,320
Yeah.

144
00:13:53,320 --> 00:14:03,600
So I think we just discovered a research project.

145
00:14:03,600 --> 00:14:09,320
Maybe a little large to embark on today, but I think that's another one that we can chalk

146
00:14:09,320 --> 00:14:17,880
onto the chalkboard and potentially look at as early as next week.

147
00:14:17,880 --> 00:14:21,320
Yeah.

148
00:14:21,320 --> 00:14:27,040
Well, I guess just go ahead.

149
00:14:27,040 --> 00:14:33,720
I guess we can go ahead and break into a little bit of work that I've done here recently and

150
00:14:33,720 --> 00:14:37,640
I can present some of that if you would like.

151
00:14:37,640 --> 00:14:38,640
So

152
00:14:38,640 --> 00:14:51,960
So can you give me a second?

153
00:14:51,960 --> 00:14:57,880
So long story short.

154
00:14:57,880 --> 00:15:06,080
Canlytics has put together a map here where we're trying to map the cannabis testing licenses

155
00:15:06,080 --> 00:15:16,000
in the United States, potentially the world as well, but for now the United States.

156
00:15:16,000 --> 00:15:21,600
And so just to get some on the map over here.

157
00:15:21,600 --> 00:15:32,640
Just been introducing Canlytics to various laboratories here in Oklahoma City and then

158
00:15:32,640 --> 00:15:48,200
popped over to Tulsa, a handful of laboratories here.

159
00:15:48,200 --> 00:15:59,840
I'm actually in Grove right now and I'll stop by Genesis testing just to say hey.

160
00:15:59,840 --> 00:16:06,920
So just to get a look at the Oklahoma industry here.

161
00:16:06,920 --> 00:16:10,800
So Oklahoma does have interesting data.

162
00:16:10,800 --> 00:16:21,840
So stories.opengov.com, Oklahoma State.

163
00:16:21,840 --> 00:16:26,440
They just have snapshots of their revenue.

164
00:16:26,440 --> 00:16:49,880
However, you know, and they also have

165
00:16:49,880 --> 00:16:57,400
a list of their licensed businesses so you can see the licensed growers, processors,

166
00:16:57,400 --> 00:17:05,240
dispensaries, laboratories in the state.

167
00:17:05,240 --> 00:17:11,080
So here we have PDFs.

168
00:17:11,080 --> 00:17:21,640
And so we would like this data in a more readily consumable format.

169
00:17:21,640 --> 00:17:24,120
And there may be others as well.

170
00:17:24,120 --> 00:17:28,560
So Canlytics, we're trying to make data accessible.

171
00:17:28,560 --> 00:17:37,960
So today we're going to make this data accessible.

172
00:17:37,960 --> 00:17:48,560
So just to kind of show you some of the work I've done here.

173
00:17:48,560 --> 00:17:55,160
So just using a handful of tools.

174
00:17:55,160 --> 00:18:06,360
We'll basically go and we'll download the PDFs.

175
00:18:06,360 --> 00:18:12,200
So we'll basically download each of these.

176
00:18:12,200 --> 00:18:23,000
And then we'll want to parse them just to get them into a consumable format.

177
00:18:23,000 --> 00:18:29,480
And then I just do a little cleaning of the names.

178
00:18:29,480 --> 00:18:34,920
Just a lowercase and snake case.

179
00:18:34,920 --> 00:19:04,600
Oh, I've committed this code to the Manifest Data Science Repository.

180
00:19:04,600 --> 00:19:18,200
So you can follow along and really read into the script here when you really have time

181
00:19:18,200 --> 00:19:21,080
to delve into that.

182
00:19:21,080 --> 00:19:28,600
And then if you just want to grab the data, then we have it already created and uploaded

183
00:19:28,600 --> 00:19:34,760
for you.

184
00:19:34,760 --> 00:19:56,560
Alright, so let's look at the data.

185
00:19:56,560 --> 00:20:01,600
As most states are moving in this direction just towards a little bit of privacy for the

186
00:20:01,600 --> 00:20:11,640
cultivators and processors, your geographic location is at the county level.

187
00:20:11,640 --> 00:20:16,720
So you don't really get much granularity than that, and so that's to respect the privacy

188
00:20:16,720 --> 00:20:18,920
of these businesses.

189
00:20:18,920 --> 00:20:30,320
However, for a lot of our analysis, the county level or even the zip code is perfectly sufficient

190
00:20:30,320 --> 00:20:35,800
for our analysis.

191
00:20:35,800 --> 00:20:53,920
And just to show you the scale of these, just in Oklahoma, there are some north of 11,000

192
00:20:53,920 --> 00:20:54,920
licensees.

193
00:20:54,920 --> 00:21:03,880
And this would include your transporters, your library, your laboratories, your waste

194
00:21:03,880 --> 00:21:14,000
disposal companies.

195
00:21:14,000 --> 00:21:19,560
It's first glance striving, or thriving.

196
00:21:19,560 --> 00:21:28,320
So we really want to try to quantify this a bit more because all we really know is just

197
00:21:28,320 --> 00:21:31,160
the number of licensees.

198
00:21:31,160 --> 00:21:35,320
But these people may have just licensed.

199
00:21:35,320 --> 00:21:38,080
They may not be operational.

200
00:21:38,080 --> 00:21:47,600
So we want to get a bit better look at the industry.

201
00:21:47,600 --> 00:22:06,960
So the next data point we can get is essentially the revenue.

202
00:22:06,960 --> 00:22:19,360
You can download this as a CSV.

203
00:22:19,360 --> 00:22:24,360
So I've downloaded that.

204
00:22:24,360 --> 00:22:33,560
And once again, we need to parse this appropriately.

205
00:22:33,560 --> 00:22:43,880
So one thing that I began to note, so you'll see this here in a second.

206
00:22:43,880 --> 00:22:52,560
So let's go ahead and read in the revenue.

207
00:22:52,560 --> 00:23:12,880
And once again, just parse the data just to get it into your format that we can use.

208
00:23:12,880 --> 00:23:20,080
All right.

209
00:23:20,080 --> 00:23:37,520
So does anything jump out at you, Charles?

210
00:23:37,520 --> 00:23:42,480
It's linear and increasing.

211
00:23:42,480 --> 00:23:44,560
Yes.

212
00:23:44,560 --> 00:23:49,640
Almost perfectly.

213
00:23:49,640 --> 00:23:54,320
Do you have much faith in that data?

214
00:23:54,320 --> 00:23:55,320
No.

215
00:23:55,320 --> 00:24:07,880
So that was something that's a little concerning to me is it's one would say how linear.

216
00:24:07,880 --> 00:24:19,600
And so that's actually what I actually did next was I so here I crossed on just a, you

217
00:24:19,600 --> 00:24:24,880
know, time index.

218
00:24:24,880 --> 00:24:29,560
That way, you know, we just have a counter.

219
00:24:29,560 --> 00:24:37,280
Then we can just calculate just a quick trend.

220
00:24:37,280 --> 00:24:47,240
This is basically just running a regression of total revenue on time.

221
00:24:47,240 --> 00:24:52,960
You know, where, you know, t equals zero.

222
00:24:52,960 --> 00:24:55,480
Right.

223
00:24:55,480 --> 00:25:15,800
Then, you know, we'll just, you know, plot the trend with total revenue.

224
00:25:15,800 --> 00:25:42,080
And so, you know, we essentially have, you know, just a perfectly linear line.

225
00:25:42,080 --> 00:25:51,000
So we just kind of, you know, now I'm just trying to think about like how to explain

226
00:25:51,000 --> 00:25:52,000
this.

227
00:25:52,000 --> 00:26:02,360
So, so honestly, the first things that came to mind is maybe the software developer just

228
00:26:02,360 --> 00:26:09,200
put in placeholders, but that just doesn't seem likely.

229
00:26:09,200 --> 00:26:25,600
Then I'm thinking, is maybe that's just what they're accounting for, or maybe they're

230
00:26:25,600 --> 00:26:28,840
only taxing a certain amount per month.

231
00:26:28,840 --> 00:26:30,160
It's hard to explain.

232
00:26:30,160 --> 00:26:37,960
So do you have any thoughts on this Charles or what are we looking at?

233
00:26:37,960 --> 00:26:42,320
No, it looks suspicious.

234
00:26:42,320 --> 00:26:58,840
But maybe that's just like how they're accounting for it.

235
00:26:58,840 --> 00:27:04,880
Like they're just like, they just, the licensees just submitted the year total and they're

236
00:27:04,880 --> 00:27:10,640
just like accruing it per month, but it doesn't really explain like how they have a through

237
00:27:10,640 --> 00:27:24,360
or am I just, or am I interpreting this data wrong?

238
00:27:24,360 --> 00:27:51,800
Are maybe more and more people getting into the system legally?

239
00:27:51,800 --> 00:28:01,200
Or have they loosened their requirements on medical marijuana?

240
00:28:01,200 --> 00:28:06,720
Let's just calculate the, because it looks like there's about a, what is it, seven million

241
00:28:06,720 --> 00:28:07,720
different.

242
00:28:07,720 --> 00:28:10,640
So let's see.

243
00:28:10,640 --> 00:28:16,640
Because basically I was wanting to start doing the statistics on this data, but like I said,

244
00:28:16,640 --> 00:28:20,640
it just doesn't look, well, it just, I don't know.

245
00:28:20,640 --> 00:28:22,240
I just need some explanation.

246
00:28:22,240 --> 00:28:24,600
Like I'm just not certain what's going on.

247
00:28:24,600 --> 00:28:27,220
So it's just.

248
00:28:27,220 --> 00:28:33,240
So it looked like it was like, just when you were hovering over it, there was like 11, 12,

249
00:28:33,240 --> 00:28:35,800
13% growth.

250
00:28:35,800 --> 00:28:44,520
And so all the, every month grew, you know, by roughly the same amount.

251
00:28:44,520 --> 00:29:09,320
Unless, unless this is just like.

252
00:29:09,320 --> 00:29:18,120
Or are we looking at something weird?

253
00:29:18,120 --> 00:29:20,960
Like are we just looking at like the aggregate or something?

254
00:29:20,960 --> 00:29:26,240
And maybe like the differences, what's going on?

255
00:29:26,240 --> 00:29:42,880
Because you see, it's not like, this is strange.

256
00:29:42,880 --> 00:29:57,440
Well, why don't we write a quick for loop.

257
00:29:57,440 --> 00:30:13,320
Yeah.

258
00:30:13,320 --> 00:30:42,040
What's going on?

259
00:30:42,040 --> 00:30:56,560
What's going on?

260
00:30:56,560 --> 00:30:59,840
Thank you for bearing with me.

261
00:30:59,840 --> 00:31:05,560
The tortoise always beats the hare.

262
00:31:05,560 --> 00:31:26,560
So it's not the same every month.

263
00:31:26,560 --> 00:31:49,480
This is interesting.

264
00:31:49,480 --> 00:31:55,680
So let's just try to capture these values and analyze the difference here for a second.

265
00:31:55,680 --> 00:32:09,920
So let's just look at this.

266
00:32:09,920 --> 00:32:18,200
That's what you would sort of expect.

267
00:32:18,200 --> 00:32:24,680
Like a revenue plot would look a bit more, doesn't that look a bit more reasonable?

268
00:32:24,680 --> 00:32:25,680
Yeah.

269
00:32:25,680 --> 00:32:36,000
So do you think this is just year to date?

270
00:32:36,000 --> 00:32:39,760
Is that what we're looking at here?

271
00:32:39,760 --> 00:32:43,000
Maybe.

272
00:32:43,000 --> 00:32:47,240
When did medical marijuana become legal in Oklahoma?

273
00:32:47,240 --> 00:32:55,200
It looks like sales began essentially in July of 2020.

274
00:32:55,200 --> 00:33:02,720
I want to say January of 2020, but it looks like we just have data from July.

275
00:33:02,720 --> 00:33:10,200
I mean, this just could be like a ramping up as people get into the system and start

276
00:33:10,200 --> 00:33:13,720
moving from like, you know, the black market to the legal market.

277
00:33:13,720 --> 00:33:14,720
Interesting.

278
00:33:14,720 --> 00:33:24,520
Let's do this.

279
00:33:24,520 --> 00:33:31,400
So I think that may just be.

280
00:33:31,400 --> 00:33:46,680
Though I think this is not necessarily what we want.

281
00:33:46,680 --> 00:33:49,360
Let's go back to.

282
00:33:49,360 --> 00:33:58,200
And keep in mind, this is the tax, not the revenue.

283
00:33:58,200 --> 00:34:01,520
So okay, so Charles, this is not what I'm thinking.

284
00:34:01,520 --> 00:34:08,160
I'm thinking this is just a running.

285
00:34:08,160 --> 00:34:15,040
Oh, yeah.

286
00:34:15,040 --> 00:34:16,640
Right.

287
00:34:16,640 --> 00:34:23,480
Because there's actual, then there's total.

288
00:34:23,480 --> 00:34:30,160
The actual and the total are always the same.

289
00:34:30,160 --> 00:34:40,080
Yes, I think this is just a.

290
00:34:40,080 --> 00:34:44,000
You know, you know, cumulative.

291
00:34:44,000 --> 00:34:46,440
You know, the cumulative graph.

292
00:34:46,440 --> 00:34:57,280
So August includes July, September includes July and.

293
00:34:57,280 --> 00:34:59,240
August.

294
00:34:59,240 --> 00:35:06,920
And so then if you look at the difference, then.

295
00:35:06,920 --> 00:35:13,680
You know, this is essentially, you know, your change in.

296
00:35:13,680 --> 00:35:19,240
Tax size tax per month, so this would be the taxable.

297
00:35:19,240 --> 00:35:25,440
This would represent the taxable revenue.

298
00:35:25,440 --> 00:35:31,000
If I'm interpreting this correctly, and so it may be worth reaching out to.

299
00:35:31,000 --> 00:35:35,200
The OMA.

300
00:35:35,200 --> 00:35:37,480
This body, the Oklahoma Medical.

301
00:35:37,480 --> 00:35:46,000
Just to to clarify that that is what we are looking at.

302
00:35:46,000 --> 00:35:54,120
So we can actually.

303
00:35:54,120 --> 00:36:02,000
Perhaps adjust our code real quick.

304
00:36:02,000 --> 00:36:19,800
To account for this.

305
00:36:19,800 --> 00:36:38,320
Pandas just do a simple.

306
00:36:38,320 --> 00:36:40,480
Excellent so pain.

307
00:36:40,480 --> 00:36:43,480
This is just a built in.

308
00:36:43,480 --> 00:36:48,320
If that we can use so.

309
00:36:48,320 --> 00:36:54,120
It does not.

310
00:36:54,120 --> 00:37:02,360
We can't explain this initial five billion.

311
00:37:02,360 --> 00:37:07,360
Because.

312
00:37:07,360 --> 00:37:22,160
Well, let's see what the mean is over here.

313
00:37:22,160 --> 00:37:39,360
Or mean.

314
00:37:39,360 --> 00:37:59,240
Where was our initial.

315
00:37:59,240 --> 00:38:19,080
Let's look at some of the statistics that they've produced.

316
00:38:19,080 --> 00:38:32,040
So they're saying that in March.

317
00:38:32,040 --> 00:38:55,520
We're not certain the year that almost one million came from.

318
00:38:55,520 --> 00:39:02,280
It's really hard to this way this data really needs some exploration because it's.

319
00:39:02,280 --> 00:39:15,960
Doesn't it seem like their industry is like staggeringly large.

320
00:39:15,960 --> 00:39:22,320
What's the like.

321
00:39:22,320 --> 00:39:23,320
Hold on.

322
00:39:23,320 --> 00:39:34,840
I was calculating the total revenue.

323
00:39:34,840 --> 00:39:59,440
I was applying the tax rate, but we need to do the change first.

324
00:39:59,440 --> 00:40:15,000
Okay, so let's try.

325
00:40:15,000 --> 00:40:39,400
It's just a fact that it's significant.

326
00:40:39,400 --> 00:40:52,760
Okay, so let's see.

327
00:40:52,760 --> 00:40:56,720
Okay, so that does look about right.

328
00:40:56,720 --> 00:41:10,440
So it looks like they're getting about 5 million in excise tax per month.

329
00:41:10,440 --> 00:41:14,040
That sounds a bit more reasonable, right?

330
00:41:14,040 --> 00:41:15,040
Yes.

331
00:41:15,040 --> 00:41:35,720
Let's see when this was published.

332
00:41:35,720 --> 00:41:38,560
This was published in 2019.

333
00:41:38,560 --> 00:41:47,960
So in 2019, they were getting about 1 million, a little less than 1 million with their excise

334
00:41:47,960 --> 00:41:48,960
tax.

335
00:41:48,960 --> 00:42:02,960
So it appears that by July of 2020, when they have this data published, it looks like the

336
00:42:02,960 --> 00:42:07,040
first data point is about 5 million per month.

337
00:42:07,040 --> 00:42:12,160
And then it looks like they're increasing about 5 million each month.

338
00:42:12,160 --> 00:42:16,760
So this must be the cumulative.

339
00:42:16,760 --> 00:42:19,840
Right.

340
00:42:19,840 --> 00:42:34,760
Okay, and so then we can actually add on this first data point here.

341
00:42:34,760 --> 00:42:41,480
Just your little hotfix here.

342
00:42:41,480 --> 00:42:51,160
Just to go ahead and get the first data point back.

343
00:42:51,160 --> 00:43:06,440
See how can we...

344
00:43:06,440 --> 00:43:21,440
Can we just add... essentially just want to get that first data point back.

345
00:43:21,440 --> 00:43:36,940
Because I never want to throw away...

346
00:43:36,940 --> 00:43:52,600
Something I'm using...

347
00:43:52,600 --> 00:44:08,520
Okay, so I didn't use that correctly.

348
00:44:08,520 --> 00:44:20,800
Do you know off the top of your head how to sign the value?

349
00:44:20,800 --> 00:44:28,320
Don't you want to use like iLock or lock?

350
00:44:28,320 --> 00:44:31,200
Potentially.

351
00:44:31,200 --> 00:44:38,520
So basically we want to fill in this first revenue observation.

352
00:44:38,520 --> 00:44:40,200
Right.

353
00:44:40,200 --> 00:44:44,480
I mean, it's something like...

354
00:44:44,480 --> 00:44:49,360
Can we do that?

355
00:44:49,360 --> 00:44:53,320
I think so.

356
00:44:53,320 --> 00:45:02,880
Well, maybe we should give a suggestion.

357
00:45:02,880 --> 00:45:11,080
Yeah, you're just sending it on a copy, which is not...

358
00:45:11,080 --> 00:45:16,200
So I don't think that's the best practice.

359
00:45:16,200 --> 00:45:20,480
You just need to say cop.copy.

360
00:45:20,480 --> 00:45:27,480
Nonetheless, for now...

361
00:45:27,480 --> 00:45:28,480
You're safe.

362
00:45:28,480 --> 00:45:32,040
You're not going to do anything...

363
00:45:32,040 --> 00:45:36,720
We'll serve our purpose well.

364
00:45:36,720 --> 00:45:37,840
Let's address this later.

365
00:45:37,840 --> 00:45:46,680
So we'll just look at it from August onwards, just so we can expedite things.

366
00:45:46,680 --> 00:45:52,200
We'll come back and address this.

367
00:45:52,200 --> 00:46:02,920
Okay, so does this look a bit more reasonable for revenue?

368
00:46:02,920 --> 00:46:08,040
It does not.

369
00:46:08,040 --> 00:46:22,680
Well, it actually does not sound very good.

370
00:46:22,680 --> 00:46:27,520
They're getting about 77 million and the excise tax is 7%.

371
00:46:27,520 --> 00:46:31,000
Yeah, so that's about right.

372
00:46:31,000 --> 00:46:34,280
So it does...

373
00:46:34,280 --> 00:46:43,800
I mean, in my opinion, it's a staggering amount of sales, right?

374
00:46:43,800 --> 00:46:56,760
You're doing 77 million in sales in August, then it dips down towards December, which

375
00:46:56,760 --> 00:47:05,760
is an interesting observation that it looks like it's inversely related with Christmas

376
00:47:05,760 --> 00:47:08,400
shopping.

377
00:47:08,400 --> 00:47:16,480
And then you just see this nice...

378
00:47:16,480 --> 00:47:33,960
You see quite a steady rise to what's this, 90 million already in May.

379
00:47:33,960 --> 00:47:41,480
So you have 89 million in May.

380
00:47:41,480 --> 00:47:46,880
You're on...

381
00:47:46,880 --> 00:47:47,880
Let's see.

382
00:47:47,880 --> 00:47:55,920
So let's calculate how much revenue they've raised this year in 2021.

383
00:47:55,920 --> 00:48:14,400
So we can say, okay, we want revenue data.

384
00:48:14,400 --> 00:48:15,560
Let's just do this quickly.

385
00:48:15,560 --> 00:48:27,080
So maybe we can just say PD.

386
00:48:27,080 --> 00:48:46,320
Let's see, 2021... right, so that worked.

387
00:48:46,320 --> 00:48:47,320
So there...

388
00:48:47,320 --> 00:48:58,600
So if you... we just want to look at revenue so far for 2021, fairly steady increase.

389
00:48:58,600 --> 00:49:02,400
And so let's go ahead and...

390
00:49:02,400 --> 00:49:25,320
So in the first five months, Oklahoma's grossed almost 400 million in cannabis sales.

391
00:49:25,320 --> 00:49:37,320
So I think Colorado is about a billion or perhaps even $2 billion industry.

392
00:49:37,320 --> 00:49:52,440
And so here, Oklahoma, five months has done 400 million.

393
00:49:52,440 --> 00:49:56,600
So let's just say...

394
00:49:56,600 --> 00:50:09,920
So one could expect if they keep on this trajectory, that it wouldn't be surprising to see Oklahoma

395
00:50:09,920 --> 00:50:16,360
gross one billion in cannabis sales this year.

396
00:50:16,360 --> 00:50:23,360
So that's...

397
00:50:23,360 --> 00:50:47,040
So that's 70 million in excise tax if I'm doing the math correctly.

398
00:50:47,040 --> 00:50:48,040
So that sounds about right.

399
00:50:48,040 --> 00:50:53,800
So they're doing a little more than five million a month.

400
00:50:53,800 --> 00:50:56,160
So they're getting about $70 million.

401
00:50:56,160 --> 00:51:00,920
And I mean, that's not chump change.

402
00:51:00,920 --> 00:51:01,920
So...

403
00:51:01,920 --> 00:51:02,920
No.

404
00:51:02,920 --> 00:51:07,520
I mean, well, this is why everybody's interested in this industry, right?

405
00:51:07,520 --> 00:51:09,920
It's going to...

406
00:51:09,920 --> 00:51:11,960
It generates a lot of money.

407
00:51:11,960 --> 00:51:17,560
Well, let's just money left on the table.

408
00:51:17,560 --> 00:51:35,560
If you're not doing it, in fact, not only are people getting benefits from the medicinal

409
00:51:35,560 --> 00:51:36,560
consumption, right?

410
00:51:36,560 --> 00:51:39,560
So Oklahoma, this is actually all medicinal sales.

411
00:51:39,560 --> 00:51:46,240
So these are all people who are using cannabis for medicinal purposes.

412
00:51:46,240 --> 00:51:50,360
So not only are they getting that benefit, but yes, the state's also...

413
00:51:50,360 --> 00:51:54,360
70 million, that's like...

414
00:51:54,360 --> 00:52:06,880
It could easily be like 35 schools or however they want to spend that money.

415
00:52:06,880 --> 00:52:14,800
That's hopefully going to have quite a positive impact.

416
00:52:14,800 --> 00:52:20,840
So I think we've made a little sense of the data here.

417
00:52:20,840 --> 00:52:36,040
And I think if we have time in the last little bit here, the interesting measure is we could

418
00:52:36,040 --> 00:52:46,720
potentially look at the average revenue per retailer or grower.

419
00:52:46,720 --> 00:52:54,320
So that way we can kind of get a measure of how large these businesses are.

420
00:52:54,320 --> 00:53:07,560
So let's just do a dispensary account essentially.

421
00:53:07,560 --> 00:53:17,280
So we're going to have to do this quite quickly, Charles, but that's why we're just crunching

422
00:53:17,280 --> 00:53:18,280
numbers.

423
00:53:18,280 --> 00:53:25,880
That's what we're all about here.

424
00:53:25,880 --> 00:53:44,320
Let's quickly find out with your call and then isolate them.

425
00:53:44,320 --> 00:53:47,960
Then count them.

426
00:53:47,960 --> 00:53:50,440
There's about 2000.

427
00:53:50,440 --> 00:54:00,200
And then keep in mind that these were probably...

428
00:54:00,200 --> 00:54:10,640
These probably fluctuated throughout the year, but we can use a proxy here.

429
00:54:10,640 --> 00:54:13,240
So we can say...

430
00:54:13,240 --> 00:54:23,240
Well not really a proxy, but an estimate, so we can just say estimated revenue per dispensary

431
00:54:23,240 --> 00:54:27,040
per month.

432
00:54:27,040 --> 00:54:53,840
So that would just be our revenue divided by the number of dispensaries.

433
00:54:53,840 --> 00:54:58,000
And then let's see if this makes sense at all.

434
00:54:58,000 --> 00:55:14,400
Here's when you need to define everything.

435
00:55:14,400 --> 00:55:19,480
And then this is yet another positive impact on the community, right?

436
00:55:19,480 --> 00:55:25,760
Because you've got your patients getting their medicinal benefit, you have the government

437
00:55:25,760 --> 00:55:31,600
getting their revenue, which they can distribute to the public through public goods like education,

438
00:55:31,600 --> 00:55:32,600
roads.

439
00:55:32,600 --> 00:55:38,080
You do see a lot of road construction in Oklahoma, which is a good thing to see, right?

440
00:55:38,080 --> 00:55:43,400
They're fixing up their roads big time.

441
00:55:43,400 --> 00:55:50,920
And then you look at the dispensaries and look at this.

442
00:55:50,920 --> 00:55:53,920
They're grossing...

443
00:55:53,920 --> 00:56:02,760
Look at the mean here.

444
00:56:02,760 --> 00:56:21,000
The average is a little north of 34,000 a month for dispensary.

445
00:56:21,000 --> 00:56:26,640
People who are opening up a cannabis dispensary, which requires a little bit of overhead, but

446
00:56:26,640 --> 00:56:31,080
not that much more than opening up a restaurant.

447
00:56:31,080 --> 00:56:38,760
They're going to be less and they can easily gross half a million dollars a year.

448
00:56:38,760 --> 00:56:44,280
That's a good sized business.

449
00:56:44,280 --> 00:56:50,000
So there's 2000 of them, right?

450
00:56:50,000 --> 00:56:56,320
I'm sure they're not all making the same revenue, but right?

451
00:56:56,320 --> 00:57:06,640
So overnight, Oklahoma turned on the switch and now you have 2200 businesses making an

452
00:57:06,640 --> 00:57:13,120
average of $400,000 a year.

453
00:57:13,120 --> 00:57:15,080
That's phenomenal.

454
00:57:15,080 --> 00:57:16,080
That's good for them.

455
00:57:16,080 --> 00:57:23,400
That's good for their employees, the people that work for them.

456
00:57:23,400 --> 00:57:26,760
You have a large multiplier effect from this, right?

457
00:57:26,760 --> 00:57:29,640
Because when a business opens, right?

458
00:57:29,640 --> 00:57:31,440
They need an accountant.

459
00:57:31,440 --> 00:57:34,440
They need to make signage.

460
00:57:34,440 --> 00:57:37,600
They need to do advertising.

461
00:57:37,600 --> 00:57:45,500
There's a huge ripple effect throughout the economy where these companies are now utilizing

462
00:57:45,500 --> 00:57:47,240
other companies.

463
00:57:47,240 --> 00:57:59,280
So I think I'll leave it there for today, but it's just phenomenal to see.

464
00:57:59,280 --> 00:58:02,520
Any comments, Charles?

465
00:58:02,520 --> 00:58:06,840
No, it's great.

466
00:58:06,840 --> 00:58:14,720
They have to hire people who spend money around the community.

467
00:58:14,720 --> 00:58:26,200
It might be interesting to look like sales by county or sales by zip code.

468
00:58:26,200 --> 00:58:30,300
So Oklahoma does have a public records request.

469
00:58:30,300 --> 00:58:37,240
So similar to Washington state, you can make public records requests to OMA.

470
00:58:37,240 --> 00:58:42,320
And so it would be worthwhile asking.

471
00:58:42,320 --> 00:58:46,960
So they just have their dashboard, which just has the totals.

472
00:58:46,960 --> 00:58:52,680
So I think it is worthwhile to ask what other data is public?

473
00:58:52,680 --> 00:58:58,200
So can we get the sales by county, so to speak?

474
00:58:58,200 --> 00:59:05,360
Even if we have to do the data wrangling and crunching ourselves, they can at least get

475
00:59:05,360 --> 00:59:07,360
us in the right direction.

476
00:59:07,360 --> 00:59:13,680
So you're right, because I think further analysis is needed, because as we've seen in Washington

477
00:59:13,680 --> 00:59:24,480
state, the distribution of licensees varies.

478
00:59:24,480 --> 00:59:33,320
So you've got your concentrated retailers along the Puget Sound, and then you've got

479
00:59:33,320 --> 00:59:40,000
the cultivation heavy Easternors.

480
00:59:40,000 --> 00:59:45,920
We've got several fruitful avenues to pursue for next week.

481
00:59:45,920 --> 00:59:52,480
So we've got failure rates to look at.

482
00:59:52,480 --> 00:59:56,360
We've got geographic location to look at.

483
00:59:56,360 --> 01:00:00,240
And there was one at the beginning that we'll have to review in the recording.

484
01:00:00,240 --> 01:00:04,200
So I think it's been a productive day.

485
01:00:04,200 --> 01:00:07,200
Yeah, that's good.

486
01:00:07,200 --> 01:00:12,200
It was a good meeting.

487
01:00:12,200 --> 01:00:19,640
So we were going to look at the processors to see if processor size affects failure rates.

488
01:00:19,640 --> 01:00:25,560
So those are the takeaways for today.

489
01:00:25,560 --> 01:00:27,760
Economies of scale are worth looking at.

490
01:00:27,760 --> 01:00:30,920
The geographic location is worth looking at.

491
01:00:30,920 --> 01:00:38,360
And as always, the number one rule is look at the data.

492
01:00:38,360 --> 01:00:42,200
Well, awesome, Charles.

493
01:00:42,200 --> 01:00:47,560
I think we'll leave it there for today and pick up next week.

494
01:00:47,560 --> 01:00:54,080
Hopefully we can get things cleared away as to the timing.

495
01:00:54,080 --> 01:00:58,000
And I look forward to crunching some numbers with you next week.

496
01:00:58,000 --> 01:00:59,000
Okay, great.

497
01:00:59,000 --> 01:01:01,760
Ask me to go to the home.

498
01:01:01,760 --> 01:01:02,760
I will.

499
01:01:02,760 --> 01:01:03,760
Get some good barbecue.

500
01:01:03,760 --> 01:01:05,960
There's really good barbecue out there.

501
01:01:05,960 --> 01:01:08,160
Ooh, I'll get some.

502
01:01:08,160 --> 01:01:13,480
And keep, well, you know, keep the secret secret.

503
01:01:13,480 --> 01:01:18,640
But, you know, definitely feel free to share anything that you can about what you've learned

504
01:01:18,640 --> 01:01:20,640
about machine learning.

505
01:01:20,640 --> 01:01:25,720
Because that's something, you know, we would love to learn more about as well.

506
01:01:25,720 --> 01:01:26,720
Okay.

507
01:01:26,720 --> 01:01:27,720
Awesome, Charles.

508
01:01:27,720 --> 01:01:33,240
Well, once again, have a fantastic day and we'll be in touch.

509
01:01:33,240 --> 01:01:34,240
Okay, see ya.

510
01:01:34,240 --> 01:01:35,240
Bye now.

511
01:01:35,240 --> 01:01:54,540
Bye now.